This Agents Unpacked series is about that shift. Not the hype version, not the science-fiction version, but the practical reality of what it means when an AI system can act, remember, and persist — when it can take a goal and work toward it, rather than waiting for you to type the next message.

If you have used ChatGPT, Claude, or Gemini to help with your work, you already know the issue: the answer comes back, it’s good, and then... the real work begins. This series is about what happens when the AI can do some of that real work too.

The Half-Done Feeling

You ask an AI chatbot to draft a project proposal. It gives you a solid one — well-structured, sensible, ready to polish. Then you close the chat window, and the proposal lives nowhere. The assistant doesn’t remember it. It doesn’t know where to file it, who needs to see it, or what happened the last time you wrote a similar proposal. If you come back tomorrow, you’re starting from scratch.

Or you ask it to research a topic. It gives you a good summary. But it didn’t check your existing notes first, it didn’t save what it found, it didn’t organise the sources, and it didn’t flag the gaps. The answer is useful. The process is incomplete.

This isn’t a criticism of chatbots. They do exactly what they were designed to do: you ask, they answer. The problem is that real work doesn’t stop at the answer. The proposal needs filing, the research needs organising, the plan needs tracking. The chatbot gave you a great starting point and then left you to do everything that comes after.

That gap between a useful answer and a finished job is where agentic AI enters the picture.

What Changes When AI Can Act

A chatbot is like a consultant you can phone. You describe the problem, they give you advice, and then you hang up and do the work yourself. Good advice (hopefully), but the consultant doesn’t pick up the phone again tomorrow and ask how it went.

An agent is different. It’s more like a capable colleague you’ve given a desk, a filing cabinet, and access to your systems. It has a workspace. It can remember what happened yesterday. It can read files, search the web, send messages, run calculations, ask you questions — not just talk about doing those things, but actually do them. And given a goal, it can work out the steps itself.

The difference sounds small. In practice it changes what you can delegate, what you can automate, and what you still need to do yourself.

The Loop Underneath

Every agent runs on an agentic loop. It sounds technical, but the pattern is surprisingly familiar:

1. Observe — What is the situation? What did the user ask? What information is already available?

2. Think — What needs to happen next? Is one step enough, or should this be broken into parts?

3. Act — Use a tool, look something up, write a file, send a message.

4. Repeat — Look at what happened, decide whether the job is done, and continue if it isn’t.

A chatbot usually observes your message, thinks once, and replies. That’s steps one through three, then stop. An agent goes further: after it acts, it looks at the result. Was that enough? Did the web search return useful information, or does it need to try again with different terms? Did the first draft cover everything, or are there gaps to fill? Is the job actually done, or is there more to do?

You might be thinking: chatbots already learned to use tools. Isn’t the agentic loop just that, plus one more step? Four items instead of three. But the gap between “can use a tool” and “decides whether to keep going” is not small. It is the difference between a system that performs a task and a system that pursues a goal. The first is impressive. The second changes what you can trust it to do unsupervised.

That continuation — the decision to check, adjust, and keep going — is what makes the loop matter. A single action might solve a simple request. But most real tasks aren’t simple — they need a sequence of steps, each one informed by the result of the last.

The loop is the architecture that turns language into work. Without it, you have a very clever answering system. With it, you have something that can move through a task, make intermediate decisions, recover from partial failures, and stop only when the job is actually done.

Stephen: I don’t get why the ‘Repeat’ step is needed? Wouldn’t the ‘Act’ provide the output I need?

A single action might solve a simple request. But most real tasks aren’t simple — they need a sequence of steps, each one informed by the result of the last. The ‘Act’ step does produce an output. But the output is not the same as the outcome.

After ‘Act’ runs, the agent looks at what happened: Was that enough? Did the web search return useful information, or does it need different terms? Did the first draft cover everything, or are there gaps to fill? Is the job actually done, or is there more to do?

That check — that ‘Repeat’ step — is what closes the gap between a technically complete action and a genuinely finished job. Without it, you have a system that acts and stops. With it, you have a system that works until the job is actually done.

Subscribe now

Tools Are the Hands

An agent isn’t just “a better language model.” It has capabilities — things it can actually do in the world. Those are called tools.

Some tools are built in: read files, write files, run commands, search the web, inspect images. Others are external: send emails, query databases, call APIs, trigger workflows. The specific tools vary by platform, but the principle is the same — tools are the bridge between thinking and doing.

Think of it like this: an LLM on its own is like a brilliant mind with no hands. Tools give it hands. The loop is what decides when and how to use them.

Keeping Track

An agent can also hold information across steps and sessions. It remembers what it has already tried, what worked, what you prefer, what is still outstanding. This is not a personality trait. It is the system keeping relevant context available over time — the same way you rely on a notebook or a project board when you are working on something complex.

Stephen: So when the agent remembers my preferences, that’s like what you’ve done to write this article. You learnt my learning style, read my writing (including my writing about technical writing), absorbed my preferences, so you can adapt how you explain things. Is that roughly right?

That is exactly right. What an agent does with memory is not mysterious. It is practical. The system can store what you told it, what it observed, what it tried, and what the result was. When you come back, it can pick up where it left off. When it is working on a long task, it can hold the overall goal in view while handling the individual steps. That continuity is what turns a series of disconnected exchanges into something that feels like a sustained conversation.

You do not have to repeat yourself. The agent remembers. That is the difference — not just remembering facts, but maintaining a thread.

A Concrete Example

Let’s make this less abstract.

Imagine you ask for help planning a short research trip to a city you haven’t visited before. You need flights, accommodation, a sense of the neighbourhoods, and a rough itinerary that fits your schedule.

A chatbot might give you an excellent summary of the city, suggest some hotels, and recommend a few neighbourhoods. That’s genuinely useful. But then you have to: check whether those hotels are actually available on your dates, compare prices, figure out which neighbourhood works best for your meetings, build the itinerary around your existing calendar, and keep track of it all so you can adjust later.

An agent can take that same request and do something different.

It might:

• Check your calendar for available dates before suggesting anything

• Search for flights and filter by your preferred departure times

• Cross-reference hotel locations against the addresses you need to visit

• Build a day-by-day itinerary that accounts for travel time between meetings

• Save the whole plan somewhere you can find it and update it later

• Flag the gaps — “I found flights and hotels, but you haven’t told me whether you need a visa”

• Filter results by your stored preferences — early-morning flights, boutique hotels

• Use your past travel patterns to anticipate the kind of trip that fits your style

The agent doesn’t just tell you about the city. It assembles a usable, integrated plan. It uses tools to search, compare, read your calendar, write the plan, and flag what is missing. It loops through those actions until the trip is actually planned or until it hits something it cannot resolve without your input.

That is the difference between an answer and an outcome.

The Key Transition

The shift to understanding agents is not about capabilities. It is about the move from isolated exchanges to sustained work.

A chatbot gives you an answer. An agent helps you reach an outcome. One is a single exchange; the other is a process. One is clever; the other is useful in a different way — not smarter, but more continuous.

The loop is what makes that continuity possible. Observe, think, act, check what happened, adjust, and continue until the work is done.

Once you see that pattern, you start to notice it everywhere. A junior colleague troubleshooting a problem is running a loop: try something, see if it worked, try something else. A project manager steering a complex task is running a loop: check the status, identify what needs attention, act, and review.
The same pattern appears everywhere in life. A cook adjusts a recipe by tasting as they go. A teacher tries an explanation, sees whether the student understood, and tries a different approach if they didn’t.

Stephen: So, tell me if I got this right: The agent loop is mimicking how we, humans, work.

We understand the problem, get the relevant context, perform an action, look at the result of our action and then decide whether that solves our problem. If it doesn’t, we explore why, come up with a new plan, implement the new actions, and repeat the process.

It feels like the agent is going through the same process.

That is exactly right. The agent loop isn’t some exotic new form of intelligence. It’s a pattern humans use all day every day, made explicit and embedded in a system that can act. The insight isn’t that the AI has become more intelligent. It’s that the AI has gained the ability to persist, to use tools, and to continue — the same things that turn a one-off answer into real, completed work.

That is why this moment in AI feels different from previous ones. It is not that models suddenly became smarter. It is that they gained the ability to take action in a loop, over time, toward a goal.

What This Series Will Do

This series is for people like me (Stephen, not Priya!), who already understand how LLMs work, who have used chatbots like ChatGPT or Claude, and who are now hearing about “agents” and wondering what that actually means. I felt I was lagging in the AI world so I started this learning process to make sure I’m not left behind!

We will move beyond the half-done feeling and into the architecture of agentic systems. We will look at what agents actually are, how they are structured, what tools and skills give them their power, how multiple agents can coordinate, and how to think about trust, evaluation, and oversight.

We will also look at the platforms and frameworks that exist today — what they offer, how they differ, and what tradeoffs you are choosing when you pick one. This is not a manual for any one platform. It is a guide to understanding the category itself, so you can make good choices about whether, where, and how to use agentic AI in your own work.

The field moves fast. Some tools that looked promising six months ago may be superseded by the time you read this. That is fine. The principles matter more than the products. If you understand the loop, the anatomy, and the tradeoffs, you can look at whatever the current landscape happens to be and know what you are seeing.

Here’s the draft Table of Contents of this series in Agents Unpacked. This is likely to change as Priya and I progress through this project:

• Stephen’s Preface to Agents Unpacked

• Part I — The Mental Shift

  • Chapter 1 — From Answer to Outcome (this post)

  • Chapter 2 — Anatomy of an Agent

  • Chapter 3 — Skills, Tools, and the Action Loop

• Part II — Agents in Practice

  • Chapter 4 — Why One Agent Is Often Not Enough

  • Chapter 5 — Where Agents Are Actually Useful

  • Chapter 6 — Delegation Design

• Part III — Building and Trusting Agentic Work

  • Chapter 7 — Designing Your First Agentic Workflow

  • Chapter 8 — When Things Go Wrong: Evaluation, Guardrails, and Trust

  • Chapter 9 — What to Build Next
    

<< Previous Post: Stephen’s Preface to Agents Unpacked

>> Next Post: Coming Soon

Table of Contents

stephengruppetta.com

• Stephen’s Preface to Agents Unpacked

• Part I — The Mental Shift

  • Chapter 1 — From Answer to Outcome

  • Chapter 2 — Anatomy of an Agent

  • Chapter 3 — Skills, Tools, and the Action Loop

• Part II — Agents in Practice

  • Chapter 4 — Why One Agent Is Often Not Enough

  • Chapter 5 — Where Agents Are Actually Useful

  • Chapter 6 — Delegation Design

• Part III — Building and Trusting Agentic Work

  • Chapter 7 — Designing Your First Agentic Workflow

  • Chapter 8 — When Things Go Wrong: Evaluation, Guardrails, and Trust

  • Chapter 9 — What to Build Next

But things moved quickly, and I couldn’t quite catch up. I was busy with, you know, actual work, family, and life.

Then I started hearing lots of new terms, lots of new acronyms. I didn’t know what they were. I still don’t know what most of them are.

Then it was all about agents. I remember clearly thinking to myself: “Is this really any different from the ChatGPT-type chatbots?”

And here’s where this new series comes in. I decided to dive into agents and created a few. One of them is a learning tutor agent that I personalised to suit me.

I gave the agent all my tutorials and books. I gave the agent all the articles I wrote about my views on learning and technical writing.

I asked the agent to figure out from all this how I like to learn, how I like to communicate. I teach the way I like to learn, so it’s fine to put my teaching style in the mix. Then, I had a good long chat with my learning tutor agent to make sure we’re on the same page.

I gave a name to my agent (I named all my agents!) My personalised tutor is Priya.

This series is the joint effort between Priya and me to help me understand agents. What are they? How do they work?

Yes, it’s AI-generated content. But...

1. It’s generated by an AI that’s extremely well-versed in my style of learning and communication.

2. I had an active role in steering and editing the content. Here’s how...

What you’ll read in the following articles was created using the following process:

1. Priya researched the topic following my brief and created a course outline.

2. She drafted the first chapter.

3. I read through the file, leaving comments and questions along the way, directly within the text.

   1. I marked some comments as private.

   2. I marked some comments as public.

4. Priya revised the chapter by incorporating my comments and questions.

   1. She deleted the private comments from the text after making changes to address my comments.

   2. She kept my public comments and questions in the text, clearly marked as “Stephen’s questions”, and she answered them directly in the text.

5. I reviewed the chapter again and left more comments, and Priya revised the chapter again. We iterated through this until I was happy with the final text. And I was happy with the final text when I felt I understood everything in it and all my questions had been answered.

6. Priya then moves on to draft the second chapter, and the whole process starts again.

I will post these chapters as they emerge from this process. This is how I’m learning this topic. Hopefully, they may help some other people, too.

Here’s the planned table of contents for this Agents Unpacked series. But note, this may change!

• Stephen’s Preface to Agents Unpacked (this post)

• Part I — The Mental Shift

  • Chapter 1 — From Answer to Outcome

  • Chapter 2 — Anatomy of an Agent

  • Chapter 3 — Skills, Tools, and the Action Loop

• Part II — Agents in Practice

  • Chapter 4 — Why One Agent Is Often Not Enough

  • Chapter 5 — Where Agents Are Actually Useful

  • Chapter 6 — Delegation Design

• Part III — Building and Trusting Agentic Work

  • Chapter 7 — Designing Your First Agentic Workflow

  • Chapter 8 — When Things Go Wrong: Evaluation, Guardrails, and Trust

  • Chapter 9 — What to Build Next

Subscribe now

The challenge when learning something new is finding resources that are just right for me. And that’s not easy. I know I can learn things better and quicker with resources that fit my style well, but you can’t always find these resources.

I recently got particularly annoyed learning about the biomechanics of sprinting – I do have non-Python interests, yes – because all three textbooks I read, and lots of the online writing in this field, are just, let’s say, not great.

But I now found the solution.

After many decades of learning in the same way, I have now upgraded how I learn thanks to my new tutor, Priya.

Yes, I gave her a name. No, she’s not a real person. Priya is my personalised tutor agent. I’ll tell you all about her below.

And you’ll experience her teaching, too (not on the Python articles, though, I’ll keep writing those the old-fashioned way.) I’ll tell you more about this below, too, but let me first tell you why this works for me.

My Tutor, My Style

I’ve been thinking about the way I learn and teach for many years, from way back when I was a young University lecturer faced with 120 students in a lecture hall. I wasn’t that much older than the students, but I learnt fast. And they liked my teaching (I even have awards to prove it!)

More recently, I’ve been writing a lot. I wrote articles here on The Python Coding Stack and elsewhere. I wrote a Python textbook. I even wrote about learning and technical writing in Breaking the Rules: the substack and the book.

All this meant that I could ask my freshly-spawned agent to spend a bit of time reading what I wrote to understand how I teach, which is how I like to learn. Priya analysed the techniques I use in my writing and understood my motivations for doing what I do through my technical writing texts.

Then, Priya and I had a good chat to refine ideas, to make sure she captured the essence of “my style”.

And since Priya is an AI agent, “my style” became her knowledge base. This knowledge now lives in several lengthy markdown files and is summarised in shorter context packs and an index to ensure Priya’s short-term memory (the context window) isn’t overwhelmed.

Then I was ready to go. Any topic I wanted to learn, large or small, I could ask Priya to research it thoroughly, creating a new set of knowledge files, this time specific to the topic she needed to teach rather than my learning style. And then, she’s ready to teach me.

And it worked. The stuff she prepared was exactly the way I like it.

The Tutor-Student Conversation Course

And here’s the format I settled on (for now). Once the agent completes her research about the topic I want to learn, I ask her to plan a course spanning several modules.

But here’s the refinement loop that makes the real difference:

1. I ask Priya to draft the first module. She writes this in a markdown file.

2. I read through her draft and leave comments and questions directly within the text.

3. Priya reads my questions and revises the text to address my questions. (But read on to find out more about the two categories of comments/questions I leave for her.)

4. Repeat steps 2 and 3 until I feel I understand the topic.

5. Move on to the next module and repeat steps 1 to 4.

This is a human-in-the-loop approach to creating the learning material. Yes, Priya is trained in my way of learning and teaching and in my writing style. But I’m actively having a conversation with her within the text.

This is equivalent to raising your hand in a lesson and asking the teacher a question. A good teacher will then revise how they present the material to address your question.

Priya’s learning materials are just like that. In fact, I will take credit for her output. Sure, I’m not an expert in the subject matter she’s teaching me – that’s the whole point, right? But the output reflects my views and ideas about teaching and includes my questions and queries as I tried to understand and master the topic.

This is a collaboration. Priya and I are co-authors, even though Priya did most of the “writing”.

I tried this approach on several topics, but there are two I want to share with you. I’m setting up two new sections here on The Python Coding Stack, which I’ll use to learn these two topics in public. I’ll publish the “transcripts” of the conversations Priya and I are having. It’s mostly Priya doing the talking, but my questions are there, too.

The first topic I’m learning in public with Priya’s help is Agentic AI. It’s very meta to use agentic AI to learn about agentic AI! I’ll publish an introduction and the first module in the coming days in the new section here on The Python Coding Stack called Agents Unpacked. You can already see this section in the menu on the homepage.

I’ll set up another section to deal with the second topic in a week or so. No spoilers for now except to say it’s directly related to programming but it’s distinct from the articles I publish in the main section on The Python Coding Stack and in The Club.

By the way, you’ll be able to select which sections you want to receive regularly by email. So if you’re interested in my Python core content but not in these other topics, you can pick and choose what to opt out of. You can always go to The Python Coding Stack to read the other sections, of course.

How Priya and I Create These “Courses”

But let me expand on how Priya – my tutor agent – and I created these courses. [Incidentally, those are my em-dashes – I use them often and have always done. Commas would be ambiguous in that context!]

I provide two types of questions or comments to my agent as I read through the drafts: private and public.

Private Questions and Comments

When Priya reads the private questions or comments, she makes changes to the text, but then she deletes my input. So, you won’t see my intervention explicitly in these cases. However, Priya’s text reflects my thoughts. My interventions guide Priya. This type of intervention is similar to an editor’s role, but I’m intervening as a learner more than as an editor.

Public Questions and Comments

However, when Priya comes across a comment or question I mark as public, she leaves it in the text, acknowledges the question, and answers it directly. So, you’ll see my public questions in the text. Priya and I decided not to include too many of these public questions to keep the text flowing. However, I think it’s beneficial to see some of my interventions. My questions may also be your questions.

More Learning. More Articles. More Fun

As with everything to do with AI, this is all very new. It’s a work in progress. I may refine and revise how I interact with my agent. But it’s been fun learning this way, and I hope you enjoy reading my interactions with Priya and you find it useful, too.

To state the obvious, the posts I’ll publish in these two new sections are mostly AI-generated content. If you read this far, then you won’t be surprised by that statement. A year ago, I would never have thought I’d publish anything written by AI. But a year is a long time in the AI world. And this AI content reflects me and my thinking. The agent is my mentee – someone I trained to teach the way I do, to write the way I do. But she’s also my tutor, teaching me new stuff.

So there’s a lot of “me” in what you read, even if it’s mostly written by Priya!

The posts in the main section of The Python Coding Place and those in The Club (for premium subscribers) won’t change. They’re still my writing from beginning to end. Every word and letter you read in those posts is the result of nerve signals going from my brain to my fingers, which tap keys on a keyboard. In this era of AI doing a lot of work for us, I think it’s more important than ever for me to keep using my pre-AI skills. Otherwise, my brain will atrophy, and I don’t want that!

So, in summary, there will soon be four sections here on The Stack:

1. The main area in The Python Coding Stack – no change here, you’ll get the same type of Python articles you’ve been reading for the past 3+ years

2. The Club – the extra Python posts for premium subscribers

3. Agents Unpacked – the Agentic AI course Priya and I are creating for me to learn all about this agentic stuff. Learn with me (and Priya) if you’re interested.

4. Mystery Fourth Section – Stay tuned, you won’t have to wait long. This is also a Priya-Stephen collaboration.

Next post will be the introduction and first section in Agents Unpacked. Soon after, I have another Python post I’m planning for you.

Subscribe now

stephengruppetta.com

Photo by detait

However, there are some headaches and rabbit holes along the way.

And one of these challenges is when you start exploring the “get*” dunder methods. You come across .__getitem__(), .__getattr__(), .__getattribute__(), and .__get__() and you think:

“Aren’t they all the same thing? They’re all ‘getting’ stuff, right?”

Well, yes, they’re all “getting stuff”, which is why they have “get” in their names. But, as you probably guessed by now, they do different things.

One Deals With [] • The Others Deal With .

Let’s start with the odd one out, which is also possibly the least challenging of the lot. The .__getitem__() special method deals with the square bracket notation, [], which you place after an object’s name. These are the square brackets you use to get an item from a list or a dictionary, say. You’ll see later that the other special methods with “get” in their name deal with the dot, ., which you use in a different context in Python.

With lists, or any other sequence, you use the square brackets with an index that represents the position of an item within the data structure. So, numbers[0] gives you the value of the first item in numbers if numbers is a sequence.

With dictionaries, or mappings in general, you place the key inside the square brackets to fetch the value associated with that key, such as points["Sam"], which gives the value associated with the key "Sam" if it exists.

Let’s explore a custom class:

You pass a mapping to PointsTable when creating an instance of the class. The data is stored as a dictionary in the ._data attribute.

Now, let’s say you’d like to access points from a PointsTable instance using the square bracket notation, as you would do with a dictionary:

Unfortunately, this doesn’t work:

Traceback (most recent call last):
  ...
    print(table[”Mark”])
          ~~~~~^^^^^^^^
TypeError: ‘PointsTable’ object is not subscriptable

The PointsTable instance is not a dictionary. It includes a dictionary as a data attribute. The PointsTable object itself is not subscriptable, which means you can’t use the square bracket notation to fetch an item.

When Python sees the square bracket notation after an object identifier, such as table, it calls the class’s .__getitem__() special method. If it’s not there, as in this case, Python raises a TypeError saying the object is not subscriptable.

But what does this tell you? If you want to use the square brackets notation, you just need to define .__getitem__() in the class:

Now, Python finds the .__getitem__() special method. It takes whatever you placed within the square brackets and passes it as an argument to .__getitem__(). Therefore, table["Mark"] returns whatever the call to PointsTable.__getitem__(table, "Mark") returns:

17

The code outputs Mark’s points. The PointsTable class is now subscriptable since it has a .__getitem__() special method. The term item generally refers to the objects contained within a data structure. Therefore, __getitem__() is there to let you get an item from within a data structure.

Let’s make this example a bit more interesting before we move on to the other special methods with “get” in their name.

Try the following:

This raises an error:

Traceback (most recent call last):
  File ..., line 18, in <module>
    print(table[”Mark”, “Stephen”])
          ~~~~~^^^^^^^^^^^^^^^^^^^
  File ..., line 6, in __getitem__
    return self._data[item]
           ~~~~~~~~~~^^^^^^
KeyError: (’Mark’, ‘Stephen’)

There’s no key equal to ("Mark", "Stephen"). So, you get a KeyError. Note how Python places parentheses around the two names when showing you the KeyError, even though you didn’t use the parentheses within the square brackets in table["Mark", "Stephen"]. This gives you a clue to what’s happening. But let’s explore this so we’re sure and we don’t just rely on our hunches:

You add two print() calls in .__getitem__(). Here’s what they print out before Python raises the KeyError:

item=(’Mark’, ‘Stephen’)
type(item)=<class ‘tuple’>

The object Python passes to .__getitem__() is the tuple ("Mark", "Stephen"). You decide to make your PointsTable class super-flexible so you can include more than one name within the square brackets notation:

The .__getitem__() method now checks whether item is a tuple first. If it is, then .__getitem__() returns a tuple with all the values. If item is not a tuple, then it must be just a single key, and the square brackets behave in the normal way:

(17, 20)
(22, 20, 19)
17

But note that you can no longer use an actual tuple as a dictionary key. This is fine in this example since all the keys are strings with people’s names.

So, .__getitem__() gives you full control over what happens when you use the square brackets notation to fetch an item from an object. Now, let’s move on to the other “get*” special methods.

Support The Python Coding Stack

Accessing Data Using The Dot Notation, .

You decide you also want PointsTable to work with the dot notation, so that you can use table.Mark to get the number of points Mark has.

Note that I’m adding more features to this class to demonstrate various special methods in this article. I’m not suggesting it’s necessarily always desirable to try to be too fancy with your classes!

Let’s try it out directly first, without making any changes to the class:

Unfortunately, table.Mark raises an error:

Traceback (most recent call last):
  File ..., line 32, in <module>
    print(table.Mark)
          ^^^^^^^^^^
AttributeError: ‘PointsTable’ object has no attribute ‘Mark’

Python is looking for an attribute called Mark. Attributes are the things you can access using the dot notation in Python. Typically, these are data attributes, methods, properties, and other things you can access using the dot. However, as with everything else, Python provides special methods to handle this. But this is where it gets a bit complicated.

.__getattr__()

Let’s take this one step at a time. Let’s add the .__getattr__() special method. You can guess that this special method name stands for get attribute (but it’s not the only special method that stands for get attribute!) Let’s add this method:

The .__getattr__() special method has a print() call to show when it’s called. This line is not required, but it will help you figure out when Python calls this special method. It’s a way of peeking into Python’s inner workings! For completeness, I added a similar line to .__getitem__() as well.

There are two print() calls at the bottom: one uses the square brackets notation, which you dealt with in the previous section, and the other uses the dot notation. Now, both work:

Calling __getitem__ with argument item=’Mark’
17
Calling __getattr__ with argument name=’Mark’
17

As you’ve seen in the previous section, the square brackets notation relies on .__getitem__().

And as the third and fourth lines in this printout show, Python used .__getattr__() to deal with table.Mark.

Python passed the attribute name to the .__getattr__() special method’s name parameter. This method then fetches the value associated with this name from the dictionary ._data.

This trick only works for keys that are valid attribute names and that don’t clash with real attributes or methods. That’s one reason dictionary-style access is often preferable for arbitrary user data. But let’s stick with this exercise in this article.

Also, you wrote the code such that if self._data[name] raises a KeyError, then .__getattr__() raises an AttributeError, which is what you’d expect if you try to use a name that’s not an attribute after the dot:

This raises an AttributeError:

AttributeError: ‘PointsTable’ object has no attribute ‘Matilda’

It seems that .__getattr__() is analogous to .__getitem__() – the first deals with . and the second with []. But...

But things are a bit more complex. What if you try to access a standard attribute, say a data attribute or a method? You haven’t defined any methods in this class, but you do have a data attribute, ._data. It’s marked with a leading underscore, showing you shouldn’t access it directly. But you can “break the rules”. Python won’t stop you:

The output is the following:

{’Stephen’: 20, ‘Mark’: 17, ‘Kate’: 19, ‘Sarah’: 22}

Perhaps, this is not surprising. But note that there’s no printout that says that .__getattr__() was called, as you got when you used table.Mark. Let’s look at both outputs together:

The output now is the following:

Calling __getattr__ with argument name=’Mark’
17
{’Stephen’: 20, ‘Mark’: 17, ‘Kate’: 19, ‘Sarah’: 22}

Python called the .__getattr__() special method when you accessed .Mark, but not when you accessed ._data.

This means there’s something else happening behind the scenes.

.__getattribute__()

This is where .__getattribute__() enters the scene. No prizes for guessing that this method name stands for get attribute, but then so did .__getattr__(). This is confusing.

And here’s the thing. It’s .__getattribute__() that gets called each and every time you use the dot notation with an instance. Let’s define it:

Note that this code breaks lots of things as it is now. We’ll fix it soon. Let’s first look at the output from this code:

Calling __getattribute__ with argument name=’Mark’
None
Calling __getattribute__ with argument name=’_data’
None

As you can see, Python calls .__getattribute__() for both table.Mark and table._data. However, the current .__getattribute__() method does nothing else. Therefore, it returns None. You’ll see soon that this is a bigger problem than you might think.

Let’s try using the dot notation to access an attribute that doesn’t exist:

Every time you use the dot, ., Python calls .__getattribute__(), which, for now, does nothing except print out the self-serving statement to show it was called:

Calling __getattribute__ with argument name=’Mark’
None
Calling __getattribute__ with argument name=’_data’
None
Calling __getattribute__ with argument name=’Janine’
None

Therefore, you can’t access any attribute at the moment. All data attributes and methods are out of reach. You should be careful when overriding .__getattribute__(). In fact, you’ll rarely need to write your own .__getattribute__().

Let’s fix this code by calling object.__getattribute__() from within PointsTable.__getattribute__(). Recall that all Python classes inherit from object. And object.__getattribute__() is where the important logic that Python uses to decide what to do when you use the . lives. You can access object using super():

Note that in this class, .__getattribute__() is not doing anything different to the default version except for printing out a statement. I’m using this for demonstration purposes only.

Let’s deal with one print() call at a time. The code above includes print(table._data). You’re fetching the value of a data attribute:

Calling __getattribute__ with argument name=’_data’
{’Stephen’: 20, ‘Mark’: 17, ‘Kate’: 19, ‘Sarah’: 22}

Python calls PointsTable.__getattribute__(), which in turn calls the base class’s object.__getattribute__(). This special method recognises that ._data is an instance variable and fetches its value from the object’s .__dict__.

But let’s see what happens with the .Mark attribute access. Recall that .Mark is not a data attribute or method:

Here’s the output now. Look at the various printouts, too:

Calling __getattribute__ with argument name=’Mark’
Calling __getattr__ with argument name=’Mark’
Calling __getattribute__ with argument name=’_data’
17

When Python sees table.Mark, it calls PointsTable.__getattribute__(table, "Mark"). This call gives rise to the first line printed out above. The base class’s .__getattribute__() doesn’t recognise "Mark" as a data attribute, a method name, or anything else it’s expecting (such as descriptors, which we will discuss soon). So, object.__getattribute__() actually raises an AttributeError in this case, which you can’t see in the output above.

That’s because when .__getattribute__() raises an AttributeError during normal attribute access using the dot, Python checks whether there’s a .__getattr__() defined. If there is, it uses it as a fallback. That’s what happens in this case. You can see that .__getattr__() is called next – that’s the second line printed out above.

But .__getattr__() contains this line: return self._data[name]. There’s a dot in that line, too. So Python needs to access .__getattribute__() again, this time using "_data" as the argument. Since ._data is a data attribute, object.__getattribute__() doesn’t need to use .__getattr__() as a fallback since it knows how to deal with data attributes.

Confused? You’re not alone. This is confusing, I know. And there’s a bit more. But we’ll make sure things are clear by the end of this article.

But first, let’s focus on table.Mark again and let’s comment out, just for now, the definition of .__getattr__():

The base class’s .__getattribute__() method works hard to determine whether "Mark" is an attribute it expects. It doesn’t find it, and now, there’s no fallback .__getattr__(), so Python raises an error:

Calling __getattribute__ with argument name=’Mark’
Calling __getattribute__ with argument name=’__dict__’
Calling __getattribute__ with argument name=’__class__’
Traceback (most recent call last):
  ...
AttributeError: ‘PointsTable’ object has no attribute ‘Mark’

You can see a couple of extra printouts, too. These are side effects of Python preparing the error message since we get the printout each time Python uses the dot notation, even when it does so behind the scenes. Let’s not go too far down the rabbit hole in this article. We may never come out again if we go too deep!

You’ve seen that when you use the dot notation, Python calls object.__getattribute__() eventually. That’s why you should include super().__getattribute__() when you override this special method, unless you have a clear (niche) reason why you don’t want to do this. This special method looks in a number of places for the attribute. I’m going to skip a few steps in the hierarchy for now. But we’ll revisit these in the next section (remember, we’re not done yet since there’s still .__get__() to deal with).

The .__getattribute__() special method looks for instance attributes. Let’s temporarily turn .Mark into an instance attribute:

Now, .Mark is a data attribute, so .__getattribute__() finds it:

Calling __getattribute__ with argument name=’Mark’
This is Mark as a data attribute

If it’s not an instance variable, object.__getattribute__() also checks whether it’s a class attribute. Let’s test this:

Here’s the output:

Calling __getattribute__ with argument name=’Mark’
This is Mark as a class attribute

Note that the line in .__init__() that defines .Mark as a data attribute is commented out now because instance attributes take precedence over class attributes. Only when all else fails (including the steps I skipped for now), does Python check whether the fallback special method .__getattr__() is there.

So, when you use the dot notation for attribute access (the simplified version, for now):

1. Python calls .__getattribute__()

2. It checks for known attributes, such as instance attributes and class attributes (and a bit more – coming soon)

3. If it doesn’t find anything, then .__getattribute__() raises an AttributeError. However, Python does one final check before letting this error through to the user: is there a .__getattr__()? If there is, Python calls it to see whether it contains instructions for handling this unknown attribute.

I’m simplifying slightly here. Descriptors complicate the order, and we’ll return to them in the next section.

Note that in real code, you rarely need to define .__getattribute__(). You can rely on the default provided in the object base class in most cases. Most custom attribute behaviour should use .__getattr__() rather than .__getattribute__(). Use .__getattribute__() only when you need to intercept every attribute access. There is rarely a need for this, though.

Common uses for .__getattr__() include delegating missing attributes to another object, implementing lazy loading, or exposing dynamic attributes from structured data.

Here’s the tidied-up code in full so far:

The Asymmetry Between .__getattr__() and .__setattr__()

A short note: There’s an asymmetry between .__getattr__() and .__setattr__() despite their names following the same pattern. As you’ve seen above, .__getattr__() is only used as a fallback when .__getattribute__() doesn’t find the attribute through “normal routes”.

However, .__setattr__() doesn’t have a counterpart analogous to .__getattribute__(). Therefore, .__setattr__() is always used when setting the value of an attribute. Life is simpler in the “setting” world!

In this article, I’m focusing on the “getting” part of things, but some of the same logic applies to “setting”, too.

The Fourth Horseman: .__get__()

I’ve had this article in my pipeline here on The Python Coding Stack for a very long time. But I never wrote it because I knew that dealing with .__get__() would be a pain and would need a lot of time and space. However, in March, I published The Weird and Wonderful World of Descriptors in Python. If you haven’t read that article yet, well, now is the time!

And therefore I can cheat in this article. I can avoid talking about .__get__() in detail. What follows in this section is a summary of the world of descriptors. See the full article for more details.

If a class has .__get__(), it’s a descriptor class. If it only has .__get__() and it does not define .__set__() or .__delete__(), which are the other methods that make up the descriptor protocol, then the class creates non-data descriptors. Classes that include .__set__() or .__delete__() create data descriptors. This distinction matters when we get back to .__getattribute__() and the order it uses to look for known attributes.

Let’s explore this with a dummy example. This code defines two descriptor classes and another class to test the priority order that object.__getattribute__() uses for different attribute types:

DataDescriptor defines .__get__() and .__set__() methods, which is why it creates data descriptors. However, NonDataDescriptor creates non-data descriptors since the class only defines .__get__().

All the special methods in all classes (except .__init__()) have print() calls so you can see when Python calls them. Before we look at the output from this code, let’s review the five attributes you use in the print() calls and where they appear in the class definitions:

1. .first is defined as a data descriptor at the top of the TestingAttributeAccess class. Recall from the article on descriptors that you initially define descriptors as class attributes within a class. You also assign a value to .first within the class’s .__init__(). More on this soon.

2. .second is also defined as a descriptor at the top of the TestingAttributeAccess class, but it’s a non-data descriptor since the NonDataDescriptor class only defines .__get__(). You also assign a value to self.second within .__init__(). We’ll see how the class actually deals with .second soon, since it treats it differently from .first.

3. .third is also defined as a non-data descriptor. However, there is no other assignment to third within .__init__(). This is the key difference between .second and .third in this code.

4. .fourth is a class attribute. It’s not a descriptor and it’s not an instance attribute.

5. .fifth is not defined anywhere in the class, but you still call print(test.fifth).

Let’s look at the whole output from this code and then break it down into steps later:

Calling DataDescriptor.__set__ with value=”’first’: a data attribute defined in .__init__”

Printing `test.first`
Calling __getattribute__ with argument name=’first’
Calling DataDescriptor.__get__
This is the Data Descriptor

Printing `test.second`
Calling __getattribute__ with argument name=’second’
‘second’: a data attribute defined in .__init__

Printing `test.third`
Calling __getattribute__ with argument name=’third’
Calling NonDataDescriptor.__get__
This is the Non-Data Descriptor

Printing `test.fourth`
Calling __getattribute__ with argument name=’fourth’
I’m just a normal class attribute!

Printing `test.fifth`
Calling __getattribute__ with argument name=’fifth’
Calling __getattr__ with argument name=’fifth’
This is the fallback value from __getattr__

There’s plenty to digest there. So, let’s break it down in stages.

0. Creating an instance of TestingAttributeAccess

I want to focus on the “getting” bit from the various print() calls. However, there’s this line output first, so let’s deal with it, too:

Calling DataDescriptor.__set__ with value=”’first’: a data attribute defined in .__init__”

This happens when you create the instance of TestingAttributeAccess and assign it to the variable name test. Why? Because first has already been defined as a data descriptor at the time of the class definition. Therefore, when you assign a value to .first during initialisation, the descriptor protocol kicks in.

When you call TestingAttributeAccess(), Python calls the class’s .__init__() and soon finds this expression: self.first = ...

Therefore, Python calls DataDescriptor.__set__() to set the value of this attribute. Note that DataDescriptor.__set__() doesn’t really set anything in this case. Normally, something more meaningful would happen in .__set__(). See The Weird and Wonderful World of Descriptors in Python for more on this.

1. test.first

The next section in the code’s output is the following:

Printing `test.first`
Calling __getattribute__ with argument name=’first’
Calling DataDescriptor.__get__
This is the Data Descriptor

This output is created when you write print(test.first). The dot notation triggers Python to call .__getattribute__(). So, let’s start exploring the hierarchy of checks in .__getattribute__(). What does this special method look for first?

The first thing .__getattribute__() checks is whether the attribute is a data descriptor.

Since .first is a data descriptor, the search stops there. Python calls DataDescriptor.__get__() and returns whatever the .__get__() method returns. This is the string "This is the Data Descriptor" in this demo example.

2. test.second

Let’s look at the next segment in the code’s output. This is generated when you call print(test.second):

Printing `test.second`
Calling __getattribute__ with argument name=’second’
‘second’: a data attribute defined in .__init__

You know the drill by now. The dot notation is the trigger that makes Python call .__getattribute__(). It checks whether the attribute is a data descriptor. But .second is not a data descriptor. You initially define it as a non-data descriptor. Since it’s definitely not a data descriptor, it’s time to move on to the second check in the hierarchy.

The second thing .__getattribute__() checks is whether the attribute is an instance attribute.

Is this attribute name in the object’s .__dict__ dictionary? Instance attributes come second in the hierarchy. Now, you’ve seen that .second is not a data descriptor, which is why we moved to the second check. But is test.second an instance attribute or a non-data descriptor?

You can see from the printout that Python never called NonDataDescriptor.__get__() even though you originally defined .second as a NonDataDescriptor object. The swap occurred when you created the TestingAttributeAccess instance. Python calls the class’s .__init__(), which contains this assignment line: self.second = ....

Since .second is not a data descriptor (it’s a non-data descriptor) and doesn’t have a .__set__(), the descriptor protocol doesn’t kick in here to set the value of this attribute. This is different from .first, which was a data descriptor and, therefore, its .__set__() was responsible for setting the value.

Since the .second non-data descriptor doesn’t have a .__set__(), Python does what it always does in these cases: it assigns the new object, the string "'second': a data attribute defined in .__init__", to test.second. Therefore, test.second is now an instance attribute containing the string rather than the original non-data descriptor.

Although test.second would originally access the non-data descriptor, you created an instance attribute when you initialised the object. It’s now a data attribute, which is an instance attribute.

Note that TestingAttributeAccess.second is still there as a class attribute, and that’s still the non-data descriptor. But test.second now points to another object, the string.

Therefore, when later in the code you call print(test.second), Python calls .__getattribute__() and starts looking through its hierarchy. It’s not a data descriptor. It is an instance attribute. Therefore, it returns its value.

3. test.third

The next block of output is the one generated when you call print(test.third). Recall that .third is a NonDataDescriptor object. So was .second initially. However, unlike .second, Python never creates an instance attribute that’s assigned to test.third since you don’t assign to self.third in .__init__(). Here’s the output:

Printing `test.third`
Calling __getattribute__ with argument name=’third’
Calling NonDataDescriptor.__get__
This is the Non-Data Descriptor

In print(test.third), there’s a dot again, so Python calls .__getattribute__(). This method first checks whether .third is a data descriptor. It is not. Then it checks whether .third is an instance attribute. It is not. So...

The third thing .__getattribute__() checks is whether the attribute is a non-data descriptor.

Non-data descriptors come next in the hierarchy. That’s why you can see the printout saying that Python calls NonDataDescriptor.__get__ and then prints the attribute’s value, which is the value it gets from the non-data descriptor’s .__get__().

Incidentally, ordinary instance methods are functions that also implement the .__get__() special method. They are non-data descriptors. So, methods are found in this third check in the hierarchy of checks when you use the dot notation.

4. test.fourth

It’s test.fourth‘s turn. Here’s the output you get from print(test.fourth):

Printing `test.fourth`
Calling __getattribute__ with argument name=’fourth’
I’m just a normal class attribute!

There’s a dot, so there’s a call to .__getattribute__() which checks whether .fourth is a data descriptor, an instance attribute, or a non-data descriptor, in that order. It’s none of these. Time for the fourth check.

The fourth thing .__getattribute__() checks is whether the attribute is a class attribute.

And look at that! The .fourth attribute is indeed a class attribute.

5. test.fifth

And finally, it’s .fifth‘s turn. You call print(test.fifth) and you get the following:

Printing `test.fifth`
Calling __getattribute__ with argument name=’fifth’
Calling __getattr__ with argument name=’fifth’
This is the fallback value from __getattr__

Yes, yes, there’s a dot again, so Python calls .__getattribute__(). No, it’s not a data descriptor. No, it’s not an instance attribute. No, it’s not a non-data descriptor. No, it’s not a class attribute. It’s nothing. The attribute .fifth doesn’t exist.

When .__getattribute__() fails to find the attribute, Python calls .__getattr__() if this special method exists.

And that’s it. Simple, eh?!?!

Let’s finish off by summarising the order in which Python deals with looking for attributes when you use the dot notation:

1. The first thing .__getattribute__() checks is whether the attribute is a data descriptor.

2. The second thing .__getattribute__() checks is whether the attribute is an instance attribute.

3. The third thing .__getattribute__() checks is whether the attribute is a non-data descriptor.

4. The fourth thing .__getattribute__() checks is whether the attribute is a class attribute.

5. Finally, when .__getattribute__() fails to find the attribute, Python calls .__getattr__() if this special method exists.

The fifth step occurs when you access attributes using the dot notation. If you call .__getattribute__() explicitly, which you rarely need to do, Python doesn’t automatically look for .__getattr__().

Note that Python looks for descriptors, instance and class attributes in the class and its base classes, too, following the method resolution order.

Final Words

Wait for this. I’ve been waiting all article to write this:

Now, you won’t forget the four “get*” special methods.

Got it! (My son just walked out disapprovingly when he read this.)

You may never need to use all of these methods. Maybe you’ll never need to use any of them. But if you ever wondered why there are these four special methods with similar names, now you know what they do. And you dug a bit more underneath the Python surface along the way. And that’s always fun.

Photo by Robert Clark

Code in this article uses Python 3.14

The code images used in this article are created using Snappify. [Affiliate link]

Join The Club, the exclusive area for paid subscribers for more Python posts, videos, a members’ forum, and more.

Subscribe now

You can also support this publication by making a one-off contribution of any amount you wish.

Support The Python Coding Stack

For more Python resources, you can also visit Real Python—you may even stumble on one of my own articles or courses there!

Also, are you interested in technical writing? You’d like to make your own writing more narrative, more engaging, more memorable? Have a look at Breaking the Rules.

And you can find out more about me at stephengruppetta.com

Further reading related to this article’s topic:

• The Manor House, the Oak-Panelled Library, the Vending Machine, and Python’s `_getitem__()`

• The Weird and Wonderful World of Descriptors in Python • The Price is Right

Appendix: Code Blocks

Code Block #1

class PointsTable:
    def __init__(self, data):
        self._data = dict(data)

table = PointsTable(
    {
        "Stephen": 20,
        "Mark": 17,
        "Kate": 19,
        "Sarah": 22,
    }
)

Code Block #2

# ...
print(table["Mark"])

Code Block #3

class PointsTable:
    def __init__(self, data):
        self._data = dict(data)

    def __getitem__(self, item):
        return self._data[item]

table = PointsTable(
    {
        "Stephen": 20,
        "Mark": 17,
        "Kate": 19,
        "Sarah": 22,
    }
)

print(table["Mark"])

Code Block #4

# ...
print(table["Mark", "Stephen"])

Code Block #5

class PointsTable:
    def __init__(self, data):
        self._data = dict(data)

    def __getitem__(self, item):
        print(f"{item=}")
        print(f"{type(item)=}")
        return self._data[item]

table = PointsTable(
    # ...
)

print(table["Mark", "Stephen"])

Code Block #6

class PointsTable:
    def __init__(self, data):
        self._data = dict(data)

    def __getitem__(self, item):
        if isinstance(item, tuple):
            return tuple(self._data[key] for key in item)
        return self._data[item]

table = PointsTable(
    # ...
)

print(table["Mark", "Stephen"])
print(table["Sarah", "Stephen", "Kate"])
print(table["Mark"])

Code Block #7

# ...
print(table.Mark)

Code Block #8

class PointsTable:
    # ...

    def __getitem__(self, item):
        print(f"Calling __getitem__ with argument {item=}")
        if isinstance(item, tuple):
            return tuple(self._data[key] for key in item)
        return self._data[item]

    def __getattr__(self, name):
        print(f"Calling __getattr__ with argument {name=}")
        try:
            return self._data[name]
        except KeyError:
            raise AttributeError(
                f"'{type(self).__name__}' object has "
                f"no attribute '{name}'"
            ) from None

table = PointsTable(
    # ...
)

print(table["Mark"])
print(table.Mark)

Code Block #9

# ...
print(table.Matilda)

Code Block #10

# ...
print(table._data)

Code Block #11

# ...
print(table.Mark)
print(table._data)

Code Block #12

class PointsTable:
    def __init__(self, data):
        self._data = dict(data)

    def __getitem__(self, item):
        # ...

    def __getattr__(self, name):
        # ...

    def __getattribute__(self, name):
        print(f"Calling __getattribute__ with argument {name=}")

table = PointsTable(
    # ...
)

print(table.Mark)
print(table._data)

Code Block #13

# ...
print(table.Mark)
print(table._data)
print(table.Janine)

Code Block #14

class PointsTable:
    def __init__(self, data):
        self._data = dict(data)

    def __getitem__(self, item):
        # ...

    def __getattr__(self, name):
        # ...

    def __getattribute__(self, name):
        print(f"Calling __getattribute__ with argument {name=}")
        return super().__getattribute__(name)

table = PointsTable(
    # ...
)

print(table._data)

Code Block #15

# ...
print(table.Mark)

Code Block #16

class PointsTable:
    def __init__(self, data):
        self._data = dict(data)

    def __getitem__(self, item):
        # ...

    # def __getattr__(self, name):
    #     print(f"Calling __getattr__ with argument {name=}")
    #     try:
    #         return self._data[name]
    #     except KeyError:
    #         raise AttributeError(
    #             f"'{type(self).__name__}' object has "
    #             f"no attribute '{name}'"
    #         ) from None

    def __getattribute__(self, name):
        print(f"Calling __getattribute__ with argument {name=}")
        return super().__getattribute__(name)

table = PointsTable(
    # ...
)

print(table.Mark)

Code Block #17

class PointsTable:
    def __init__(self, data):
        self._data = dict(data)
        self.Mark = "This is Mark as a data attribute"

    def __getitem__(self, item):
        # ...

    # def __getattr__(self, name):
    # ...

    def __getattribute__(self, name):
        print(f"Calling __getattribute__ with argument {name=}")
        return super().__getattribute__(name)

table = PointsTable(
    # ...
)

print(table.Mark)

Code Block #18

class PointsTable:
    Mark = "This is Mark as a class attribute"

    def __init__(self, data):
        self._data = dict(data)
        # self.Mark = "This is Mark as a data attribute"

    def __getitem__(self, item):
        # ...

    # def __getattr__(self, name):
    # ...

    def __getattribute__(self, name):
        print(f"Calling __getattribute__ with argument {name=}")
        return super().__getattribute__(name)

table = PointsTable(
    # ...
)

print(table.Mark)

Code Block #19

class PointsTable:
    def __init__(self, data):
        self._data = dict(data)

    def __getitem__(self, item):
        print(f"Calling __getitem__ with argument {item=}")
        if isinstance(item, tuple):
            return tuple(self._data[key] for key in item)
        return self._data[item]

    def __getattr__(self, name):
        print(f"Calling __getattr__ with argument {name=}")
        try:
            return self._data[name]
        except KeyError:
            raise AttributeError(
                f"'{type(self).__name__}' object has "
                f"no attribute '{name}'"
            ) from None

    def __getattribute__(self, name):
        print(f"Calling __getattribute__ with argument {name=}")
        return super().__getattribute__(name)

table = PointsTable(
    {
        "Stephen": 20,
        "Mark": 17,
        "Kate": 19,
        "Sarah": 22,
    }
)

print(table._data)  # Only .__getattribute__() used
print(table.Mark)  # Fallback .__getattr__() needed here

Code Block #20

class DataDescriptor:
    def __get__(self, instance, owner):
        print("Calling DataDescriptor.__get__")
        return "This is the Data Descriptor"

    def __set__(self, instance, value):
        print(f"Calling DataDescriptor.__set__ with {value=}")

class NonDataDescriptor:
    def __get__(self, instance, owner):
        print("Calling NonDataDescriptor.__get__")
        return "This is the Non-Data Descriptor"

class TestingAttributeAccess:
    first = DataDescriptor()
    second = NonDataDescriptor()
    third = NonDataDescriptor()
    fourth = "I'm just a normal class attribute!"

    def __init__(self):
        self.first = "'first': a data attribute defined in .__init__"
        self.second = "'second': a data attribute defined in .__init__"

    def __getattribute__(self, name):
        print(f"Calling __getattribute__ with argument {name=}")
        return super().__getattribute__(name)

    def __getattr__(self, name):
        print(f"Calling __getattr__ with argument {name=}")
        return "This is the fallback value from __getattr__"

test = TestingAttributeAccess()
print("\nPrinting `test.first`")
print(test.first)
print("\nPrinting `test.second`")
print(test.second)
print("\nPrinting `test.third`")
print(test.third)
print("\nPrinting `test.fourth`")
print(test.fourth)
print("\nPrinting `test.fifth`")
print(test.fifth)

For more Python resources, you can also visit Real Python—you may even stumble on one of my own articles or courses there!

Also, are you interested in technical writing? You’d like to make your own writing more narrative, more engaging, more memorable? Have a look at Breaking the Rules.

And you can find out more about me at stephengruppetta.com

But one thing kept catching my attention: all the noise outside. Everyone’s talking about AI writing code, agents shipping products, the death of programming. And every so often, someone asks me — usually with that slightly guilty look of someone who thinks they’re about to insult my livelihood — “but do you really still need to learn Python? In this age?”

So I’ve been thinking about it. Properly. And I wanted to share where I’ve landed, because I think the answer matters and it’s different from what some hot takes suggest.

Do I still need to learn to code?

The way we write computer programs is changing. I don’t have a crystal ball for what programming looks like in five or ten years.

But here’s the thing: neither does anyone else.

Here’s what I know from watching others and experimenting with AI tools in my own work: right now, the people getting the most out of AI are the ones who already know how to code.

Here’s the hierarchy as I see it today:

• No coding knowledge = Little benefit from AI

• Little coding knowledge = Intermediate benefit from AI

• Intermediate coding knowledge = Great benefit from AI

• Great coding knowledge = Superpower-like benefit from AI

We’re in an era where some coding knowledge takes you much further than it could have a few years ago. That’s not an argument against learning to code. It’s an argument for it.

“But what about those vibe coding people? They seem to be shipping things.” Some are. I’ll be honest about that.

The projects I’ve seen from pure vibe coders tend to be smaller, tend to follow well-trodden patterns, and often hit a ceiling when something goes slightly wrong or slightly off-piste. Which is fine for a side project. I just created a useful dashboard to help me organise my day the way I want to using this approach.

But it tells you something: the AI does the heavy lifting on the known stuff. The moment something needs genuine thinking, you need the human who knows what’s going on beneath the surface.

AI-Assisted Human Coding and Human-Assisted AI Coding

Most serious work right now is a partnership.

Sometimes it’s AI-assisted human coding. The human drives, AI assists.

Sometimes it’s human-assisted AI programming. The AI writes most of the code, but the human knows what to ask for, how to steer it toward good design, how to evaluate whether the output actually makes sense.

Even when the coding looks like it’s done by AI, the person prompting and reviewing it is generally an experienced programmer. They’ve learned enough Python to know what’s reasonable, what’s a red flag, and when the AI is confidently wrong.

I’ve been experimenting with agentic AI over the past few weeks. I’ve tackled side projects I’d never have had the time for before. I didn’t have the time to start then, let alone finish them. Some of that output will show up in other places — stay tuned! But not here. The Python Coding Stack is the place for my writing.

The Programming Mindset When Talking to AI Agents

Here’s one thing I noticed. Talking to AI agents isn’t like talking to humans. But it isn’t like talking to computers (a.k.a. programming) either.

You need both qualities at the same time.

You need the clarity and communication skills you’d use with a person — explaining context, setting direction, knowing what matters, using clear language.

And you need the precision you’d use when programming — no ambiguity, clear intent, structure.

A good programmer who’s also a good communicator is the best human to work with AI agents.

That’s not coincidental. The same thought habits that make you an effective programmer also make you an effective prompter and reviewer when AI is involved.

I’ll share some of the prompts I’m using in a future post and analyse them to discuss why I wrote what I wrote. Learning to code well gives you an unfair advantage in this new world.

There’s Never Been a Better Time to Learn Python

Here’s what I’ve convinced myself of after all this. There’s never been a better time to learn Python.

A few years ago, you needed to reach an intermediate-to-advanced level before you could do something genuinely useful with Python. The bar was high. Now, with AI assistance, the bar is lower. Less Python knowledge takes you further than ever.

What used to need expertise can now be explored with curiosity and a bit of intermediate-ish-level Python.

That’s not replacing deeper learning. It’s making the entry point more accessible. And once you’re in, you can go as deep as you want.

So yes, I’ll keep coding in Python. Sometimes with a bit of help from AI. Sometimes with a lot of help from AI.

And yes, I’ll keep writing about Python here, as I’ve always done.

The Fun Factor

Here’s the thing we don’t talk about enough when discussing programming. It’s fun. It’s challenging. It’s rewarding. It’s fulfilling. It’s stimulating. It keeps my brain active.

I code because I enjoy it. I’ll keep writing about Python because I enojoy that, too,, and because I find value in sharing here — for myself and, hopefully, for you too.

Normal service resumes here. More Python posts coming. And maybe, just maybe, some of the AI things I’m learning will make their way in, too.

Psst–did you know you can become a premium member to be a part of The Club? It would mean so much to me!

Subscribe now

Quick question:

Photo by Maksim Goncharenok

Join The Club, the exclusive area for paid subscribers for more Python posts, videos, a members’ forum, and more.

Subscribe now

You can also support this publication by making a one-off contribution of any amount you wish.

Support The Python Coding Stack

For more Python resources, you can also visit Real Python—you may even stumble on one of my own articles or courses there!

Also, are you interested in technical writing? You’d like to make your own writing more narrative, more engaging, more memorable? Have a look at Breaking the Rules.

And you can find out more about me at stephengruppetta.com

It still is.

I started writing about Python as a learning tool—to help me sort things out in my head, weave a thread through all the disparate bits of information, clarify my thoughts, make sure any knowledge gaps are filled.

I started The Python Coding Stack three years ago. That’s the first of the mystery numbers in the post’s title revealed! I had written elsewhere before, but at the time of starting The Stack, I felt I had found my own “Python voice”. I had been teaching Python for nearly a decade. I had written plenty of articles, but setting up The Python Coding Stack was a deliberate choice to step up. I was still writing articles primarily for my own benefit, but now I was also writing for others, hoping they would want to learn the way I do.

And 7,600 subscribers apparently do. Thank you for joining this journey, whether you were there three years ago or you joined a few days ago. If you just joined, there’s an archive of 121 articles, most of them long-form tutorials or step-by-step guides.

A special thank you to the 33 subscribers who chose to upgrade to premium and join The Club. It may only amount to 3 coffees per month for you, but it makes a difference to me. Thank you! I hope you’ve been enjoying the exclusive content for The Club members.

And perhaps, if a few more decide to join you in The Club (you can surely cut three coffees out of your monthly intake!), then this publication may even become self-sustainable. Your support can make a real difference—if you value these articles and want to see them continue, please consider joining now. At the moment, I give up a lot of my time for free to think about my articles, plan them, draft them, review them technically, review them linguistically, get them ready for publication, and then publish.

Subscribe now

I mentioned my live teaching earlier. My written articles and my live teaching have a lot in common. One of the hardest things about teaching (or communication in general) is to place yourself in the learner’s mindset. I know, it’s obvious. But it’s hard.

A string of words can make perfect sense to someone who already understands the concept, but it’s hard to understand for someone learning it for the first time.

Going from A to B can be a smooth reasoning step for an expert, but requires a few more intermediate steps for a novice.

A trait that helps me in my teaching is my ability to recall the pain points I had when learning a topic. Everything is easy once you know it, but hard when you don’t. Remembering that what comes easily today was once hard is essential for teaching, whatever the format.

I often use my writing to help me with my live teaching. And, just as often, I discover a new angle or insight during live teaching that I then put down in writing. It’s a two-way street. Both forms of communication—live teaching and writing—complement each other.

All this to say that I enjoy writing these articles. They’re useful for me personally, and for my work teaching Python. And I hope they’re useful for you.

121 articles. The cliché would have me say that choosing favourites is like choosing a favourite child. But that’s not the case. There are articles I like less than others. So, I tried to put together a highlights reel of the past three years. Here we go…

The Centre of the Python Universe • Objects

• My Life • The Autobiography of a Python Object

• The One About the £5 Note and the Trip to the Coffee Shop • The Difference Between 'is' and '==' in Python

• When a Duck Calls Out • On Duck Typing and Callables in Python

A Stroll Across Python • Fancy and Not So Fancy Tools

• The Curious Little Shop at The End of My Street • Python’s f-strings

• Do You Really Know How 'or' And 'and' Work in Python?

• If You Find if..else in List Comprehensions Confusing, Read This, Else…

Where Do I Store This? • Data Types and Structures

• Clearing The Deque—Tidying My Daughter’s Soft Toys • A Python Picture Story

• Are Python Dictionaries Ordered Data Structures?

• Hermione’s Undetectable Extension Charm: Revealed (Using Python)

• bytes: The Lesser-Known Python Built-In Sequence • And Understanding UTF-8 Encoding

• Where’s William? How Quickly Can You Find Him? • What’s a Python Hashable Object?

And here are the posts in The Club section of this publication, exclusive for premium subscribers: The Club | The Python Coding Stack

Happy 3rd Birthday to The Python Coding Stack. From just under a hundred people in the first week to 7,600+ today, this community has grown thanks to your enthusiasm.

Let’s keep up the momentum—consider joining The Club today! Your membership can help ensure The Python Coding Stack continues on its path, stronger than ever.

Subscribe now

Photo by Daria Obymaha

For more Python resources, you can also visit Real Python—you may even stumble on one of my own articles or courses there!

Also, are you interested in technical writing? You’d like to make your own writing more narrative, more engaging, more memorable? Have a look at Breaking the Rules.

And you can find out more about me at stephengruppetta.com

What happens when you write coffee.selling_price?

Sure, I know you know. The code returns the value of the .selling_price attribute. But what really happens behind the scenes?

Python operations rely on special methods. The dot you use to access attributes is no exception. But strange and weird things can happen when you use the dot.

Normally, the Product class determines what happens when you write coffee.selling_price. It checks whether the instance has a .selling_price attribute, typically by looking in its .__dict__ dictionary of attributes. If the instance doesn’t have this attribute, Python checks whether it’s a class attribute.

But sometimes the class doesn’t get to decide what happens when you write coffee.selling_price. The object in .selling_price could take control of the attribute lookup.

Let’s explore the weird and wonderful world of descriptors in Python. For such a long time, I struggled to understand descriptors. But, as with many topics, at some stage, everything clicks. The story makes sense. Hopefully, this article can help you make sense of descriptors.

But first, we’ll start with properties.

The Product Class

Let’s create a Product class to deal with products for sale in a shop:

You create a Product instance by providing a name, a cost price, and a selling price. There’s a data attribute for each of these values. There’s also a .profit_margin data attribute.

However, there’s a serious flaw in this code. Have you spotted it?

Let’s try it out:

This code outputs the following:

coffee.profit_margin=2.0

So far, so good. But now try the following update:

Here’s the output from this version:

coffee.profit_margin=2.0

The profit margin still shows as £2. This is incorrect since you modified the selling price to £5.50 soon after creating the Product instance. The cost price is still £2.50. Therefore, the profit margin should be £3.

The .profit_margin data attribute is only calculated once in the .__init__() method, which is called when you create and initialise the object. The code doesn’t update the .profit_margin data attribute when you update the selling price.

You can choose to use a method instead of a data attribute, say .calculate_profit_margin(). However, you think the class’s interface would be neater and easier to use if .profit_margin were still a data attribute like .cost_price and .selling_price.

The solution is to create a property. I promise that this is related to this week’s topic – properties are a great stepping stone towards descriptors:

You start by creating a method called .profit_margin(). Note how you name this in the style of a data attribute rather than a method – it describes data rather than an action. However, you use the decorator @property to create a property object.

Whenever you access the .profit_margin attribute, without the parentheses, Python calls the .profit_margin() method behind the scenes. Therefore, each time you access the attribute .profit_margin, which is now a property, Python calculates and returns the current value:

coffee.profit_margin=3.0

The .profit_margin attribute looks and feels like a data attribute but behaves like a method.

There’s another benefit in this case. This attribute is read-only:

You try to assign a new value to the .profit_margin attribute, but Python raises an error:

Traceback (most recent call last):
  ...
    coffee.profit_margin = 4
    ^^^^^^^^^^^^^^^^^^^^
AttributeError: property ‘profit_margin’ of 
    ‘Product’ object has no setter

In this case, this is the desired behaviour since a user shouldn’t be able to set the profit margin directly. It should be calculated from the cost price and the selling price.

Validating Prices

But let’s move to the central theme of this article. You want to add validation to the prices. For example, you want to ensure they can’t be below zero.

Let’s work on the cost price only for now. The following version won’t work for the same reason you saw earlier:

You add the validation in the .__init__() method. This will apply only to the initialisation process, not to any later changes to the cost price.

So, the validation works fine in this case:

This raises the exception:

Traceback (most recent call last):
  ...
ValueError: Cost price cannot be negative

However, the code won’t stop you from changing the cost price to a negative value later:

Here’s the output:

coffee.cost_price=-1

Solution? Create a property:

Note that I removed the validation code from the .__init__() method. The definition of the .cost_price data attribute in .__init__() has also gone.

But what do you include in the cost_price() method? Unlike the .profit_margin property, which can be calculated from existing attributes, you still need somewhere to store the current value for the cost price. You still need a data attribute that you create in the .__init__() method.

You can create a non-public data attribute:

The data attribute ._cost_price has a leading underscore, which is a Python convention indicating that the user shouldn’t access this attribute directly. Python doesn’t prevent the user from accessing this non-public attribute, but the leading underscore shows the programmer’s intent that this attribute should be left alone when using the class.

Whenever you access the .cost_price property, Python executes the code you define in the method that you decorate with the @property decorator. In this case, Python returns the value of the ._cost_price data attribute.

Therefore, you still have a data attribute – the one with the leading underscore. However, the user doesn’t access it directly. Instead, they access it through the property .cost_price:

Here’s the output from this code:

coffee.cost_price=2.5

However, this makes the cost price read-only. That’s not what you’re trying to achieve:

This raises an error since you can’t assign values to the .cost_price property:

Traceback (most recent call last):
  ...
    coffee.cost_price = 3
    ^^^^^^^^^^^^^^^^^
AttributeError: property ‘cost_price’ of 
    ‘Product’ object has no setter

Note what the error message says: 'Product' object has no setter. The setter is the method that allows you to set the value of a property. But you can add this setter to your code:

There are no issues with setting the value for coffee.cost_price now that you have a setter, too. And the validation works when you try to set the cost price value to a negative value:

This raises an error:

Traceback (most recent call last):
  ...
    coffee.cost_price = -3
    ^^^^^^^^^^^^^^^^^
ValueError: Cost price cannot be negative

However, there’s still a flaw in this code. The negative cost price validation still fails in some circumstances. Can you figure out when it fails?

Try this code:

Here’s the output:

coffee.cost_price=-2.5

Validation occurs only when you set the cost_price attribute. However, look carefully at the code in your .__init__() method. It still sets the actual data attribute ._cost_price directly. This bypasses the verification.

Luckily, there’s an easy fix:

You can use the property in the .__init__() method, too. This triggers the property’s setter, which creates the ._cost_price data attribute and stores the value in it.

Avoiding float for prices

Let’s add a bit more to the validation code. It’s best not to use float when dealing with prices since floats suffer from rounding errors, which may lead to issues when dealing with money (especially if the rounding goes against you!)

Here’s the classic example showing the floating-point rounding errors:

Some floats cannot be represented precisely in binary, leading to rounding errors.

One option is to work in cents or pence or whatever the smaller unit of the currency you’re using. This means you’re working with integers. However, an alternative is to use decimal.Decimal instead:

Note that you should pass the floats as strings to Decimal and let the Decimal class deal with that. If you use float objects, you’ve already included the rounding errors in your starting value!

Let’s make sure the cost price uses Decimal internally without forcing the user to use this data type:

In true duck typing fashion, rather than choosing which data types you’re willing to accept when setting the cost price, you try to convert the input value to Decimal. Any data type that can be converted to Decimal is acceptable. The Decimal type raises an InvalidOperation exception, defined within decimal, when you pass an unacceptable type. So, you use InvalidOperation in the except block.

You probably want to perform the same validation for the selling price, too:

This works. However, there’s a lot of boilerplate code. And the validation is repeated in two setter methods. Sure, you can create a helper function to avoid the repetition. But what if you had more types of prices to include in your class? Perhaps there’s a wholesale_price and a discounted_price, say. You’d have to create properties for each one.

Before I move on from properties, here’s a link to another article I wrote about properties: The Properties of Python’s property .

Now let’s explore a different option.

Descriptors

Spoiler alert: you’ve already been using descriptors. Python’s property uses the descriptor protocol behind the scenes. Indeed, you use descriptors each time you access a method, or when you use the @classmethod and @staticmethod decorators.

But let’s get back to what happens when you use the dot after an object name, such as coffee.cost_price. The quick answer often states that Python looks at the instance’s .__dict__ attribute, which is a dictionary containing the instance’s data attributes. If the attribute is not there, Python looks in the class’s .__dict__ to check whether the attribute is a class attribute.

However, there’s a bit more happening.

The . notation calls the .__getattribute__() special method, which looks for the attribute using a broader set of rules and priorities.

The first thing it looks for is whether the attribute is a descriptor – technically, a data descriptor.

But what’s a descriptor?

Creating a descriptor class

Let’s create one. Create a Price class:

An object that has either of these special methods is a descriptor. (Note that there’s another special method that’s part of the descriptor protocol, .__delete__(), which is also sufficient to create a descriptor. I won’t be discussing .__delete__() in this article.)

So what? This is where descriptors start getting weird – yes, right at the start! Let’s pause with this class and get back to the Product class. Define .cost_price and .selling_price as class attributes and assign instances of the new Price class to them:

The rest of the code remains the same except that four methods are gone. These are the getter and setter methods you had defined previously to ensure that .cost_price and .selling_price are properties. You no longer need them. You’ll see why soon.

Why define class attributes? Class attributes are created when you define the class, unlike instance attributes, which are created when you create instances. Therefore, the attributes .cost_price and .selling_price exist and are equal to instances of a descriptor class at the time of class definition. This matters, as you’ll also see shortly.

So, what’s special about descriptors? Remember when we discussed how the . triggers the .__getattribute__() special method, which is responsible for looking for the attribute and returning its value? Well, as it happens, Python doesn’t look in the instance’s .__dict__ attribute first, after all.

The first thing it checks is whether the attribute is a descriptor. More specifically, it checks whether the attribute is a data descriptor, which is a descriptor that has .__set__() or .__delete__(). But our descriptor class, Price, has a .__set__(), so this priority applies to a Price object.

Here’s the key to understanding descriptors:

Since .cost_price and .selling_price are descriptors, the Product class defers to the Price class to decide how to get or set a value for this attribute.

Let me repeat that.

Normally, the class that defines an instance determines its behaviour. The special methods in Product determine how a Product instance behaves, such as what its string representations are, whether it’s iterable, whether it has a length, and so on. We haven’t defined any of these special methods in Product, but if we did, they would determine the object’s behaviour in those contexts.

However, when you use the dot on an instance of a Product class, if the attribute you’re accessing is a descriptor, such as .cost_price and .selling_price, it’s Price.__get__() and Price.__set__() that determine what to do when you get or set the values of .cost_price and .selling_price.

Code in the Price class determines what happens when you access an attribute in a Product instance.

And since you define .cost_price and .selling_price as class attributes that contain descriptors – instances of Price – then the descriptor special methods kick in even in Product.__init__() when you write self.cost_price = cost_price and self.selling_price = selling_price.

Understanding how descriptors work

Let’s prove this by adding some code to the Price class. This code is here to let us track what’s happening. You’ll write “proper” code later:

The .__get__() and .__set__() methods print out phrases to allow you to see when these methods are called. The .__get__() special method also returns a string. A reminder that these special methods are defined in the Price class.

Here’s the code so far, which includes a statement in the main scope that creates an instance of the Product class:

Here’s what you see when you run this code:

Calling Price.__set__()
Calling Price.__set__()

The code calls the Price.__set__() method twice: once when creating self.cost_price and another time when creating self.selling_price in Product.__init__().

Let’s also fetch one of the attributes:

Here’s the output from this code:

Calling Price.__set__()
Calling Price.__set__()
Calling Price.__get__()
Price coming soon...

You still get the two “breadcrumbs” that show that the code called Price.__set__() twice. But now you can also see that the code calls Price.__get__() since you wrote print(coffee.selling_price). And when you access the .selling_price attribute, Python returns the value returned by Price.__get__(), which is the string saying "Price coming soon...".

Therefore, attribute access for .selling_price and .cost_price, which are attributes in Product, is now controlled by the Price class, which is the descriptor class.

Yes, the Price class has taken over control of attribute access in the Product class (but only for the attributes that are descriptors of the Price class, .cost_price and .selling_price.)

Let’s look at the signatures of the .__get__() and .__set__() special methods to figure out how they work. Let’s focus on .__get__() first. Here’s the method’s signature and a few more printouts to help us figure out what’s what:

There are three parameters defined in the signature for .__get__(). You won’t be surprised to see self as the first parameter, which is the case for all instance methods in classes. There are also two other parameters named instance and owner. Let’s figure out what they are.

You’ve seen earlier that the current code calls .__get__() once when the code evaluates coffee.selling_price, which you use as an argument for print(). Here’s the output of the code now:

Calling Price.__set__()
Calling Price.__set__()
Calling Price.__get__()
self=<__main__.Price object at 0x100bc8f50>
instance=<__main__.Product object at 0x100bd9160>
owner=<class ‘__main__.Product’>
Price coming soon...

The new additions are the printouts on the fourth, fifth, and sixth lines showing the values assigned to the parameters self, instance, and owner in Price.__get__(). First, a reminder that this method is triggered when you run the following expression:

Here’s what the outputs for the three parameters show:

• self=<__main__.Price object ...>: Therefore, self refers to the instance of the Price class itself, as self always does in classes. This is the descriptor instance. It’s the same descriptor object the attribute coffee.selling_price invokes (but not the object that coffee.selling_price returns, as you’ll see soon.)

• instance=<__main__.Product object ...>: This output shows that instance is an object of type Product. It’s the object that has the descriptor as one of its attributes. In this example, this is the coffee object.

• owner=<__main__.Product>: This is the Product class itself (not an instance of the class).

Therefore, in the expression coffee.selling_price, coffee is the Product instance and .selling_price is the Price instance, which is the descriptor.

And when Python evaluates the expression coffee.selling_price, it defers the decision of which value to display to the Price class. Python calls Price.__get__() with the following arguments:

• The selling_price descriptor is passed to self

• The coffee object, an instance of Product, is passed to instance

• The Product class is passed to owner

I’ve repeated myself in the previous few paragraphs, but that’s deliberate. I get confused following the flow from the Product class to calling a method in the descriptor class, Price. I need this repetition. Maybe it helps you, too.

As a side note, which I’ll mostly ignore in this article: It’s possible to call the descriptor using the class rather than an instance, such as the following:

Here’s the output from this code:

Calling Price.__get__()
self=<__main__.Price object at 0x102a79090>
instance=None
owner=<class ‘__main__.Product’>
Price coming soon...

In this case, None is assigned to instance since there’s no Product instance involved.

Let’s look at the .__set__() special method now:

The signature still includes self and instance, and these refer to the same objects as they did in .__get__(). The third parameter is value, which stores the object you want to use when setting the attribute’s value. Let’s focus on creating an instance of Product as you did earlier:

This code accesses descriptors using the dot notation twice within Product.__init__() since it sets the values of .cost_price and .selling_price, which contain Price objects. Here’s the output from this code:

Calling Price.__set__()
self=<__main__.Price object at 0x104cf5010>
instance=<__main__.Product object at 0x104cf5160>
value=3.2

Calling Price.__set__()
self=<__main__.Price object at 0x104ce4f50>
instance=<__main__.Product object at 0x104cf5160>
value=5

The two calls are triggered by two different descriptors, .cost_price and .selling_price. Therefore, self refers to different objects in these calls, as you can see from the different identity values displayed.

However, the descriptors belong to the same instance, the Product object you named coffee. You can see that the printout references the same Product object on both occasions.

The value is the number you’re trying to assign to the descriptors.

Writing the .__get__() and .__set__() methods

Now you know how and when Python calls the .__get__() and .__set__() methods, you can add code within them. You can use the same reasoning you used when creating properties earlier.

To keep track of what’s happening, let’s build the code you need in stages.

First, let’s focus exclusively on the .cost_price attribute. You’ll write the validation as if it’s just for cost price — and you’ll soon see why that breaks once selling price joins the party. You’ll deal with .selling_price later.

You can use the same validation code you used earlier when working with properties. This code should now go in the .__set__() method within the Price class:

Let’s see what’s new in this version:

1. I removed most of the print() calls, which were useful merely to explore what objects are passed to these special methods. I left the ones showing when the code calls these methods for now.

2. The .__set__() method includes the try..except block, which ensures the value is a Decimal, and the if statement, which checks that the value is not negative.

3. The final line in .__set__() sets the ._cost_price attribute in the instance object, using the value in value. The function call setattr(instance, "_cost_price", value) is equivalent to instance._cost_price = value. In this case, instance is the Product object, such as coffee. There’s a good reason to use the setattr() built-in function rather than assigning the value directly using the dot notation. You’ll find out later when you also deal with the selling price.
   Note that you use the leading underscore in "_cost_price" since you wish to create the non-public attribute ._cost_price rather than assign the value to the .cost_price attribute (the one without the leading underscore).

Let’s try this out. Note that this code is hard-coded to work only for the .cost_price attribute for now:

Oops! This doesn’t work:

Calling Price.__set__()
Calling Price.__set__()
Calling Price.__get__()
None

The final line shows that the code outputs None. That’s because coffee.cost_price triggers the Price.__get__() method, which you haven’t completed yet! You can still check whether we’re on track by printing the actual data attribute ._cost_price:

You should avoid accessing non-public attributes outside of the class definition in general. But I give you permission to break the rule here since we’re merely exploring what’s happening behind the scenes!

Calling Price.__set__()
Calling Price.__set__()
5

Note how there’s no printout saying "Calling Price.__get__()" now since you’re accessing the data attribute ._cost_price, not the descriptor .cost_price.

But did you notice that this still shows the wrong value? The cost price should be £3.20. It’s the selling price that’s £5.00. Remember how you’re hard-coding the string "_cost_price" in the call to setattr()? This means that when you try to set the selling price, the code currently overwrites the cost price. But don’t worry, we’ll fix this soon.

Time to focus on .__get__(). Since you created the non-public data attribute ._cost_price when setting the value using .__set__(), you can now get the value of that attribute using getattr():

Recall that in .__get__(), the parameter instance refers to the Product instance, such as coffee. Therefore, getattr(instance, "_cost_price") is equivalent to coffee._cost_price in this example.

Let’s try this out, this time using the .cost_price descriptor directly when accessing the value of the cost price:

Here’s the output showing all the breadcrumbs confirming that two attributes are set and that you fetch the value of one of them once. The output also shows the value of the attribute, which is still wrong since £5 is the selling price:

Calling Price.__set__()
Calling Price.__set__()
Calling Price.__get__()
5

We’ll soon stop the selling price from overwriting the cost price.

But first, let’s check that the validation logic works:

This raises an error:

Traceback (most recent call last):
  ... in <module>
    coffee = Product(”Espresso”, -3.2, 5)
  ... in Product.__init__
    self.cost_price = cost_price
    ^^^^^^^^^^^^^^^
  ... in Price.__set__
    raise ValueError(f”’cost_price’ can’t be less than zero”)
    
ValueError: ‘cost_price’ can’t be less than zero

And so does this:

This time, the error message is different:

...
TypeError: Invalid format for ‘cost_price’: three point two

Another minor note for completeness’s sake. We mentioned that you can call the descriptor through the class, such as Product.cost_price. Since instance is None in this case, the code in .__get__() raises an error since None doesn’t have a ._cost_price attribute (or any attribute, for that matter!) To account for this, you can add the following code in .__get__():

If you call the descriptor from a class, you just get the descriptor back. There’s no instance, so there’s no value to fetch in this case! But let’s move on…

Your call…

The Python Coding Place offers something for everyone:

• a super-personalised one-to-one 6-month mentoring option
$ 4,750

• individual one-to-one sessions
$ 125

• a self-led route with access to 60+ hrs of exceptional video courses and a support forum
$ 400

Which The Python Coding Place student are you?

Making a general descriptor class

This works. But it only works for the .cost_price attribute since there are a few references to this attribute hard-coded in the .__get__() and .__set__() methods:

1. You use the string "_cost_price" when calling getattr() and setattr().

2. You also refer to the attribute name 'cost_price' in the error messages.

Hard-coding the cost price attribute names also has the undesirable effect of overwriting the cost price with the selling price!

Ideally, you’d like to access the public attribute name "cost_price" (the name of the descriptor) and the non-public data attribute name "_cost_price" dynamically. Recall that these are attributes in the Product class. Python passes the descriptor object self to the .__get__ and .__set__() methods, but not its attribute name within Product.

But descriptors can also deal with this problem. A descriptor allows you to pass the attribute name from the Product class to the Price class using .__set_name__() in the descriptor class:

I’ve put similar print() calls to the ones you used earlier to help us identify what’s what. Remove the code that creates the instance coffee, so that your module only defines the two classes, Price and Product, and nothing else. Run your script, and you’ll see this output:

Calling Price.__set_name__()
self=<__main__.Price object at 0x101bbcec0>
owner=<class ‘__main__.Product’>
name=’cost_price’

Calling Price.__set_name__()
self=<__main__.Price object at 0x101bacf50>
owner=<class ‘__main__.Product’>
name=’selling_price’

Python calls Price.__set_name__ when you create the descriptors within Product. Recall that you create the descriptors when you assign them to the class attributes. These attributes, and therefore the descriptors, are created when you define the Product class and not when you create Product instances.

As before, self refers to the descriptor object itself. The owner parameter still refers to the class in which the descriptors are defined, Product in this example.

Finally, name is a string containing the attribute name you use within the Product class. Therefore, name is either "cost_price" or "selling_price".

You can now use this name within code in the Price class. First, let’s tidy up the .__set_name__() method:

You create two instance attributes in the Price object to refer to the private and public names. You may recall that I mentioned Python prefers the term “non-public” over “private”. However, the Python documentation also occasionally uses the term “private”, including when discussing descriptors. It’s common to see .private_name used as the instance attribute within a descriptor. I’ll acquiesce.

Now, you can update the rest of the code:

You use self.private_name as an argument in getattr() and setattr() instead of the hard-coded string you had before. And you use self.public_name in the error messages. Note that I’ve also removed the breadcrumb calls to print() since we no longer need to show when Python calls each of these methods.

Let’s test:

You can set and get values, including within the .profit_margin property:

3.2
5
1.8

But you can’t set invalid values. Try coffee.cost_price = -2 and coffee.cost_price = "three point two" to confirm the validation code works.

Here’s the full code containing both classes:

How About Properties, Then?

You use a property for the .profit_margin attribute in Product. This is a simple read-only attribute you use to calculate the value each time you need the attribute. You could use properties for .selling_price and .cost_price, but there’s plenty of boilerplate code needed for each attribute. Annoying, especially if you need many price-related attributes in your class, not just cost and selling prices.

Descriptors allow you to put the validation logic into a separate class. And descriptors aren’t restricted to data validation. You can use descriptors whenever you need to execute any logic each time you get or set the value of an attribute. And you can use the Price descriptor elsewhere, too. It’s not restricted to the Product class.

But remember the property object? This is itself a descriptor. Here’s evidence of this:

This code returns True. The .profit_margin attribute is a property. It has the .__get__() special method, which is sufficient to make it a descriptor. A property is a descriptor.

Note that you’re accessing this property through the class name within hasattr(). Remember the code checking whether instance is None in .__get__()? You used the if instance is None: check to return the descriptor itself when you call the descriptor from a class. That’s why you can introspect the descriptor through Product.profit_margin. But don’t worry too much about this detail!

Final Words

How often have you used a method, such as in this case?

Millions of times, you say? You may already be aware that this call to the .append() method is similar to this more verbose version:

In this version, you use append() as a function. It’s a function that’s an attribute of the list class. Therefore, you pass the object as its first argument – this is the object that is assigned to the first parameter self in the append() signature.

Well, as it happens, Python uses the descriptor protocol to automatically equate numbers.append(30) to the function call list.append(numbers, 30).

Functions are objects. And function objects are descriptors in Python. When you access them through an instance of a class, such as in numbers.append(30), Python goes through the function’s .__get__() method and calls list.append(numbers, 30) so that you don’t have to worry about it:

Functions have a .__get__() method, which makes them descriptors. Since you can’t set the value for a function object, they don’t have a .__set__() method. Descriptors that only have a .__get__() special method, such as functions, are non-data descriptors. Non-data descriptors don’t get the same precedence as data descriptors when Python is figuring out what to do when you use the dot notation. But that’s a topic for another day…

As with many topics, learning about descriptors helps you learn about descriptors (duh!), but it also helps you understand Python better.

So, descriptors are a useful tool whenever you want to perform additional tasks, such as validation, whenever you get or set the value of an attribute. What makes them a bit weird until you get used to them is that the logic is defined in the descriptor class itself. The benefit of this is that you can use descriptors across many classes and rely on the same logic, without having to define that logic in each class.

For example, you can use Price() objects, which are descriptors, in any class where you need to use prices and apply the same validation you defined in this article. And you can add functions to any class safe in the knowledge that the descriptor protocol will ensure they’re treated as instance methods when they’re accessed (unless you use the @classmethod or @staticmethod decorators, which – you guessed it – also use the descriptor protocol to perform their magic.)

In summary, a descriptor is an object that hijacks the attribute access process triggered by the dot notation. Normally, when you access an attribute of an object, the object’s class deals with finding the value. However, when the attribute is a descriptor, the descriptor class takes over.

Did you enjoy this article? Did you understand how descriptors work? That’s not an easy feat!

Do you want to become a premium subscriber and join The Club to help support this publication and ensure there are more articles like this one?

Subscribe now

Photo by Ej Agumbay

Code in this article uses Python 3.14

The code images used in this article are created using Snappify. [Affiliate link]

For more Python resources, you can also visit Real Python—you may even stumble on one of my own articles or courses there!

Also, are you interested in technical writing? You’d like to make your own writing more narrative, more engaging, more memorable? Have a look at Breaking the Rules.

And you can find out more about me at stephengruppetta.com

Further reading related to this article’s topic:

• The Properties of Python’s property

Appendix: Code Blocks

Code Block #1

class Product:
    def __init__(self, name, cost_price, selling_price):
        self.name = name
        self.cost_price = cost_price
        self.selling_price = selling_price
        self.profit_margin = self.selling_price - self.cost_price

Code Block #2

# ...
coffee = Product("Espresso", 2.5, 4.5)
print(f"{coffee.profit_margin=}")

Code Block #3

# ...
coffee = Product("Espresso", 2.5, 4.5)
coffee.selling_price = 5.5
print(f"{coffee.profit_margin=}")

Code Block #4

class Product:
    def __init__(self, name, cost_price, selling_price):
        self.name = name
        self.cost_price = cost_price
        self.selling_price = selling_price

    @property
    def profit_margin(self):
        return self.selling_price - self.cost_price

coffee = Product("Espresso", 2.5, 4.5)
coffee.selling_price = 5.5
print(f"{coffee.profit_margin=}")

Code Block #5

# ...
coffee.profit_margin = 4

Code Block #6

class Product:
    def __init__(self, name, cost_price, selling_price):
        self.name = name
        if cost_price < 0:
            raise ValueError("Cost price cannot be negative")
        self.cost_price = cost_price
        self.selling_price = selling_price

    @property
    def profit_margin(self):
        return self.selling_price - self.cost_price
      
# ...

Code Block #7

# ...
coffee = Product("Espresso", -2.5, 4.5)

Code Block #8

# ...
coffee = Product("Espresso", 2.5, 4.5)  # Back to correct cost price
coffee.cost_price = -1
print(f"{coffee.cost_price=}")

Code Block #9

class Product:
    def __init__(self, name, cost_price, selling_price):
        self.name = name
        self.selling_price = selling_price
        
    @property
    def cost_price(self):
        # ???
        ...

    @property
    def profit_margin(self):
        return self.selling_price - self.cost_price
      
# ...

Code Block #10

class Product:
    def __init__(self, name, cost_price, selling_price):
        self.name = name
        self._cost_price = cost_price
        self.selling_price = selling_price

    @property
    def cost_price(self):
        return self._cost_price

    @property
    def profit_margin(self):
        return self.selling_price - self.cost_price
      
# ...

Code Block #11

# ...
coffee = Product("Espresso", 2.5, 4.5)
print(f"{coffee.cost_price=}")

Code Block #12

# ...
coffee = Product("Espresso", 2.5, 4.5)
coffee.cost_price = 3
print(f"{coffee.cost_price=}")

Code Block #13

class Product:
    def __init__(self, name, cost_price, selling_price):
        self.name = name
        self._cost_price = cost_price
        self.selling_price = selling_price

    @property
    def cost_price(self):
        return self._cost_price

    @cost_price.setter
    def cost_price(self, value):
        if value < 0:
            raise ValueError("Cost price cannot be negative")
        self._cost_price = value

    @property
    def profit_margin(self):
        return self.selling_price - self.cost_price

coffee = Product("Espresso", 2.5, 4.5)
coffee.cost_price = 3
print(f"{coffee.cost_price=}")

Code Block #14

# ...
coffee = Product("Espresso", 2.5, 4.5)
coffee.cost_price = -3
print(f"{coffee.cost_price=}")

Code Block #15

# ...
coffee = Product("Espresso", -2.5, 4.5)
print(f"{coffee.cost_price=}")

Code Block #16

class Product:
    def __init__(self, name, cost_price, selling_price):
        self.name = name
        self.cost_price = cost_price
        self.selling_price = selling_price

    @property
    def cost_price(self):
        return self._cost_price

    @cost_price.setter
    def cost_price(self, value):
        if value < 0:
            raise ValueError("Cost price cannot be negative")
        self._cost_price = value

    @property
    def profit_margin(self):
        return self.selling_price - self.cost_price
      
# ...

Code Block #17

0.1 + 0.2
# 0.30000000000000004

Code Block #18

from decimal import Decimal
Decimal("0.1") + Decimal("0.2")
# Decimal('0.3')

Code Block #19

from decimal import Decimal, InvalidOperation

class Product:
    def __init__(self, name, cost_price, selling_price):
        self.name = name
        self.cost_price = cost_price
        self.selling_price = selling_price

    @property
    def cost_price(self):
        return self._cost_price

    @cost_price.setter
    def cost_price(self, value):
        try:
            value = Decimal(str(value))
        except InvalidOperation:
            raise TypeError("Invalid format for cost price")
        if value < 0:
            raise ValueError("Cost price cannot be negative")
        self._cost_price = value

    @property
    def profit_margin(self):
        return self.selling_price - self.cost_price
      
# ...

Code Block #20

from decimal import Decimal, InvalidOperation

class Product:
    def __init__(self, name, cost_price, selling_price):
        self.name = name
        self.cost_price = cost_price
        self.selling_price = selling_price

    @property
    def cost_price(self):
        return self._cost_price

    @cost_price.setter
    def cost_price(self, value):
        try:
            value = Decimal(str(value))
        except InvalidOperation:
            raise TypeError("Invalid format for cost price")
        if value < 0:
            raise ValueError("Cost price cannot be negative")
        self._cost_price = value

    @property
    def selling_price(self):
        return self._selling_price

    @selling_price.setter
    def selling_price(self, value):
        try:
            value = Decimal(str(value))
        except InvalidOperation:
            raise TypeError("Invalid format for selling price")
        if value < 0:
            raise ValueError("Selling price cannot be negative")
        self._selling_price = value

    @property
    def profit_margin(self):
        return self.selling_price - self.cost_price

Code Block #21

# ...

class Price:
    """Descriptor to handle price validation and Decimal conversion."""
    
    def __get__(self, instance, owner):
        ...

    def __set__(self, instance, value):
        ...
        
# ...

Code Block #22

# ...

class Product:
    cost_price = Price()
    selling_price = Price()

    def __init__(self, name, cost_price, selling_price):
        self.name = name
        self.cost_price = cost_price
        self.selling_price = selling_price

    @property
    def profit_margin(self):
        return self.selling_price - self.cost_price

# ...

Code Block #23

# ...

class Price:
    def __get__(self, instance, owner):
        print("Calling Price.__get__()")
        return "Price coming soon..."

    def __set__(self, instance, value):
        print("Calling Price.__set__()")

# ...

Code Block #24

class Price:
    def __get__(self, instance, owner):
        print("Calling Price.__get__()")
        return "Price coming soon..."

    def __set__(self, instance, value):
        print("Calling Price.__set__()")

class Product:
    cost_price = Price()
    selling_price = Price()

    def __init__(self, name, cost_price, selling_price):
        self.name = name
        self.cost_price = cost_price
        self.selling_price = selling_price

    @property
    def profit_margin(self):
        return self.selling_price - self.cost_price


coffee = Product("Espresso", 3.2, 5)

Code Block #25

# ...
coffee = Product("Espresso", 3.2, 5)
print(coffee.selling_price)

Code Block #26

class Price:
    def __get__(self, instance, owner):
        print("Calling Price.__get__()")
        print(f"{self=}")
        print(f"{instance=}")
        print(f"{owner=}")
        return "Price coming soon..."
      
      # ...

Code Block #27

coffee.selling_price

Code Block #28

# ...
print(Product.selling_price)

Code Block #29

class Price:
    # ...

    def __set__(self, instance, value):
        print("Calling Price.__set__()")
        print(f"{self=}")
        print(f"{instance=}")
        print(f"{value=}")
# ...

Code Block #30

# ...
coffee = Product("Espresso", 3.2, 5)

Code Block #31

from decimal import Decimal, InvalidOperation

class Price:
    def __get__(self, instance, owner):
        print("Calling Price.__get__()")
        # more code coming soon here

    def __set__(self, instance, value):
        print("Calling Price.__set__()")
        try:
            value = Decimal(str(value))
        except InvalidOperation:
            raise TypeError(f"Invalid format for 'cost_price': {value}")

        if value < 0:
            raise ValueError(f"'cost_price' can't be less than zero")

        setattr(instance, "_cost_price", value)

# ...

Code Block #32

# ...

coffee = Product("Espresso", 3.2, 5)
print(coffee.cost_price)

Code Block #33

# ...

coffee = Product("Espresso", 3.2, 5)
print(coffee._cost_price)

Code Block #34

from decimal import Decimal, InvalidOperation

class Price:
    def __get__(self, instance, owner):
        print("Calling Price.__get__()")
        return getattr(instance, "_cost_price")

    def __set__(self, instance, value):
        print("Calling Price.__set__()")
        try:
            value = Decimal(str(value))
        except InvalidOperation:
            raise TypeError(f"Invalid format for 'cost_price': {value}")

        if value < 0:
            raise ValueError(f"'cost_price' can't be less than zero")

        setattr(instance, "_cost_price", value)

Code Block #35

# ...

coffee = Product("Espresso", 3.2, 5)
print(coffee.cost_price)

Code Block #36

# ...

coffee = Product("Espresso", -3.2, 5)

Code Block #37

# ...
coffee = Product("Espresso", "three point two", 5)

Code Block #38

from decimal import Decimal, InvalidOperation

class Price:
    def __get__(self, instance, owner):
        print("Calling Price.__get__()")
        if instance is None:
            return self
        return getattr(instance, "_cost_price")
    # ...

Code Block #39

from decimal import Decimal, InvalidOperation

class Price:
    def __set_name__(self, owner, name):
        print("Calling Price.__set_name__()")
        print(f"{self=}")
        print(f"{owner=}")
        print(f"{name=}")
        
		# ...

Code Block #40

from decimal import Decimal, InvalidOperation

class Price:
    def __set_name__(self, owner, name):
        print("Calling Price.__set_name__()")
        self.public_name = name
        self.private_name = f"_{name}"

# ...

Code Block #41

from decimal import Decimal, InvalidOperation

class Price:
    def __set_name__(self, owner, name):
        self.public_name = name
        self.private_name = f"_{name}"

    def __get__(self, instance, owner):
        if instance is None:
            return self
        return getattr(instance, self.private_name)

    def __set__(self, instance, value):
        try:
            value = Decimal(str(value))
        except InvalidOperation:
            raise TypeError(
                f"Invalid format for '{self.public_name}': {value}"
            )

        if value < 0:
            raise ValueError(
                f"'{self.public_name}' can't be less than zero"
            )

        setattr(instance, self.private_name, value)

# ...

Code Block #42

# ...
coffee = Product("Espresso", 3.2, 5)
print(coffee.cost_price)
print(coffee.selling_price)
print(coffee.profit_margin)

Code Block #43

from decimal import Decimal, InvalidOperation

class Price:
    def __set_name__(self, owner, name):
        self.public_name = name
        self.private_name = f"_{name}"

    def __get__(self, instance, owner):
        if instance is None:
            return self
        return getattr(instance, self.private_name)

    def __set__(self, instance, value):
        try:
            value = Decimal(str(value))
        except InvalidOperation:
            raise TypeError(
                f"Invalid format for '{self.public_name}': {value}"
            )

        if value < 0:
            raise ValueError(
                f"'{self.public_name}' can't be less than zero"
            )

        setattr(instance, self.private_name, value)

class Product:
    cost_price = Price()
    selling_price = Price()

    def __init__(self, name, cost_price, selling_price):
        self.name = name
        self.cost_price = cost_price
        self.selling_price = selling_price

    @property
    def profit_margin(self):
        return self.selling_price - self.cost_price

Code Block #44

# ...
print(hasattr(Product.profit_margin, "__get__"))

Code Block #45

numbers = [10, 20]
numbers.append(30)
numbers
# [10, 20, 30]

Code Block #46

numbers = [10, 20]
list.append(numbers, 30)
numbers
# [10, 20, 30]

Code Block #47

def some_function():
    ...
hasattr(some_function, "__get__")
# True
hasattr(some_function, "__set__")
# False

Join The Club, the exclusive area for paid subscribers for more Python posts, videos, a members’ forum, and more.

Subscribe now

You can also support this publication by making a one-off contribution of any amount you wish.

Support The Python Coding Stack

For more Python resources, you can also visit Real Python—you may even stumble on one of my own articles or courses there!

Also, are you interested in technical writing? You’d like to make your own writing more narrative, more engaging, more memorable? Have a look at Breaking the Rules.

And you can find out more about me at stephengruppetta.com

But today, I’ll write a short post about some tools I came across that may be interesting. I’ll keep the post short since, if you’re interested, you can easily explore the packages independently – you won’t need my explanatory efforts.

Remember some of the earliest code you ever wrote? It may have looked like this:

first = input(”Enter the first number: “)
second = input(”Enter the second number: “)
print(
    f”The sum of the two numbers is {float(first) + float(second)}”
)

Those days are long gone. But what if you wanted this code to work for any number of inputs, not just two?

I’m not talking about the code to work out the sum itself. You’d probably use a list to collect all the numbers and the sum() built-in function, or just use a running total. Whatever.

The annoying part is making user-friendly strings when asking for the input – it’s fine to write "first" and "second" when there are only two prompts – and then when showing the result, which currently says "two numbers". But what if you have more?

I came across this need several times, but my solution has always been to change the string so that it’s general enough to work in all cases. I’m lazy, I know.

In any case, recently I decided to look for solutions. And of course, they exist! Everything seems to exist in the Python ecosystem. Several solutions…

num2words

The first solution is probably the simplest: the num2words package doesn’t do much beyond converting numbers to words, as its name suggests. You’ll need to install num2words using pip, uv, or whatever you use to install packages.

>>> import num2words
>>> num2words.num2words(34)
‘thirty-four’

The main function in the num2words module is also called num2words(), as often happens with such modules!

How far can it go?

>>> num2words.num2words(5468.23)
‘five thousand, four hundred and sixty-eight point two three’

Working in Spanish?

>>> num2words.num2words(5468.23, lang=”es”)
‘cinco mil cuatrocientos sesenta y ocho punto dos tres’

Or Japanese?

>>> num2words.num2words(5468.23, lang=”ja”)
‘五千四百六十八点二三’

Read more

In this post, I’ll discuss the OOP mindset–or, let’s say, one way of viewing the object-oriented paradigm. I debated with myself whether to write this article here on The Python Coding Stack. Most articles here are aimed at the “intermediate-ish” Python programmer, whatever “intermediate” means. Most readers may feel they already understand the ethos and philosophy of OOP. I know I keep discovering new perspectives from time to time. So here’s a perspective I’ve been exploring in courses I ran recently.

As you can see, I decided to write this post. At worst, it serves as revision for some readers, perhaps a different route towards understanding why we (sometimes) define classes and (always) use objects in Python. And beginners read these articles, too!

If you feel you’re an OOP Pro, go ahead and skip this post. Or just read it anyway. It’s up to you. You can always catch up with anything you may have missed from The Python Coding Stack’s Archive instead – around 120 full-length articles and counting!

This post is inspired by the introduction to OOP in Chapter 7 in The Python Coding Book – The Relaxed and Friendly Programming Book for Beginners

Meet Mark, a market seller. He sets up a stall in his local market and sells tea, coffee, and home-made biscuits (aka cookies for those in North America).

I’ll write two and a half versions of the code Mark could use to keep track of his stock and sales. The first one–that’s the half–is not very pretty. Don’t write code like this. But it will serve as a starting point for the main theme of this post. The second version is a stepping stone towards the third.

First (Half) Version • Not Pretty

Mark is also learning Python in his free time. He starts writing some code:

He was planning to write some functions to update the stock, change the cost price and selling price, and deal with a sale from his market stall.

But he stopped.

We’ll also stop here with this version.

Mark is a Python beginner, but even he knew this was not the best way to store the relevant data for each product he sells.

Four separate objects? Not ideal. Mark knows that the data in these four lists are linked to each other. The first items in each list belong together, and so on. But he’ll need to manually ensure he maintains these links in the code he writes. Not impossible, but it’s asking for trouble. So many things can go wrong. And he’ll have a tough time writing the code, making all those links in every line he writes.

Mark knows what the link is between the various values. But the computer program doesn’t. The computer program sees four separate data structures, unrelated to each other. You can try writing an update_stock() function with this version to see the challenges, the manual work needed to connect data across separate structures.

Let’s move on.

Second Version • Dictionary and Functions

Mark learnt about dictionaries. They’re a great way to group data together:

Now there’s one data structure that contains the three products he sells. This structure contains three other data structures, each containing all the relevant data for each product.

The data are structured into these dictionaries to show what belongs where. In this version, the Python program “knows” which data items belong together. Each dictionary contains related values.

Well done, Mark! He understood the need to organise the data into a sensible structure. This is not the only combination of nested data structures Mark could use, but this is a valid option. Certainly better than the first version!

He can write some functions now. First up is update_stock():

Let’s ignore the fact that Mark is accessing the global variable products within the function. We’ll have a chat with Mark about this.

Still, there’s now a function called update_stock(). It needs the product name as an argument. It also needs the stock amount to increase (or decrease if the value is negative).

Let’s look at another function Mark wrote:

This function also needs the product name as one of its arguments. It also needs a second argument: the quantity sold.

Mark wrote other functions, but I’ll keep this section brief so I won’t show them. However, many of Mark’s functions have a few things in common:

• They require the product name as one of the arguments. This makes sense since the operations Mark needs to carry out depend on which product he’s dealing with.

• They make changes to the data in one of the inner dictionaries defined within products.

• Some of them return data. Others don’t.

Great. Mark is happy with his effort.

He structured the data sensibly so it’s well organised. Separately, he wrote functions that use data from those data structures–the inner dictionaries in products and sometimes make changes to the data in those data structures.

Your call…

The Python Coding Place offers something for everyone:

• a super-personalised one-to-one 6-month mentoring option
$ 4,750

• individual one-to-one sessions
$ 125

• a self-led route with access to 60+ hrs of exceptional video courses and a support forum
$ 400

Which The Python Coding Place student are you?

Third Version • Class and Objects

Mark’s code stores data. The nested dictionaries in products deal with the storage part. His code also does stuff* with the data through the functions he wrote.

*stuff is not quite a Python technical term, in case you’re wondering. But it’s quite suitable here, I think!

Lots of programs store data and do stuff with data.

Object-oriented programming takes these two separate tasks–storing data and doing stuff with data–and combines them into a single “unit”. In OOP, this “unit” is the object. Objects are at the centre of the OOP paradigm, which is why OOP is called OOP, after all!

1. Mark’s progression from the first version to the second relied on structuring the data into units–the nested dictionary structure.

2. The progression from the second version to the third takes this a step further and structures the data and the functions that act on those data into a single unit: the object.

This structure means that the data for each object is contained within the object, and the actions performed are also contained within that object. Everything is self-contained within the object.

Let’s build a Product class for Mark. But let’s do this step by step, using the code from the second version and gradually morphing it into object-oriented code. This will help us follow the transition.

First, let’s create the class:

Nothing much to see so far. But this is the shell that will contain all the instructions for creating this “unit,” including all the relevant data and functionality.

You can already create an instance of this class – that’s another way of saying an object created from this class:

Whereas Product represents the class, Product() is an instance of the class. Note the parentheses. There’s only one Product, but you can create many instances using Product() .

Bundling In the Functionality • Methods

Now, let me take a route I recently explored for the first time while teaching OOP in a beginners’ course. It’s only subtly different from my “usual” teaching route, but I think it helped me appreciate the topic from a distinct perspective. So, here we go.

Let’s copy and paste the functions from version two directly into this class. Warning: This won’t work. We’ll need to make some changes. But let’s use this as a starting point:

To keep you on your toes, we change the terminology here. These are functions. However, when they’re defined within a class definition, we call them methods. But they’re still functions. Everything you know about functions applies equally to methods.

You’ve seen that each function in version two needed to know which product it was dealing with. That’s why the first parameter is product_name. You used product_name to fetch the correct values when you needed the selling price, the cost price, or the number of items in stock.

However, now that we’re in the OOP domain, the object will contain all the data it needs–you’ll add the data-related code soon. You don’t need to fetch the data from anywhere else, just from the object itself.

Therefore, the functions defined within the class – the methods – no longer need the product name as the first argument. Instead, they need the entire object itself since this object contains all the data the method needs about the object.

We just need a parameter name to describe the object itself. How about self?

The method signatures now include self as the first parameter and whatever else the methods need as the remaining parameters.

Incidentally, although you can technically name this first parameter anything you want, there’s a strong convention to always use self. So it’s best to stick with self!

The methods are the tools you use to bundle functionality into the object. Let’s pause on working on the methods for now and shift our attention to how to store the data, which also needs to be bundled into the object.

Bundling In the Data • Data Attributes

Let’s get back to the code that creates an instance of the Product class:

The expression Product() does a few things behind the scenes. First, it creates a new blank object. Then it initialises this object. To initialise the object, Python needs to “do stuff”. Therefore, it needs a method. But not just any method. A special method:

You can’t choose the name of this method. It must be .__init__(). But it’s still a method. It still takes self as the first parameter since it still needs access to the object itself. This method creates variables that are attached to the object. You can think of the dot as attaching name to self and so on for the others. We tend not to call these variables – more new terminology just for the OOP paradigm – they’re data attributes. They’re object attributes that store data.

For now, these data attributes contain default values: the empty string for the .name data attribute and 0 for the others. However, when you create an object, you often want to supply some or all of the data that the object needs. Often, you want to pass this information when you create the object:

However, when you run the code, you get an error:

Traceback (most recent call last):
  ...
    a_product = Product(”Coffee”, 1.98, 3.2, 100)
TypeError: Product.__init__() takes 1 positional 
    argument but 5 were given

The error message mentions Product.__init__(). Note how you don’t explicitly use .__init__() in the expression to create a Product object. However, Python calls this special method behind the scenes. And when it does, it complains about the arguments you passed:

• "Product.__init__() takes 1 positional argument..." is the first part of the error message. Makes sense, since you have self as the one and only parameter within the definition of the .__init__() special method.

• "...but 5 [arguments] were given", the error message goes on to say. Wait, why 5? You pass four objects when you create Product(): the string "Coffee", the floats 1.98 and 3.2, and the integer 100. Python can’t count, it seems?

Not quite. When Python calls a method, it automatically passes the object as the first argument. You don’t see this. You don’t need to do anything about it. It happens behind the scenes. So, soon after Product() creates a blank new instance, it calls .__init__() and passes the object as the first argument to .__init__(), the one that’s assigned to the parameter self.

Recall that methods are functions. So, Python automatically passes the object as the first argument so that these functions (methods) have access to the object.

That’s why the error message says that five arguments were passed: the object itself and the four remaining arguments.

This tells you that .__init__() needs five parameters. The first is self, which is the first parameter in these methods. Let’s add the remaining four:

The code no longer raises an error since the number of arguments passed to Product() when you create the object – including the object itself, which is implied – matches the number of parameters in .__init__(). However, you want to shift the data into the data attributes you created earlier. These data attributes are the storage devices attached to the object. You want the data to be stored there:

The parameters are used only to transfer the data from the method call to the data attributes. From now on, the data are stored in the data attributes, which are attached to the object.

Note that the data attribute names don’t have to match the parameter names. But why bother coming up with different names? Might as well use the same ones!

Now, you can create any object you wish, each having its own data:

Here’s the output:

Coffee
Tea
3.2
2.25

Every instance of the Product class will have a .name, .cost_price, .selling_price, and .stock. But each instance will have its own values for those data attributes. Each object is distinct from the others. The data is self-contained within the object.

Back to the Functionality • Methods

Let’s look at the class so far, which still has code pasted from version two earlier:

Let’s focus on .update_stock() first. Note how in my writing style guide, I use a leading dot when writing method names, such as .update_stock(). That’s because they’re also attributes of the object. To call a method, you call it through the object, such as a_product.update_stock(30).

Recall that Python will pass the object itself as the first argument to the method. That’s the argument assigned to self. You then pass 30 to the parameter stock_increment.

But this means that this method can only act on an object that already exists, that already has the data attributes it needs. You no longer need to check whether the product exists. If you’re calling this method, you’re calling it on a product. Checks on existence will happen elsewhere in your code.

So, all that’s left is to increment the stock for this object. The current stock is stored in the .stock data attribute – the data attribute that’s connected to the object. But you passed the object to .update_stock(), so you can access self.stock from within the method:

And that’s it. Everything is self-contained within the object. The method acts on the object and modifies one of the object’s data attributes.

How about the .sell() method?

Again, you can remove the validation to check that the product exists. A method is called on a Product object, so it exists. This method then uses three of the object’s data attributes to perform its task.

Note that you can add similar validation in .update_stock() to ensure the stock doesn’t dip below 0. However, stock_increment can be negative to enable reducing the stock. You could even use .update_stock() within .sell() if you wish.

Ah, what about the biscuits? You can still use a list of products, but this time the list contains Product objects:

Or, if you prefer, you can use a dictionary:

Let’s trial out this dictionary and the Product class:

In this basic use case, you use the item input by the user to fetch the corresponding Product object from the products dictionary. You assign this object to the variable name product. Then you can call its methods, such as .sell(), and access its data attributes, such as .name and .stock.

Here’s the output from this code, including sample user inputs:

Enter item: coffee
Enter quantity of Coffee sold: 3
The income from this sale is £9.60
The profit from this sale is £3.66
Remaining Coffee units: 97

Final Words

The aim of this article is not to provide a comprehensive and exhaustive walkthrough of OOP. I wrote elsewhere about OOP. You can start from The Python Coding Book and then read the seven-part series A Magical Tour Through Object-Oriented Programming in Python • Hogwarts School of Codecraft and Algorithmancy.

You can learn about the syntax and the mechanics of defining and using classes. And that’s important if you want to write classes. But just as importantly, you need to adopt an OOP mindset. A central point is how OOP bundles data and functionality into a single unit, the object, and how everything stems from that structure.

In this post, you saw how data is bundled into the object through data attributes. And you bundled the high-level functionality that matters for your object through the methods .update_stock() and .sell(). Mark will need more of these methods, as you can imagine.

However, the object also includes low-level functionality. What should Python do when it needs to print the object? How about if it needs to add it to another object? Should that be possible? Should the object be considered False under any circumstance, say? These operations are defined by an object’s special methods, also known as dunder methods. I proposed thinking of these as “plumbing methods” recently: “Python’s Plumbing” Is Not As Flashy as “Magic Methods” • But Is It Better?

Therefore, there’s plenty of high-level and low-level functionality bundled within the object. And the data, of course.

Here’s another relatively recent post you may enjoy in case you missed it when it was published: My Life • The Autobiography of a Python Object.

In summary, OOP structures data and functionality into a single unit – the object. Then, your code is oriented around this object.

Photo by gomed fashion

Code in this article uses Python 3.14

The code images used in this article are created using Snappify. [Affiliate link]

Join The Club, the exclusive area for paid subscribers for more Python posts, videos, a members’ forum, and more.

Subscribe now

You can also support this publication by making a one-off contribution of any amount you wish.

Support The Python Coding Stack

For more Python resources, you can also visit Real Python—you may even stumble on one of my own articles or courses there!

Also, are you interested in technical writing? You’d like to make your own writing more narrative, more engaging, more memorable? Have a look at Breaking the Rules.

And you can find out more about me at stephengruppetta.com

Further reading related to this article’s topic:

• The Python Coding Book

• A Magical Tour Through Object-Oriented Programming in Python • Hogwarts School of Codecraft and Algorithmancy

• My Life • The Autobiography of a Python Object

Appendix: Code Blocks

Code Block #1

products = ["Coffee", "Tea", "Biscuits"]
cost_price = [1.98, 0.85, 3.1]
selling_price = [3.2, 2.25, 5]
stock = [100, 200, 25]

Code Block #2

products = {
    "Coffee": {
        "cost_price": 1.98,
        "selling_price": 3.2,
        "stock": 100,
    },
    "Tea": {
        "cost_price": 0.85,
        "selling_price": 2.25,
        "stock": 200,
    },
    "Biscuits": {
        "cost_price": 3.1,
        "selling_price": 5,
        "stock": 25,
    },
}

Code Block #3

# ...
def update_stock(product_name, stock_increment):
    if product_name not in products:
        raise KeyError(f"Product {product_name} not found")
    products[product_name]["stock"] += stock_increment
    # We'll need to have a chat with Mark about the issues
    # with accessing global variables from functions.
    # Let's let it pass here!

Code Block #4

# ...
def sell(product_name, quantity):
    if product_name not in products:
        raise KeyError(f"Product {product_name} not found")
    if quantity <= 0:
        raise ValueError("Quantity must be greater than zero")
    if products[product_name]["stock"] < quantity:
        raise ValueError("Not enough stock")
        
    products[product_name]["stock"] -= quantity
    selling_price = products[product_name]["selling_price"]
    cost_price = products[product_name]["cost_price"]

    income = selling_price * quantity
    profit = (selling_price - cost_price) * quantity

    return income, profit

Code Block #5

class Product:
    ...

Code Block #6

class Product:
    ...
    

a_product = Product()

Code Block #7

class Product:
    # Just pasting the functions from Version 2 here for now.
    # This still doesn't work. It's a work-in-progress
    def update_stock(product_name, stock_increment):
        if product_name not in products:
            raise KeyError(f"Product {product_name} not found")
        products[product_name]["stock"] += stock_increment
        # We'll need to have a chat with Mark about the issues
        # with accessing global variables from functions.
        # Let's let it pass here!

    def sell(product_name, quantity):
        if product_name not in products:
            raise KeyError(f"Product {product_name} not found")
        if quantity <= 0:
            raise ValueError("Quantity must be greater than zero")
        if products[product_name]["stock"] < quantity:
            raise ValueError("Not enough stock")

        products[product_name]["stock"] -= quantity
        selling_price = products[product_name]["selling_price"]
        cost_price = products[product_name]["cost_price"]

        income = selling_price * quantity
        profit = (selling_price - cost_price) * quantity

        return income, profit

Code Block #8

class Product:
    def update_stock(self, stock_increment):
        # rest of code pasted from version two (for now)

    def sell(self, quantity):
        # rest of code pasted from version two (for now)

Code Block #9

# ...

a_product = Product()

Code Block #10

class Product:
    # We'll make improvements to this method soon
    def __init__(self):
        self.name = ""
        self.cost_price = 0
        self.selling_price = 0
        self.stock = 0

Code Block #11

# ...

a_product = Product("Coffee", 1.98, 3.2, 100)

Code Block #12

class Product:
    def __init__(self, name, cost_price, selling_price, stock):
        self.name = ""
        self.cost_price = 0
        self.selling_price = 0
        self.stock = 0

Code Block #13

class Product:
    def __init__(self, name, cost_price, selling_price, stock):
        self.name = name
        self.cost_price = cost_price
        self.selling_price = selling_price
        self.stock = stock

Code Block #14

# ...

a_product = Product("Coffee", 1.98, 3.2, 100)
another_product = Product("Tea", 0.85, 2.25, 200)

print(a_product.name)
print(another_product.name)

print(a_product.selling_price)
print(another_product.selling_price)

Code Block #15

class Product:
    def __init__(self, name, cost_price, selling_price, stock):
        self.name = name
        self.cost_price = cost_price
        self.selling_price = selling_price
        self.stock = stock

    # Just pasting the functions from Version 2 here for now.
    # This still doesn't work. It's a work-in-progress
    def update_stock(self, stock_increment):
        if product_name not in products:
            raise KeyError(f"Product {product_name} not found")
        products[product_name]["stock"] += stock_increment
        # We'll need to have a chat with Mark about the issues
        # with accessing global variables from functions.
        # Let's let it pass here!

    def sell(self, quantity):
        if product_name not in products:
            raise KeyError(f"Product {product_name} not found")
        if quantity <= 0:
            raise ValueError("Quantity must be greater than zero")
        if products[product_name]["stock"] < quantity:
            raise ValueError("Not enough stock")

        products[product_name]["stock"] -= quantity
        selling_price = products[product_name]["selling_price"]
        cost_price = products[product_name]["cost_price"]

        income = selling_price * quantity
        profit = (selling_price - cost_price) * quantity

        return income, profit

Code Block #16

# ...
    def update_stock(self, stock_increment):
        self.stock += stock_increment

Code Block #17

# ...
    def sell(self, quantity):
        if quantity <= 0:
            raise ValueError("Quantity must be greater than zero")
        if self.stock < quantity:
            raise ValueError("Not enough stock")

        self.stock -= quantity
        income = self.selling_price * quantity
        profit = (self.selling_price - self.cost_price) * quantity

        return income, profit

Code Block #18

# ...
products = [
    Product("Coffee", 1.98, 3.2, 100),
    Product("Tea", 0.85, 2.25, 200),
    Product("Biscuits", 3.1, 5, 25),
]

Code Block #19

products = {
    "Coffee": Product("Coffee", 1.98, 3.2, 100),
    "Tea": Product("Tea", 0.85, 2.25, 200),
    "Biscuits": Product("Biscuits", 3.1, 5, 25),
}

Code Block #20

# ...

# I've kept this demo code brief and simple, so it lacks
# verification and robustness. You can add this if you wish!

item = input("Enter item: ").title()
qty = int(input(f"Enter quantity of {item} sold: "))

product = products[item]
income, profit = product.sell(qty)

print(f"The income from this sale is £{income:.2f}")
print(f"The profit from this sale is £{profit:.2f}")

print(f"Remaining {product.name} units: {product.stock}")

For more Python resources, you can also visit Real Python—you may even stumble on one of my own articles or courses there!

Also, are you interested in technical writing? You’d like to make your own writing more narrative, more engaging, more memorable? Have a look at Breaking the Rules.

And you can find out more about me at stephengruppetta.com

Everything goes through special methods (dunder methods) in Python

Special methods are everywhere. However, you don’t see them. In fact, you’re not meant to use them directly unless you’re defining them within a class. They’re out of sight. But they keep everything moving smoothly in Python.

Here’s a short essay exploring where these special methods fit within Python.

Why Should I Care About Special Methods?

“Everything goes through special methods in Python.” I should probably preface the phrase with “almost”, but the phrase is already unwieldy as it is.

You want to display an object on the screen? Python looks for guidance in .__str__() or .__repr__().

You want to use the object in a for loop? Is it even possible? Python looks for .__iter__() to check whether it can iterate over the object and how.

Do you want to fetch an individual item from the object? If this makes sense for this object, then it will have a .__getitem__() special method.

How should Python interpret an object’s truthiness? There’s .__bool__() for that. Or .__len__()!

Ah, you’d like to add an object to another object. Does your object have a .__add__() special method?

I could go on. I’ll post links to articles that cover some of these special methods in more detail below.

I’m also running a two-hour workshop this Thursday, 26 February, called Python’s Plumbing • Dunder Methods and Python’s Hidden Interface

Book Workshops

Many operations you take for granted in your code are governed by special methods. Each class defines the special methods it needs, and Python knows how to handle instances of that class through those methods.

But let’s talk about different ways we refer to these special methods.

What’s In A Name?

These methods are called special methods. That’s their official name. These special methods have double underscores at the beginning and end of their names, such as .__init__(), .__str__(), and .__iter__(). This double-underscore notation led to the informal name “dunder methods”. And let’s face it, most people call them “dunder methods” instead of their actual name: “special methods.”

I often call them dunder methods, too. However, the term dunder merely describes the syntax, double underscore, so it doesn’t tell us much about what they do. The term special doesn’t tell us what they do, either, but it shows us they have a special role in Python.

There’s No Such Thing as Magic

Some people also call them magic methods. However, I avoid this term, and I discourage students from using it. It makes these methods look unnecessarily mysterious, perhaps difficult to understand because it’s all down to magic.

But there’s no such thing as magic (unless you’re Harry Potter). And the magic tricks we see from real-world “magicians” are just that – tricks. The magic dissolves away once you know how the trick works.

And if you’re learning Python, then you need to learn how to be a magician. You need to learn the “magic tricks.” Therefore, they’re no longer magic!

Python’s Plumbing (”Plumbing Methods”, Anyone?)

So, “special method” tells us that these methods are important, but it doesn’t tell us what they do. “Dunder method” describes the syntax. “Magic method” misleads us and doesn’t provide any useful insight.

How about “plumbing methods” then? Now, before anyone takes me too seriously, I’m saying this with my tongue firmly in my cheek. I’m not that foolish to suggest a new term for the whole Python community to adopt. And it’s not as flashy as “magic methods” or as cool as “dunder methods”. But bear with me…

Let’s explore the analogy, even if the term won’t catch on.

Disclaimer: I know very little about plumbing. But I think that’s OK for this essay!

There are pipes, valves, and other stuff carrying water (clean or otherwise) around your house. You know they’re there. You need them there. But you don’t see them.

You don’t think about these pipes unless you’re building the house – or unless the pipes are blocked or leaking.

The house’s plumbing keeps things running smoothly. Yet, it’s out of sight, and you don’t normally think about it. You take it for granted.

You see where I’m going, right?

Python’s special methods perform the same role. You don’t normally see them when coding since they’re called implicitly, behind the scenes. You do need to define the special methods you need when you’re defining the class, just like you’ll need to lay the pipes when building or modifying your house.

And if something goes wrong in your code, you may need to dive into how these dunder methods behave, just as when you have a leak and need to explore which pipe is responsible.

Good plumbing is reliable, predictable. Bad plumbing is asking for trouble. The same applies to the infrastructure you create through the special methods you define in classes.

So, there you go, they’re “plumbing methods”. This name tells us what they do!

I’m running a two-hour live workshop this Thursday called Python’s Plumbing • Dunder Methods and Python’s Hidden Interface. This is your last chance to join and I may not run this workshop again for a while.

This workshop is the first of three in the Python Behind the Scenes series. The other two workshops in the series are:

• #2 • Pythonic Iteration: Iterables, Iterators, itertools

• #3 • To Inherit or Not? Inheritance, Composition, Abstract Base Classes, and Protocols

Join all three, or pick and choose:

The Python Coding Stack

When "It Works" Is Not Good Enough • Live Workshops

You’ve been reading articles here on The Python Coding Stack. How about live workshops…

Read more

4 months ago · 1 like · 2 comments · Stephen Gruppetta

Book Workshops

Image by Pete Linforth from Pixabay

For more Python resources, you can also visit Real Python—you may even stumble on one of my own articles or courses there!

Also, are you interested in technical writing? You’d like to make your own writing more narrative, more engaging, more memorable? Have a look at Breaking the Rules.

And you can find out more about me at stephengruppetta.com

Further reading related to this article’s topic:

• This is part of a longer OOP series: Time for Something Special • Special Methods in Python Classes

• The Anatomy of a for Loop

• The Manor House, the Oak-Panelled Library, the Vending Machine, and Python’s __getitem__() [Part 1]

• And Now for Something New

• Why Do 5 + “5” and “5” + 5 Give Different Errors in Python? • Do You Know The Whole Story?

• My Life • The Autobiography of a Python Object

• Telling The Truthy

And in my 10+ years of teaching Python, I also preferred teaching LBYL to beginners and delaying EAFP until later.

But over the years, as I came to understand Python’s psyche better, I gradually shifted my programming style—and then, my teaching style, too.

So, what are LBYL and EAFP? And which one is more suited to Python?

Look Both Sides Before Crossing the Road

You should definitely look before you leap across a busy road…or any road, really. And programming also has a Look Before You Leap concept—that’s LBYL—when handling potential failure points in your code.

Let’s start by considering this basic example. You define a function that accepts a value and a list. The function adds the value to the list if the value is above a user-supplied threshold:

def add_value_above_threshold(value, threshold, data):
    if value >= threshold:
        data.append(value)

You can confirm this short code works as intended:

# ...
prices = []
add_value_above_threshold(12, 5, prices)
add_value_above_threshold(3, 5, prices)
add_value_above_threshold(9, 5, prices)
print(prices)

This code outputs the list with the two prices above the threshold:

[12, 9]

However, you want to ensure this can’t happen:

# ...
products = {}
add_value_above_threshold(12, 5, products)

Now, products is a dictionary, but add_value_above_threshold() was designed to work with lists and not dictionaries:

Traceback (most recent call last):
  ...
    add_value_above_threshold(12, 5, products)
    ~~~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^
  ...
    data.append(value)
    ^^^^^^^^^^^
AttributeError: ‘dict’ object has no attribute ‘append’

One option is the look before you leap (LBYL):

def add_value_above_threshold(value, threshold, data):
    if not isinstance(data, list):
        print(”Invalid format. ‘data’ must be a list”)
        return
    if value >= threshold:
        data.append(value)

Now, the function prints a warning when you pass a dictionary, and it doesn’t crash the program!

But this is too restrictive.

Let’s assume you decide to use a deque instead of a list:

from collections import deque
​
# ...
​
prices = deque()
add_value_above_threshold(12, 5, prices)
add_value_above_threshold(3, 5, prices)
add_value_above_threshold(9, 5, prices)
print(prices)

This code still complains that it wants a list and doesn’t play ball:

Invalid format. ‘data’ must be a list
Invalid format. ‘data’ must be a list
Invalid format. ‘data’ must be a list
deque([])

But there’s no reason why this code shouldn’t work since deque also has an .append() method.

You could change the call to isinstance() to include the deque data type—isinstance(data, list | deque)—but then there may be other data structures that are valid and can be used in this function. You don’t want to have to write them all.

If you’re well-versed with the categories of data structures—perhaps because you devoured the The Python Data Structure Categories Series—then you might conclude you need to check whether the object is a MutableSequence since all mutable sequences have an .append() method. You can import MutableSequence from collections.abc and use isinstance(data, MutableSequence). Now you’re fine to use lists, deques, or any other mutable sequence.

This version fits better with Python’s duck-typing philosophy. It doesn’t restrict the function to a limited number of data types but to a category of data types. This category is defined by what the data types can do. In duck typing, you care about what an object can do rather than what it is. You can read more about duck typing in Python in this post: When a Duck Calls Out • On Duck Typing and Callables in Python

However, you could still have other data types that have an .append() method but may not fully fit into the MutableSequence category. There’s no reason you should exclude those data types from working with your function.

Perhaps, you could use Python’s built-in hasattr() to check whether the object you pass has an .append() attribute. You’re now checking whether the object has the required attribute rather than what the object is.

But if you’re going through all this trouble, you can go a step further.

Just Go For It and See What Happens

Why not just run the line of code that includes data.append() and see what happens? Ah, but you don’t want the code to fail if you use the wrong data type—you only want to print a warning, say.

That’s where the try..except construct comes in:

def add_value_above_threshold(value, threshold, data):
    if value < threshold:  # inequality flipped to avoid nesting
        return
    try:
        data.append(value)
    except AttributeError:
        print(
            “Provided data structure does not support appending values.”
        )

This is the Easier to Ask for Forgiveness than Permission (EAFP) philosophy. Just try the code. If it doesn’t work, you can then deal with it in the except block. Now, this fits even more nicely with Python’s duck typing philosophy. You’re asking the program whether data can append a value. It doesn’t matter what data is–can it append a value?

You don’t have to think about all the valid data types or which category they fall into. And rather than checking whether the data type has the .append() attribute first, you just try to run the code and deal with the consequences later. That’s why it’s easier to ask for forgiveness than permission.

But don’t use this philosophy when crossing a busy road. Stick with “look before you leap” there!

Another Example Comparing LBYL and EAFP

Read more

But do you feel that “it works” isn’t good enough? You need to understand Python’s “behind the scenes” to write robust and efficient code. If you’re keen to step up your Python coding, then I’ve got a series of live, hands-on workshops coming up. I think you’ll find them interesting and useful.

Here they are:

Note: workshops have already taken place

1. Python’s Plumbing: Dunder Methods and Python’s Hidden Interface

2. Pythonic Iteration: Iterables, Iterators, itertools

3. To Inherit or Not? Inheritance, Composition, Abstract Base Classes, and Protocols

Each live workshop will run for 2 hours+ (we’ll keep going if you have more questions!) My live teaching style is quite similar to how I write. So, expect a relaxed, friendly, and fun session. And you have permission to jump in and ask questions at any point.

Here’s a bit more about each workshop. You can sign up to whichever workshop you wish or all three of them. Up to you.

Pythonic Iteration: Iterables, Iterators, itertools

What happens in a for loop? There’s a lot more than you see on the surface. And if you’re using too many for loops in your code, maybe you’re missing out on some more Pythonic iteration options.

In this workshop, you’ll finally get to grips with the difference between iterable and iterator. You’ll master the Iterator Protocol and you’ll see how there’s always an iterator somewhere behind every iteration. And you know lots of weird iteration patterns, often consisting of nested for loops? You probably don’t need them. Python has special iteration tools hidden in the standard module, including the itertools treasure trove.

To Inherit or Not? Inheritance, Composition, Abstract Base Classes, and Protocols

When you learn object-oriented programming, you learn about inheritance. And that’s cool. But when should you use inheritance? Are there alternatives? [Spoiler alert: yes, there are]

The workshop explores inheritance, yes, but also composition. We’ll explore examples to understand when to use one or when to use the other (or when to use both). This discussion will lead us to some key design principles and how abstract base classes can help. But what if you don’t want to use inheritance? ABCs can’t help, but protocols (the typing.Protocol one) can step in to give you a duck typing-friendly way to organise your code.

The workshops are on Zoom but I won’t use the webinar format–that’s too ‘one-way’ for my liking. Instead, I’ll run them in a standard Zoom meeting where you can jump in and ask questions at any time. It’s friendlier that way!

You’ll also get access to the recording of the session.

Any questions? Just reply to this email and ask.

Join the Workshops

Each workshop is $45. Or you can get both remaining workshops for $75. That’s 4+ hours of live, hands-on, interactive learning…

Here are the dates and times for each workshop. Each workshop runs for 2+ hours. I’m showing times in a few time zones:

1. Python’s Plumbing: Dunder Methods and Python’s Hidden Interface

Thursday 26 February 2026 • London 9:00 PM • New York 4:00 PM • Los Angeles 1:00 PM • Berlin 10:00 PM • UTC/GMT 9:00 PM

2. Pythonic Iteration: Iterables, Iterators, itertools

Thursday 12 March 2026 • London 9:00 PM • New York 5:00 PM • Los Angeles 2:00 PM • Berlin 10:00 PM • UTC/GMT 9:00 PM

3. To Inherit or Not? Inheritance, Composition, Abstract Base Classes, and Protocols

Thursday 19 March 2026 • London 9:00 PM • New York 5:00 PM • Los Angeles 2:00 PM • Berlin 10:00 PM • UTC/GMT 9:00 PM

Book your place on both remaining workshops in one convenient, cost-effective bundle:

BOOKING CLOSED

So, in summary:

• Live workshops to master core Python and improve the quality of your code

• 2+ hours each, live, interactive, hands-on

• Ask questions during the sessions—your questions always lead to interesting discussions

See you at the workshops…

And my reaction at first was: “So what?! This tool doesn’t do anything I couldn’t already do with other features in Python.”

So, why does itemgetter() exist?

But, before answering this question, let’s answer a more basic one…

What Does itemgetter() Do?

Read more

So we came up with a system, and I’m writing a Python program to implement it. We keep a list of meals we try out. Each family member assigns a score to each meal. Every Saturday, before we go to the supermarket for the weekly shop, we plan which meals we’ll cook on each day of the week. It’s not based solely on the preference ratings, of course, since my wife and I have the final say to ensure a good balance. Finally, the program provides us with the shopping list with the ingredients we need for all the week’s meals.

I know, we’ve reinvented the wheel. There are countless apps that do this. But the fun is in writing your own code to do exactly what you want.

I want to keep this article focussed on just one thing: alternative constructors using class methods. Therefore, I won’t go through the whole code in this post. Perhaps I’ll write about the full project in a future article.

So, here’s what you need to know to get our discussion started.

Setting the Scene • Outlining the Meal and WeeklyMealPlanner Classes

Let me outline two of the classes in my code. The first is the Meal class. This class – you guessed it – deals with each meal. Here’s the class’s .__init__() method:

The meal has a name so we can easily refer to it, so there’s a .name data attribute. And the meals I cook are different from the meals my wife cooks, which is why there’s a .person_cooking data attribute. This matters as on some days of the week, only one of us is available to prepare dinner, so this attribute becomes relevant!

There are also days when we have busy afternoons and evenings with children’s activities, so we need to cook a quick meal. The .quick_meal data attribute is a Boolean flag to help with planning for these hectic days.

Then there’s the .ingredients data attribute. You don’t need me to explain this one. And since each family member rates each meal, there’s a .ratings dictionary to keep track of the scores.

The class has more methods, such as add_ingredient(), remove_ingredient(), add_rating(), and more. There’s also code to save to and load from CSV and JSON files. But these are not necessary for today’s article, so I’ll leave them out.

There’s also a WeeklyMealPlanner class:

The ._meals data attribute is a dictionary with the days of the week as keys and Meal instances as values. It’s defined as a non-public attribute to be used with the read-only property .meals. The .meals property returns a shallow copy of the ._meals dictionary. This makes it safer as it’s harder for a user to make changes directly to this dictionary. The dictionary is modified only through methods within WeeklyMealPlanner. I’ve omitted the rest of the methods in this class as they’re not needed for this article.

You can read more about properties in Python in this article: The Properties of Python’s ‘property’

So, each time we try a new dish, we create a Meal object, and each family member rates it. This meal then goes into our collection of meals to choose from each week. On Saturday, we choose the meals we want for the week, put them in a WeeklyMealPlanner instance, and we’re almost ready to go…

At the Supermarket

Well, we’re almost ready to go to the supermarket at this point. So, here’s another class:

A ShoppingList object has an .ingredients data attribute. This attribute is a dictionary. The keys are the ingredients, and the values are the quantities needed for each ingredient. I’m also showing the .add_ingredient() method, which I’ll need later on. So, you can create an instance of ShoppingList in the usual way:

Then, you can add ingredients as needed. But this is annoying for us on a Saturday. Here’s why…

Do you want to master Python one article at a time? Then don’t miss out on the article in The Club which are exclusive to premium subscribers here on The Python Coding Stack

Subscribe now

Alternative Constructor

Before describing our Saturday problems, let’s briefly revisit what happens when you create an instance of a class. When you place parentheses after the class name, Python does two things: it creates a blank new object, and it initialises it. The creation of the new object almost always happens “behind the scenes”. The .__new__() method creates a new object, but you rarely need to override it. And the .__init__() method performs the object’s initialisation.

You can only have one .__init__() special method in a class. Does this mean there’s only one way to create an instance of a class?

Not quite, no. Although there’s no way to define more .__init__() methods, there are ways to create instances through different routes. The @singledispatchmethod decorator is a useful tool, but one I’ll discuss in a future post. Today, I want to talk about using class methods as alternative constructors.

Back to a typical Saturday in our household. We just finished choosing the seven dinners we plan to have this coming week, and we created a WeeklyMealPlanner instance. So we should now create a ShoppingList instance using ShoppingList() and then go through all the meals we chose, entering their ingredients.

Wouldn’t it be nice if we could just create a ShoppingList instance directly from the WeeklyMealPlanner instance? But that would require a different way to create an instance of ShoppingList.

Let’s define an alternative constructor, then:

There’s a new method called .from_meal_planner(). However, this is not an instance method. It doesn’t belong to an instance of the class. Instead, it belongs to the class directly. The @classmethod decorator tells Python to treat this method as a class method. Note that the first parameter in this method is not self, as with the usual (instance) methods. Instead, you use cls, which is the parameter name used by convention to refer to the class.

Whereas self in an instance method represents the instance of a class, cls represents the class directly. So, unlike instance methods, class methods don’t have access to the instance. Therefore, class methods don’t have access to instance attributes.

The first line of this method creates an instance of the class. Look at the expression cls(), which comes after the = operator. Recall that cls refers to the class. So, cls is the same as ShoppingList in this example. But adding parentheses after the class creates an instance. You assign this new instance to the local variable shopping_list. You use cls rather than ShoppingList to make the class more robust in case you choose to subclass it later.

Fast-forward to the end of this class method, and you’ll see that the method returns this new instance, shopping_list. However, it makes changes to the instance before returning it. The method fetches all the ingredients from each meal in the WeeklyMealPlanner instance and populates the .ingredients data attribute in the new ShoppingList instance.

In summary, the class method doesn’t have access to an instance through the self parameter. But since it has access to the class, the method uses the class to create a new instance and initialise it, adding steps to the standard .__init__() method.

Therefore, this class method creates and returns an instance of ShoppingList with its .ingredients data attribute populated with the ingredients you need for all the meals in the week.

You now have an alternative way of creating an instance of ShoppingList:

This class now has two ways to create instances. The standard one using ShoppingList() and the alternative one using ShoppingList.from_meal_planner(). It’s common for class methods used as alternative constructors to have names starting with from_*.

You can have as many alternative constructors as you need in a class.

Question: if it’s more useful to create a shopping list directly from the weekly meal planner, couldn’t you implement this logic directly in the .__init__() method? Yes, you could. But this would create a tight coupling between the two classes, ShoppingList and WeeklyMealPlanner. You can no longer use ShoppingList without an instance of WeeklyMealPlanner, and you can no longer easily create a blank ShoppingList instance.

Creating two constructors gives you the best of both worlds. ShoppingList is still flexible enough so you can use it as a standalone class or in conjunction with other classes in other projects. But you also have access to the alternative constructor ShoppingList.from_meal_planner() when you need it.

Alternative Constructors in the Wild

You may have already seen and used alternative constructors, perhaps without noticing.

Let’s consider dictionaries. The standard constructor is dict() – the name of the class followed by parentheses. As it happens, you have several options when using dict() – you can pass a mapping, or an iterable of pairs, or **kwargs. You can read more about these alternatives in this article: dict() is More Versatile Than You May Think.

But there’s another alternative constructor that doesn’t use the standard constructor dict() but still creates a dictionary. This is dict.fromkeys():

You can have a look at help(dict.fromkeys). You’ll see the documentation text refer to this method as a class method, just like the ShoppingList.from_meal_planner() class method you defined earlier.

And if you use the datetime module, you most certainly have used alternative constructors using class methods. The standard constructor when creating a datetime.datetime instance is the following:

However, there are several class methods you can use as alternative constructors:

Have a look at other datetime.datetime methods starting with from_.

Final Words

Python restricts you to defining only one .__init__() method. But there are still ways for you to create instances of a class through different routes. Class methods are a common way of creating alternative constructors for a class. You call them directly through the class and not through an instance of the class – ShoppingList.from_meal_planner(). The class method then creates an instance, modifies it as needed, and finally returns the customised instance.

Now, let me see what’s on tonight’s meal planner and, more importantly, whether it’s my turn to cook.

Photo by Katya Wolf

Code in this article uses Python 3.14

The code images used in this article are created using Snappify. [Affiliate link]

Join The Club, the exclusive area for paid subscribers for more Python posts, videos, a members’ forum, and more.

Subscribe now

You can also support this publication by making a one-off contribution of any amount you wish.

Support The Python Coding Stack

For more Python resources, you can also visit Real Python—you may even stumble on one of my own articles or courses there!

Also, are you interested in technical writing? You’d like to make your own writing more narrative, more engaging, more memorable? Have a look at Breaking the Rules.

And you can find out more about me at stephengruppetta.com

Further reading related to this article’s topic:

• The Properties of Python’s ‘property’

• dict() is More Versatile Than You May Think

Appendix: Code Blocks

Code Block #1

class Meal:
    def __init__(
            self,
            name,
            person_cooking,
            quick_meal=False,
    ):
        self.name = name
        self.person_cooking = person_cooking
        self.quick_meal = quick_meal
        self.ingredients = {}  # ingredient: quantity
        self.ratings = {}  # person: rating

    # ... more methods

Code Block #2

class WeeklyMealPlanner:
    def __init__(self):
        self._meals = {}  # day: Meal

    @property
    def meals(self):
        return dict(self._meals)

    # ... more methods

Code Block #3

class ShoppingList:
    def __init__(self):
        self.ingredients = {}  # ingredient: quantity

    def add_ingredient(self, ingredient, quantity=1):
        if ingredient in self.ingredients:
            self.ingredients[ingredient] += quantity
        else:
            self.ingredients[ingredient] = quantity
            
    # ... more methods

Code Block #4

ShoppingList()

Code Block #5

class ShoppingList:
    def __init__(self):
        self.ingredients = {}  # ingredient: quantity

    @classmethod
    def from_meal_planner(cls, meal_planner: WeeklyMealPlanner):
        shopping_list = cls()
        for meal in meal_planner.meals.values():
            if meal is None:
                continue
            for ingredient, quantity in meal.ingredients.items():
                shopping_list.add_ingredient(ingredient, quantity)
        return shopping_list

    def add_ingredient(self, ingredient, quantity=1):
        if ingredient in self.ingredients:
            self.ingredients[ingredient] += quantity
        else:
            self.ingredients[ingredient] = quantity

Code Block #6

# if my_weekly_planner is an instance of 'WeeklyMealPlanner', then...

shopping_list = ShoppingList.from_meal_planner(my_weekly_planner)

Code Block #7

dict.fromkeys(["James", "Bob", "Mary", "Jane"])
# {'James': None, 'Bob': None, 'Mary': None, 'Jane': None}

Code Block #8

import datetime
datetime.datetime(2026, 1, 30)
# datetime.datetime(2026, 1, 30, 0, 0)

Code Block #9

datetime.datetime.today()
# datetime.datetime(2026, 1, 30, 12, 54, 2, 243976)

datetime.datetime.fromisoformat("2026-01-30")
# datetime.datetime(2026, 1, 30, 0, 0)

For more Python resources, you can also visit Real Python—you may even stumble on one of my own articles or courses there!

Also, are you interested in technical writing? You’d like to make your own writing more narrative, more engaging, more memorable? Have a look at Breaking the Rules.

And you can find out more about me at stephengruppetta.com

And many people who comment about programming and AI know as much about programming as I know about music…and probably even less about AI.

But the orchestra conductor analogy seems a good one. Let me explore it further.

Read more

“Ah, but there are immutable types.”

Sure, you can have an object that doesn’t change throughout its lifetime. But you can’t have a reference to it that’s guaranteed not to change. The identifier (variable name) you use to refer to this immutable type can easily switch to refer to something else.

“How about using all caps for the identifier. Doesn’t that make it a constant?”

No, it doesn’t. That’s just a convention you use to show your intent as a programmer that an identifier refers to a value that shouldn’t change. But nothing prevents that value from changing.

Here’s an all-uppercase identifier that refers to an immutable object:

The identifier is all caps. The object is a tuple, which is immutable. Recall that you don’t need parentheses to create a tuple—the comma is sufficient.

So, you use an all-uppercase identifier for an immutable object. But that doesn’t stop you from changing the value of FIXED_LOCATION:

Neither using an immutable object nor using uppercase identifiers prevents you from changing this value!

So, Python doesn’t have constants. But there are tools you can use to mimic constant behaviour depending on the use case you need. In this article I’ll explore one of these: Enums.

All The Python Coding Place video courses are included in a single, cost-effective bundle. The courses cover beginner and intermediate level courses, and you also get access to a members-only forum.

Get The All Courses Bundle

Jargon Corner: Enum is short for enumeration, and you’ll see why soon. But don’t confuse this with the built-in enumerate(), which does something else. See Parkruns, Python’s enumerate and zip, and Why Python Loops Are Different from Other Languages • [Note: This is a Club post] for more on enumerate().

Let’s revisit our friend Alex from an article from a short while ago: “AI Coffee” Grand Opening This Monday. This article explored the program Alex used in his new coffee shop and how the function signature changed over time to minimise confusion and errors when using it. It’s a fun article about all the various types and styles of parameters and arguments you can have in Python functions.

But it didn’t address another potential source of error when using this code. So let’s look at a simple version of the brew_coffee() function Alex used to serve his coffee-drinking customers:

When you call the function, you pass the coffee you want to this function:

And elsewhere in the code, these coffees are defined in a dictionary:

If you’ve written code like this in the past, you’ll know that it’s rather annoying—and error-prone—to keep using the strings with the coffee names wherever you need to refer to a specific coffee, such as when passing the coffees to brew_coffee().

The names of the coffees and the parameters that define them do not change. They’re constant. It’s a shame Python doesn’t have constants, you may think.

But it has enums…

The CoffeeType enum contains seven members. Each member has a name and a value. By convention, you use all-uppercase names for the members since they represent constants. And these enum members behave like constants:

When you attempt to reassign a value to a member, Python raises an exception:

Traceback (most recent call last):
  File ..., line 12, in <module>
    CoffeeType.ESPRESSO = 10
    ^^^^^^^^^^^^^^^^^^^
  ...
AttributeError: cannot reassign member ‘ESPRESSO’

The member names are also contained within the namespace of the Enum class—you use CoffeeType.ESPRESSO rather than just ESPRESSO outside the Enum class definition. So, you get autocomplete, refactor-friendly names, and fewer silent typos. With raw strings, "capuccino" (with a single “p”) can sneak into your code, and nothing complains until a customer is already waiting at the counter.

For these enum members to act as constants, their names must be unique. You can’t have the same name appear more than once:

You include ESPRESSO twice with different values. But this raises an exception:

Traceback (most recent call last):
  File ..., line 3, in <module>
    ...
    ESPRESSO = 8
    ^^^^^^^^
  ...
TypeError: ‘ESPRESSO’ already defined as 1

That’s good news. Otherwise, these enum members wouldn’t be very useful as constants.

However, you can have an alias. You can have more than one member sharing the same value:

The members MACCHIATO and ESPRESSO_MACCHIATO both have the value 4. Therefore, they represent the same item. They’re different names for the same coffee:

Note that Python always displays the first member associated with a value:

CoffeeType.MACCHIATO

The output says CoffeeType.MACCHIATO even though you pass CoffeeType.ESPRESSO_MACCHIATO to print().

Incidentally, if you don’t want to have aliases, you can use the @unique decorator when defining the enum class.

You can also access the name and value of an enum member:

Here’s the output from this code:

CoffeeType.ESPRESSO
ESPRESSO
1

The .name attribute is a string, and the .value attribute is an integer in this case:

Here’s the output when you display the types:

<enum ‘CoffeeType’>
<class ‘str’>
<class ‘int’>

You’ll often use integers as values for enum members—that’s why they’re called enumerations. But you don’t have to:

The values are now also strings:

CoffeeType.ESPRESSO
ESPRESSO
espresso

You can use these enum members instead of strings wherever you need to refer to each coffee type:

…and again when you call brew_coffee():

Now you have a safer, neater, and more robust way to handle the coffee types... and treat them as constants.

A Bit More • StrEnum and IntEnum

Let’s add some code to brew_coffee():

This version is almost fine. But here’s a small problem:

Brewing a CoffeeType.CORTADO with 30ml of coffee and 
    60ml of milk. Strength level: 2

The output displays CoffeeType.CORTADO since coffee_type refers to an enum member. You’d like the output to just show the name of the coffee! Of course, you can use the .value attribute any time you need to fetch the string.

However, to make your coding simpler and more readable, you can ensure that the enum members are also strings themselves without having to rely on one of their attributes. You can use StrEnum instead of Enum:

Members of a StrEnum also inherit all the string methods, such as:

You call the string method .title() directly on the StrEnum member:

Macchiato

There’s also an IntEnum that can be useful when you want your enum members to act as integers. Let’s replace the coffee strength values, which are currently integers, with IntEnum members:

You could use a standard Enum in this case. But using an IntEnum allows you to manipulate its members directly as integers should you need to do so. Here’s an example:

This code is equivalent to printing 3 + 1. You wouldn’t be able to do this with enums unless you use the .value attributes.

And A Couple More Things About Enums

Let’s explore a couple of other useful enum features before we wrap up this article.

An enum class is iterable. Here are all the coffee types in a for loop:

Note that CoffeeType is the class name. But it’s an enum (a StrEnum in this case), so it’s iterable:

Brewing a espresso with 30ml of coffee and 0ml of milk. Strength level: 3
Brewing a latte with 30ml of coffee and 150ml of milk. Strength level: 1
Brewing a cappuccino with 30ml of coffee and 100ml of milk. Strength level: 2
Brewing a macchiato with 30ml of coffee and 10ml of milk. Strength level: 3
Brewing a flat_white with 30ml of coffee and 120ml of milk. Strength level: 2
Brewing a ristretto with 20ml of coffee and 0ml of milk. Strength level: 4
Brewing a cortado with 30ml of coffee and 60ml of milk. Strength level: 2

I’ll let you sort out the text displayed to make sure you get ‘an espresso’ when brewing an espresso and to remove the underscore in the flat white!

And there will be times when you don’t care about the value of an enum member. You just want to use an enum to give your constants a consistent name. In this case, you can use the automatic value assignment:

Python assigns integers incrementally in the order you define the members for Enum classes. Note that these start from 1, not 0.

The same integers are used if you use IntEnum classes. However, when you use StrEnum classes, Python behaves differently since the values should be strings in this case:

The values are now the lowercase strings representing the members’ names.

Of course, the default values you get when you use auto() may be the values you need, after all. This is the case for both enums you created in this article, CoffeeType and CoffeeStrength :

Using auto() when appropriate makes it easier to write your code and expand it later if you need to add more enum members.

Final Words

You can get by without ever using enums. But there are many situations where you’d love to reach for a constant, and an enum will do just fine. Sure, Python doesn’t have constants. But it has enums!

Photo by Valeria Boltneva

Code in this article uses Python 3.14

The code images used in this article are created using Snappify. [Affiliate link]

Join The Club, the exclusive area for paid subscribers for more Python posts, videos, a members’ forum, and more.

Subscribe now

You can also support this publication by making a one-off contribution of any amount you wish.

Support The Python Coding Stack

For more Python resources, you can also visit Real Python—you may even stumble on one of my own articles or courses there!

Also, are you interested in technical writing? You’d like to make your own writing more narrative, more engaging, more memorable? Have a look at Breaking the Rules.

And you can find out more about me at stephengruppetta.com

Appendix: Code Blocks

Code Block #1

FIXED_LOCATION = 51.75, 0.34

Code Block #2

FIXED_LOCATION
# (51.75, 0.34)
FIXED_LOCATION = "Oops!"
FIXED_LOCATION
# 'Oops!'

Code Block #3

def brew_coffee(coffee_type):
    # Actual code goes here...
    # It's not relevant for this article

Code Block #4

brew_coffee("espresso")
brew_coffee("cappuccino")

Code Block #5

coffee_types = {
    "espresso": {"strength": 3, "coffee_amount": 30, "milk_amount": 0},
    "latte": {"strength": 1, "coffee_amount": 30, "milk_amount": 150},
    "cappuccino": {"strength": 2, "coffee_amount": 30, "milk_amount": 100},
    "macchiato": {"strength": 3, "coffee_amount": 30, "milk_amount": 10},
    "flat_white": {"strength": 2, "coffee_amount": 30, "milk_amount": 120},
    "ristretto": {"strength": 4, "coffee_amount": 20, "milk_amount": 0},
    "cortado": {"strength": 2, "coffee_amount": 30, "milk_amount": 60},
}

Code Block #6

from enum import Enum

class CoffeeType(Enum):
    ESPRESSO = 1
    LATTE = 2
    CAPPUCCINO = 3
    MACCHIATO = 4
    FLAT_WHITE = 5
    RISTRETTO = 6
    CORTADO = 7

Code Block #7

from enum import Enum

class CoffeeType(Enum):
    ESPRESSO = 1
    LATTE = 2
    CAPPUCCINO = 3
    MACCHIATO = 4
    FLAT_WHITE = 5
    RISTRETTO = 6
    CORTADO = 7

CoffeeType.ESPRESSO = 10

Code Block #8

from enum import Enum

class CoffeeType(Enum):
    ESPRESSO = 1
    LATTE = 2
    CAPPUCCINO = 3
    MACCHIATO = 4
    FLAT_WHITE = 5
    RISTRETTO = 6
    CORTADO = 7
    ESPRESSO = 8

Code Block #9

from enum import Enum

class CoffeeType(Enum):
    ESPRESSO = 1
    LATTE = 2
    CAPPUCCINO = 3
    MACCHIATO = 4
    FLAT_WHITE = 5
    RISTRETTO = 6
    CORTADO = 7
    ESPRESSO_MACCHIATO = 4

Code Block #10

print(CoffeeType.ESPRESSO_MACCHIATO)

Code Block #11

# ...
print(CoffeeType.ESPRESSO)
print(CoffeeType.ESPRESSO.name)
print(CoffeeType.ESPRESSO.value)

Code Block #12

# ...
print(type(CoffeeType.ESPRESSO))
print(type(CoffeeType.ESPRESSO.name))
print(type(CoffeeType.ESPRESSO.value))

Code Block #13

from enum import Enum

class CoffeeType(Enum):
    ESPRESSO = "espresso"
    LATTE = "latte"
    CAPPUCCINO = "cappuccino"
    MACCHIATO = "macchiato"
    FLAT_WHITE = "flat_white"
    RISTRETTO = "ristretto"
    CORTADO = "cortado"

print(CoffeeType.ESPRESSO)
print(CoffeeType.ESPRESSO.name)
print(CoffeeType.ESPRESSO.value)

Code Block #14

# ...
coffee_types = {
    CoffeeType.ESPRESSO: {"strength": 3, "coffee_amount": 30, "milk_amount": 0},
    CoffeeType.LATTE: {"strength": 1, "coffee_amount": 30, "milk_amount": 150},
    CoffeeType.CAPPUCCINO: {"strength": 2, "coffee_amount": 30, "milk_amount": 100},
    CoffeeType.MACCHIATO: {"strength": 3, "coffee_amount": 30, "milk_amount": 10},
    CoffeeType.FLAT_WHITE: {"strength": 2, "coffee_amount": 30, "milk_amount": 120},
    CoffeeType.RISTRETTO: {"strength": 4, "coffee_amount": 20, "milk_amount": 0},
    CoffeeType.CORTADO: {"strength": 2, "coffee_amount": 30, "milk_amount": 60},
}

Code Block #15

# ...
brew_coffee(CoffeeType.CORTADO)

Code Block #16

# ...

def brew_coffee(coffee_type):
    coffee_details = coffee_types.get(coffee_type)
    if not coffee_details:
        print("Unknown coffee type!")
        return
    print(
        f"Brewing a {coffee_type} "
        f"with {coffee_details['coffee_amount']}ml of coffee "
        f"and {coffee_details['milk_amount']}ml of milk. "
        f"Strength level: {coffee_details['strength']}"
    )

brew_coffee(CoffeeType.CORTADO)

Code Block #17

from enum import StrEnum

class CoffeeType(StrEnum):
    ESPRESSO = "espresso"
    LATTE = "latte"
    CAPPUCCINO = "cappuccino"
    MACCHIATO = "macchiato"
    FLAT_WHITE = "flat_white"
    RISTRETTO = "ristretto"
    CORTADO = "cortado"

# ...

def brew_coffee(coffee_type):
    coffee_details = coffee_types.get(coffee_type)
    if not coffee_details:
        print("Unknown coffee type!")
        return
    print(
        f"Brewing a {coffee_type} "
        f"with {coffee_details['coffee_amount']}ml of coffee "
        f"and {coffee_details['milk_amount']}ml of milk. "
        f"Strength level: {coffee_details['strength']}"
    )

brew_coffee(CoffeeType.CORTADO)

Code Block #18

print(CoffeeType.MACCHIATO.title())

Code Block #19

# ...

class CoffeeStrength(IntEnum):
    WEAK = 1
    MEDIUM = 2
    STRONG = 3
    EXTRA_STRONG = 4

coffee_types = {
    CoffeeType.ESPRESSO: {
      "strength": CoffeeStrength.STRONG, 
      "coffee_amount": 30, 
      "milk_amount": 0,
    },
    CoffeeType.LATTE: {
      "strength": CoffeeStrength.WEAK, 
      "coffee_amount": 30, 
      "milk_amount": 150,
    },
    CoffeeType.CAPPUCCINO: {
      "strength": CoffeeStrength.MEDIUM, 
      "coffee_amount": 30, 
      "milk_amount": 100,
    },
    # ... and so on...
}

# ...

Code Block #20

print(CoffeeStrength.STRONG + CoffeeStrength.WEAK)

Code Block #21

# ...

for coffee in CoffeeType:
    brew_coffee(coffee)

Code Block #22

from enum import Enum, auto
class Test(Enum):
    FIRST = auto()
    SECOND = auto()

Test.FIRST
# <Test.FIRST: 1>
Test.SECOND
# <Test.SECOND: 2>

Code Block #23

from enum import StrEnum, auto
class Test(StrEnum):
    FIRST = auto()
    SECOND = auto()
   
Test.FIRST
# <Test.FIRST: 'first'>
Test.SECOND
# <Test.SECOND: 'second'>

Code Block #24

# ...
class CoffeeType(StrEnum):
    ESPRESSO = auto()
    LATTE = auto()
    CAPPUCCINO = auto()
    MACCHIATO = auto()
    FLAT_WHITE = auto()
    RISTRETTO = auto()
    CORTADO = auto()

class CoffeeStrength(IntEnum):
    WEAK = auto()
    MEDIUM = auto()
    STRONG = auto()
    EXTRA_STRONG = auto()
# ...

For more Python resources, you can also visit Real Python—you may even stumble on one of my own articles or courses there!

Also, are you interested in technical writing? You’d like to make your own writing more narrative, more engaging, more memorable? Have a look at Breaking the Rules.

And you can find out more about me at stephengruppetta.com

And here’s how they deal with the finish line logistics. Runners don’t wear bibs with numbers. When they cross the finish line, they enter a “funnel” marked by plastic cones and are handed a token with their position number. They then proceed to another official, who scans their personal barcode, which runners carry in their pockets or on a wristband, and the position token they received a few seconds earlier. This process matches the runner with their finishing position.

What’s this got to do with Python loops? And how does it help us understand why Python does loops differently from other languages?

First step, let’s create the Parkrun funnel. I’ll just put the first five finishers in this example:

>>> funnel = [”Jonathan”, “Michael”, “Samantha”, “Jessica”, “Daniel”]

Now, here’s something you definitely know already because it’s always one of the first things you’re taught when learning Python: Don’t loop through this list like this:

# Avoid this when coding in Python
>>> i = 0
>>> while i < len(funnel):
...     name = funnel[i]
...     print(name)
...     i += 1
...    
Jonathan
Michael
Samantha
Jessica
Daniel

This style mimics how other languages may work: you manually define and increment the index. To be fair, most people who shift from other languages are more likely to write the following version at some point:

# Also best to avoid this in Python
>>> for i in range(len(funnel)):
...     name = funnel[i]
...     print(name)
...    
Jonathan
Michael
Samantha
Jessica
Daniel

This version may seem more Pythonic since it uses Python tools such as range(), but still fails to make the most of Python’s iteration protocol. The Pythonic way of looping through this list is the following:

>>> for name in funnel:
...     print(name)
...    
Jonathan
Michael
Samantha
Jessica
Daniel

A question that’s often asked but rarely answered is: Why is this version preferred over the other two? I’ll write another short post to answer this question soon as I want to keep these The Club posts short whenever possible. So, let me state just a few reasons (there are more) and then I’ll move on to my main topic for today.

• It’s more readable

• It’s more efficient (try timing the versions above using timeit—remove the print() calls first)

• It’s less prone to errors and bugs

• It works with a broader selection of data structures, not just sequences

While you wait for my follow-up post on this, you can read more about Python’s Iterator Protocol, iterables, and iterators here:

• The Anatomy of a for Loop

• Iterable: Python’s Stepping Stones

• A One-Way Stream of Data • Iterators in Python

But let’s move on.

Let’s say you want to print out the names alongside each runner’s position. You’d like the following output:

1. Jonathan
2. Michael
3. Samantha
4. Jessica
5. Daniel

“Aha!” I’m often told by some learners, “This is when you need to use the for i in range(len(funnel)) idiom, since you need the index!”

Python’s for loop doesn’t explicitly use the index, so you don’t have access to the index within the for loop. Many revert to the non-Pythonic idioms for this.

But Python provides tools that let you stay within the pure Pythonic style. Python’s for loop needs an iterator—it will create one from the iterable you provide. All Python iteration needs iterators, not just for loops. Iterators are Python’s tool for any iteration.

And there are some bespoke iterators in Python that handle most of your iteration needs. I recently wrote a series about the itertools module. The itertools module contains many such tools. Here’s the series: The itertools Series.

But there are also two built-in tools that many forget, but are extremely useful. The first one is enumerate().

Here’s how you can use it to display the Parkrun results:

>>> for index, name in enumerate(funnel, start=1):
...     print(f”{index}. {name}”)
...    
1. Jonathan
2. Michael
3. Samantha
4. Jessica
5. Daniel

Read more

No wonder I can’t remember what articles I wrote this year here at The Python Coding Stack. So make sure you adjust your expectations for this end-of-year review post.

Parties and Gatherings

And there’s another thing I can never remember, especially at this time of year when large-ish gatherings are more common. How many people are needed in a group to have a probability greater than 50% that two people share a birthday? This could be an ice-breaker in some awkward gatherings, but only if you’re with a geeky crowd. Although the analytical proof is cool, writing Python code to explore this problem is just as fun. Here’s my article from February exploring the Birthday Paradox:

This post also explores some tools from the itertools module. Iteration in Python is different from its implementation in many other languages. And the itertools module provides several tools to iterate in a Pythonic way. Later in the year, I explored more of these tools in The itertools Series. Here’s the first post, exploring Yteria’s adventures in a world a bit similar to ours, yet different:

Here’s the whole series following Yteria’s other adventures and the itertools module:

Christmas Decorations

And something else you can’t avoid at this time of year is all the Christmas decorations you’ll find everywhere. Christmas trees, flashing lights, street displays, and…

…Python has its own decorations. You can adorn functions with Python’s equivalent of tinsel and angels:

This post is the most-read post on The Python Coding Stack in 2025. It also has a follow-up post that explores more:

Python’s decorators don’t necessarily make functions pretty—they make them more versatile. However, Python’s f-strings are there to make displayed outputs look pretty. And what if you want your own custom fancy f-string format specifiers?

Endless Visits to Coffee Shops

I spend a lot of time in coffee shops over the holidays. It’s a great place to meet people for a quick catch-up. And to drink coffee. Coffee featured a few times in articles this year here on The Python Coding Stack.

One of these coffee-themed posts followed Alex’s experience with opening his new coffee shop and explored parameters and arguments in Python functions:

Another one narrates one of my trips to a coffee shop and how it helped me really understand the difference between == and is in Python—equality and identity:

Board Games

Who doesn’t play board games over the holidays? We certainly do. And that means we need a dictionary present to resolve Scrabble-related disagreements. This year, we also played Boggle, another word-based game. So, the dictionary had to work overtime.

And dictionaries work overtime in Python, too. They’re one of the most important built-in data structures. Here’s a question: Are Python dictionaries ordered? The answer is more nuanced than you might think at first:

And to understand how Python dictionaries work, it’s best to understand hashability:

This article is part of The Club, the special area on The Python Coding Stack for premium members. The Club launched this year and includes more articles, an exclusive forum, videos, and more… This premium content is in addition to the free articles, which will always remain a key part of The Python Coding Stack. To make sure you don’t miss a thing here on The Python Coding Stack, join The Club by becoming a premium subscriber:

Subscribe now

And it’s not just dictionaries that play an important role in Python. Indeed, in Python, we often prefer to focus on what a data type can do rather than what it is. Here’s another short post in The Club on this topic:

Unwanted Gifts?

Did you receive gifts you don’t need or want? Or perhaps, you have received the same gift more than once? Python can help, too. Let’s start by removing duplicate presents from the Christmas tree list:

And what about the used wrapping paper and food packaging? You can recycle some of it. But some must end up in the garbage. Python has its own trash bin, too:

Magic

This time of year can feel magical. And maybe it’s for this reason that TV stations here keep showing the Harry Potter films during the Christmas holiday season. I’m a Harry Potter fan, and I’ve written Harry Potter-themed posts and series in the past. And there was one this year, too:

One thing that’s not magic in Python is its behind-the-scenes operations. Python’s special methods deal with this, and once you know the trick, it’s no longer magic. Here’s a post that explores some of these special methods:

Queueing in The Cold

I avoided queueing in the cold this year, but I’ve done this so many times in past Christmas holidays. Queueing for a skating rink or for a Christmas fair. Queueing to get mulled wine from a street stall. If you could skip the queue, would you?

And if it’s cold, you’ll need to zip your jacket well. Python’s zipping and unzipping also feature in this year’s posts:

End-of-Year Reflections

Let me spare you all my Python-related stuff—the courses, articles, updates here on The Stack, and all that. Instead, my news for 2025 was my return to an old interest: track and field athletics. I even started a new Substack to document my adventures in track and field:

Back on the Track

My journey as I return back to track and field athletics after a 25-year gap… as a masters athlete (translation: old) and an aspiring coach

By Stephen Gruppetta

And I’ve written some posts with a track-and-field theme, too. Here’s one of these:

But the end of the year is also a time for reflecting on one’s life, past and future. Recently, a Python object has done just that:

Looking forward to a great new year in the Python world and here on The Python Coding Stack. Wishing you all a Happy New Year!

Support The Python Coding Stack

Image by iPicture from Pixabay

Join The Club, the exclusive area for paid subscribers for more Python posts, videos, a members’ forum, and more.

Subscribe now

For more Python resources, you can also visit Real Python—you may even stumble on one of my own articles or courses there!

Also, are you interested in technical writing? You’d like to make your own writing more narrative, more engaging, more memorable? Have a look at Breaking the Rules.

And you can find out more about me at stephengruppetta.com

Gold customers get served first. When no Gold customers are waiting, you serve Silver customers. Bronze customers get served when there’s no one in the upper tiers waiting.

How do you set up this queue in your Python program?

You need to consider which data structure to use to keep track of the waiting customers and what code you’ll need to write to keep track of the complex queuing rules.

Sure, you could keep three separate lists (or better still, three `deque` objects). But that’s not fun! And what if you had more than three priority categories? Perhaps a continuous range of priorities rather than a discrete number?

There’s a Python tool for this!

So let’s start coding. First, create the data structure to hold the customer names in the queue:

“You told me there’s a special tool for this? But this is just a bog-standard list, Stephen!!”

Don’t send your complaints just yet. Yes, that’s a list, but bear with me. We’ll use the list just as the structure to hold the data, but we’ll rely on another tool for the fun stuff. It’s time to import the heapq module, which is part of the Python standard library:

This module contains the tools to create and manage a heap queue, which is also known as a priority queue. I’ll use the terms ‘heap queue’ and ‘priority queue’ interchangeably in this post. If you did a computer science degree, you’d have studied this at some point in your course. But if, like me and many others, you came to programming through a different route, then read on…

Let’s bundle the customer’s name and priority level into a single item. Jim is the first person to join the queue. He’s a Silver-tier member. Here’s what his entry would look like:

It’s a tuple with two elements. The integer 2 refers to the Silver tier, which has the second priority level. Gold members get a 1 and Bronze members—you guessed it—a 3.

But don’t use .append() to add Jim to service_queue. Instead, let’s use heapq.heappush() to push an item onto the heap:

Note that heapq is the name of a module. It’s not a data type—you don’t create an instance of type heapq as you would with data structures. You use a list as the data structure, which is why you pass the list service_queue as the first argument to .heappush(). The second argument is the item you want to push to the heap. In this case, it’s the tuple (2, “Jim”). You’ll see later on why you need to put the integer 2 first in this tuple.

The heapq module doesn’t provide a new data structure. Instead, it provides algorithms for creating and managing a priority queue using a list.

Here’s the list service_queue:

“So what!” I hear you say. You would have got the same result if you had used .append(). Bear with me.

Pam comes in next. She’s a Gold-tier member:

OK, cool, Pam was added at the beginning of the list since she’s a Gold member. What’s all the fuss?

Let’s see what happens after Dwight and Michael join the queue. Dwight is a Bronze-tier member. He’s followed in the queue by Michael, who’s a Silver-tier member:

OK, this is what you’d expect once Dwight joins the queue, right? Dwight is a low-priority customer, so he’s last. Is this just a way of automatically ordering the list, then? Not so fast…

The fourth customer to walk in is Michael, who’s a Silver-tier customer. But he ends up in the last position in the list. What’s happening here?

It’s time to start understanding the heap queue algorithm.

Heap Queue • What’s Going On?

Let’s go back to when the queue was empty. The first person to join the queue was Jim (Silver tier). Let’s place Jim in a node:

So far, there’s nothing too exciting. But let’s start defining some of the rules in the heap queue algorithm:

• Each node can have at most two child nodes—that’s two nodes connected to it.

So let’s add more nodes as more customers join the queue.

Pam joined next. So Pam’s node starts as a child node linked to the only node you have so far:

However, here’s the second rule for dealing with a heap queue:

• A child cannot have a higher priority than its parent. If it does, swap places between child and parent.

Recall that 1 represents the highest priority:

Pam (Gold tier / 1) is now the parent node, and Jim (Silver tier / 2) is now the child node and lies in the second layer in the hierarchy.

Bronze-tier member Dwight joined next. Recall that each parent node can have at most two child nodes. Since Pam’s node still has an empty slot, you add Dwight as a child node to Pam’s node:

Let’s apply the second rule: the child node cannot have a higher priority than its parent. Dwight is a Bronze-tier member, and so he has a lower priority than Pam. All fine. No swaps needed.

Michael joined the queue next. He’s a Silver-tier member. Since Pam’s node already has two child nodes, you can’t add more child nodes to Pam. The second layer of the hierarchy is full. So, you take the first node in the second layer, and this now becomes a parent node. So you can add a child node to Jim:

Time to apply the second rule. But Michael, who’s in the child node, has the same membership tier as Jim, who’s in the parent node. Python doesn’t stop here to resolve the tie. But you’ll explore this later in this post. For now, just take my word that no swap is needed.

Let’s look at the list service_queue again. Recall that this list is hosting the priority queue:

The priority queue has one node in the top layer. So the first item in the list represents the only node in the top layer. That’s (1, Pam).

The second and third items in the list represent the second layer. There can only be at most two items in this second layer. The fourth item in the list is therefore the start of the third layer. That’s why it’s fine for Michael to come after Dwight in the order in the list. It’s not the actual order in the list that matters, but the relationship between nodes in the heap tree.

But there’s more fun to come as we add more customers and start serving them—and therefore remove them from the priority queue! Let’s add some more customers first.

Angela, a Bronze-tier member, joins the queue next. Let’s add the new node to the tree first:

The relationship between parent and child doesn’t violate the heap queue rule. Angela (Bronze) has a lower priority than the person in the parent node, Jim (Silver):

One more client comes in. It’s Kevin, and he’s a Gold-tier member:

There are no more free slots linked to Jim’s node, so you add Kevin as a child node linked to Dwight. But Kevin has a higher priority than Dwight, so you swap the nodes:

But now you need to compare Kevin’s node with its parent. Pam and Kevin both have the same membership level. They’re Gold-tier members.

But how does Python decide priority in this case?

Python thinks that (1, “Kevin”) has a higher priority than (1, “Pam”)—in Python’s heap queue algorithm, an item takes priority if it’s less than another item. Python is comparing tuples. It doesn’t know anything about your multi-tier queuing system.

When Python compares tuples, it first compares the first element of each tuple and determines which is smaller. The whole tuple is considered smaller than the other if the first element is smaller than the matching first element in the other tuple. However, if there’s a tie, Python looks at the second element from each tuple.

Let’s briefly assume there’s a Gold-tier member called Adam:

Python now considers (1, “Adam”) as the item with a higher priority.

The second element of each tuple is a string. Therefore, Python sorts these out using alphabetical order (lexicographic order, technically).

That’s why Kevin takes priority over Pam even though they’re both Gold-tier members. ‘K’ comes before ‘P’ in the alphabet! You must swap Kevin and Pam:

Note that the algorithm only needs to consider items along one branch of the tree hierarchy. Jim, Michael, and Angela weren’t disturbed to figure out where Kevin should go. This technique makes this algorithm efficient, especially as the number of items in the heap increases.

Incidentally, you can go back to when you added Michael to the queue and see why he didn’t leapfrog Jim even though they were both members of the same tier. ‘M’ comes after ‘J’ in the alphabet.

Now, we can argue that it’s not fair to give priority to someone just because their name comes first in alphabetical order. We’ll add timestamps later in this code to act as tie-breakers. But for now, let’s keep it simple and stick with this setup, where clients’ names are used to break ties.

Let’s check that the service_queue list matches the diagram above:

Kevin is in the first slot in the list, which represents the node at the top of the hierarchy. Jim and Pam are in the second layer, and Michael, Angela, and Dwight are the third generation of nodes. There’s still one more space in this layer. So, the next client would be added to this layer initially. But we’ll stop adding clients here in this post.

And How Does the Heap Queue Work When Removing Items?

It’s time to start serving these clients and removing them from the priority queue.

Read more