It would be nice if, just once, we could have a thread about interesting esoteric Python behaviours without people taking it as an invitation to dump their laundry list of things they personally don't (or do, for that matter) like about Python, or to make inane comparisons to other languages (especially JavaScript, for very unclear reasons) while being simply uninformed.
I used to like Python in the 2010s when it felt like a breath of fresh air relative to PHP and Perl.
Now it feels like a weird PHP itself that is slow, brittle, and dangerous to write code at scale in.
The loose typing, potluck standard library, and horrible package manager (insofar as the community does not know how to package code) all feel so dated.
Python certainly has some baggage, especially the typing system (which is still not finished, if you're looking at static typing and so is implemented differently by type checkers) and pip's safety, or lack thereof. But comparing it to PHP or Perl is rhetoric leading you one step too far.
Comparing it with PHP is unfair… to PHP. The amount of hard work that the PHP community has done to advance and keep their language relevant is impressive and admirable, and Python is perhaps the most extreme counterexample there is.
The Python community has spent the last 15 years refusing to improve in any meaningful way, or to learn anything from their peers. As someone who used to choose only jobs that would let me work with Python, I’ve gone through every phase of grief, and now just try to forget that it exists.
Lol, Python has had incredible improvements over the last decade plus, while uv fixed packaging. It's the best/comprehensive glue language ever made, even with a few remaining warts.
Most of the time downvoters don't explain their downvote, but I'll explain mine. I voted this comment down because it's just plain incorrect.I worked with PHP for nearly ten years (and I never want to go back). Maybe PHP has improved since I worked with it (PHP 7.4 was the most recent version when I last worked with it, I have never used PHP 8), but I doubt it.
But to describe the Python community as "spen[ding] the last 15 years refusing to improve in any meaningful way" is just laughably wrong. I can't give details as I haven't been doing much Python work, but even so I know of multiple changes, such as the typing system, or packaging improvements, which have significantly improved the language AFAICT. If there's a reason why you would not consider those to be "improv[ing] in any meaningful way", please enlighten me.
You mentioned two biggies, but also the GIL removal, async, performance improvements, f-string, walrus, fast dicts w. merge ops, data classes, pattern matching, friendlier repl, and hundreds of smaller yearly improvements.
Pattern matching? Nice, I'd managed to miss that one completely, as well as the fact that Python had introduced dictionary-merging (according to a quick search, Python 3.9 introduced the | (pipe) operator for dict unions). I did know about the others you mentioned, but couldn't call them to mind when writing my comment.
But reading through a Python script that I had Claude Code write for me taught me another one: apparently there's now a / operator on strings, because Claude wrote `path = "some" / "dir" / "filename.txt"` without importing anything outside of the stdlib. I presume it is shorthand for calling os.path.join and will therefore apply the correct path separator on Linux vs Windows.
You are both focusing on Python improving in any way but the person you're responding to said "improve in any meaningful way, or to learn anything from their peers"
Which I will say is overstated but I can see where they are coming from even when you bring into scope things like types, async, etc.
First types. This is what the type hints in Python allow you to do:
def foo(x: int) -> int:
return x
foo("hello")
We can say that this is great Python has type hints, but at the same time the lesson they learned is wrong because the feature to have isn't type hints but actually enforcing them so the above code cannot be written. In my eye, this is an anti-feature.
Second async, this is also the wrong lesson to have learned from other languages. Adding async/await is a bandaid over the problem that the synchronous imperative model clashes with asynchronous distributed semantics. The async/await keyword are a way to try to bridge between the two, but it creates a "function coloring" problem that all these languages which added async/await have.
The lesson Python should have learned is to not add these keywords and go back to its glue language root, allowing actually natively asynchronous languages to coordinate asynchronous processes, while Python code handles the synchronous core. Python doesn't have to be everything, the wrong lesson was to try to be the one language to rule them all.
You also brought up the packaging improvements, which I feel were the wrong lesson learned. The problem with the Python packaging ecosystem is well known since it's been expressed in the XKCD comic. The lesson from other languages is: one blessed compiler toolchain integrated into the packaging story makes for a better user experience. This is the npm, cargo lesson. For Python to really learn it, uv or equivalent would be the blessed way of managing Python projects. Instead it's still a very fragmented landscape with many competing solutions, which goes against Python's own zen.
Moving on to pattern matching, again I feel the wrong lesson was learned. Pattern matching is a feature from the functional paradigm that in my opinion became more popular with developers when they became exposed to it in Rust, and so Python joined in and added the feature as well. But the reason it's such a nice feature in Rust is because it will refuse to compile any code that does not do exhaustive matching on all variants. This is great because it catches problems early and forces you to consider the non happy path where things can error. So pattern matching alone isn't the feature it's pattern matching PLUS structured Enums and exhaustive patterns.
So in Python you can do this:
x = 3
match x:
case 1:
print("one")
case 2:
print("two")
print("done")
Output will be "done" rather than an error on the match, which is what should happen if they had learned the right lesson, because this is no better than an if or switch. Again I consider this an anti-feature -- better to not have at all if it doesn't work as expected.
Anyway, I'm not trying to say Python is bad, I'm just saying I get where the other poster is coming from when they say Python has refused to learn the right lessons. Although I would not agree they haven't improved a lot over 15 years.
Not every improvement is perfect, that’s true. I in particular fought against walrus syntax and don’t use pattern matching for other reasons. Python is also boxed in by its history. Some of the things you want done are incongruent with its design.
Many people like them however. But you’re being way too charitable to that dumbass comment.
I guess what I’m trying to do is draw a distinction between “improvement” and “change”. We can agree to disagree but my view is the things I called anti-features are not improvements at all, but rather loaded footguns. Python would be more coherent language without async/unenforced type hints/non exhaustive pattern matching. Like you say, it’s constrained by its history and I feel a lot of the recent changes are to make Python more palatable to the devs who are using it for AI and ML rather than keeping it true to to its nature.
Type hints are very helpful on big projects. And are enforced by other tools. Honestly, this kind of criticism shows a lack of experience, and denial of reality for an almost 40 year old language.
It sounds like scoffing at a Silver medalist to me.
In fact, Python is better at what it does than almost anything from the era. That’s why we use it. Nothing is ever going to be perfect because better solutions are emergent, and reverse time travel does not exist.
> Honestly, this kind of criticism shows a lack of experience, and denial of reality for an almost 40 year old language.
Okay that's where you lost me. You can talk about the language from your perspective, and like I said we can agree to disagree, but it crosses a line when you want to comment on others' experience and put them down just because they disagree with you. No on is scoffing here, what I wrote was a considered criticism. Have a nice day.
It’s about as near a well-documented fact as there is in this industry, and not intended to be an insult. Sometimes we need a wake-up call, and yes they’re not usually enjoyable. I know as well as anyone else, having received my share over the years.
I don't understand how people talk about how Python is "easy to learn for beginners" or "easy to understand." To me it's so hard to remember and follow all the weirdness. Racket / Scheme / I dare say even Haskell would just be so much simpler for learners.
I'm with Conal Elliot when he said on Type Theory for All that it is sooo much harder to understand a program in Python.
> Racket / Scheme / I dare say even Haskell would just be so much simpler for learners.
I have used all three languages; and you clearly have no idea of the notion of usability of a language. So many things contradict this, let me list them off the top of my head
- Getting a running toolchain working: Prexisting (most OSes bundle a Python interpreter) or a package install away for Python. Scheme / Racket is some odd mix of custom IDEs with Dr. in the name, or someone's 20 page essay on how SLIME is the best thing ever. Haskell gets into odd stuff with ghci, cabal, and stack, and all of them are extremely slow.
- Tutorials: Python has a ton of them, they all get you printing to stdout and calculating things in about 10 minutes. Scheme / Racket typically spends multiple chapters navel-gazing about lists, cons, and such. Haskell is actually better in terms of the Hello World stuff, but ghci v/s ghc bites you again; and no one has a clear idea of which one to use.
- Advanced concepts: Python has mainstream but halfhearted OOP; and things like decorators and metaprogramming. Quickly intelligible if you learned something else like Java or C++. Or if you learned shell scripts you can get quite a bit done with just imperative. Racket/Scheme: 3 chapters in and you're still trying to figure out tail recursion. Haskell: Instead of just doing fun things with take and foldl you're being hit with trivia about typeclasses.
> Haskell: that was funny but an absurd criticism ghc vs ghci?? Nobody has that problem.
Back when I was a beginner actually interested in getting out of the beginner step of Haskell, this was an issue for me times. So there's at least one person :)
Also anecdotally I have seen people ask this in Freenode #haskell as well (the “Freenode” probably tells you how long ago this was) ; and there a few issues [1] and [2] where I see beginners having the same/similar issue. The second one is particularly funny, 4 people give 5 solutions and no one seems to know what the actual fix is. Instead you have people arguing whether a repeated do works or not. This would never happen with Python, just saying :)
> Racket: criticizing for having a beginner-friendly IDE doesn’t make a lot of sense.
Sorry, perhaps too harsh but I don't think it's as beginner-friendly as you think. I think it would probably help if they made the design more modern and welcoming. All these details about you can rewrite entire languages in Lisp and we can't even at least get a GUI that looks like it was written after 2007?
I can understand that "advanced" python programs may be difficult to understand for beginners (lots of implicit/hidden behaviors, possibility to change basically everything one should expect, etc).
But to _learn_ programming, I really, really don't see how using Haskell would be simpler than Python. Perhaps if you have a specific background (e.g., math), but else python is almost pseudo code already. You'll really have to convince me that a more abstract language is better...
Haskell’s hard to interpret error messages alone disqualify it from being a beginner language.
Python: errors based on incorrect indentation (many beginners don’t use nice IDEs), or don’t understand the meaning of the hints) and scope (don’t forget your “global” if you’re hacking in PyGame) are challenges.
Starting with 3.13, even the REPL is a sufficiently nice IDE to avoid any careless indentation errors from poor formatting (as opposed to ones caused by actually not understanding how many levels of indentation the current line of code should have).
This is literally my point. A python example would be like, calling readline to store someone's name into a name variable then printing a string that said "Hello name". Scheme examples are some abstract list manipulations.
Who cares tho. The agents deal with all of that, if you’re still looking at the code or caring about anything other than the loops and orbs you’re at the wrong level of abstraction. The important thing is the models have tons of python in their training data.
I feel the same way. It was, back then “the second best language for everything, the first best at nothing”
Can’t take credit for the quote, read it somewhere.
The whole language changed when they kicked what’s-his-name out, and it’s a tool I almost never reach for anymore, whereas 15 years ago it was my Swiss Army knife.
I felt the same way about moving to Python versus PHP and Perl.
I still really enjoy using python though. It's not really a fair comparison because I hadn't used PHP and Perl for as long but I just don't hit some mystifying issue every single session like I did with those languages when I'm using python. I honestly have never even read about that __debug__ constant. It's fun to hear about it but it's just not something that's comes up much.
Perl is a lot like that also, you can read about some really weird old features like $[ but you never see that in practice, you just write code with variables and functions and so on.
The __debug__ constant is really weird - any block of code guarded with `if __debug__:` will be entirely omitted from the bytecode under PYTHONOPTIMIZE=1. This and `assert` are the only two examples of real “conditional compilation” in Python. This is also the reason why you cannot assign to __debug__: doing so would make it possible to invalidate the compiler’s assumption about `if __debug__:` statements.
I honestly have never even heard of this constant and I feel like I've been using python for a pretty long time. Although maybe my memory for some things just gets garbage collected if I don't use it enough. Does it actually get used that often in real world code? Seems like it might be kind of risky.
I see asserts used in production code as part of flow control way too frequently, so I assume the majority of python users aren't aware of the -O flag, much less this behavior- which I too haven't ever heard of.
Of recently, I've noticed claude is a big fan of asserts too.
I feel like it's the kind of thing you might wind up caring about if you're micro-optimizing your python, but in my experience that's a losing game and you're better served rewriting it in another language than bothering with trying to speed up the execution of the raw python code (it's not that you can't optimize python code, but only in broader strokes. If you are looking at the bytecode you're in too deep and every time I've seen it tried the code has been ported shortly afterwards).
Ditto. I’ve certainly never used it and can’t recall seeing it in any codebases I’ve worked on or looked at. Sounds interesting though!
I’ve of course certainly heard of, seen, and used `assert`, but more often than not, outside of pytest, I see its use way more in potential footgun scenarios—I doubt that many people know that assertions can be silenced, and that they’d probably be better off raising exceptions in many cases where they’re using `assert`.
I made a CTF problem where `assert` was used as a critical safety check - and where "accidentally" running the program under -O (for speed!) resulted in a security vulnerability. A large fraction of the people who attempted the problem seemingly missed this bug.
I would not be surprised in the least if that pattern existed in the wild. In fact, it's quite common to see this in C/C++ codebases too: people will use assert() to check a security-relevant property, and then disable those checks in their release builds "because it can't happen".
Because Python lets you get as far as it does without formally learning everything, but is also expected to suit a huge variety of use cases, it ends up with lots of hidden details that are irrelevant to most users.
The most direct way to find out about `__debug__` is to read `python -h` (or the usage message, which is not all that easy to trigger) in full, and then head over to the documentation.
> Does it actually get used that often in real world code?
If 0. Etc. are also compiled out, at compile time __debug__ is simply False or True and the existing optimization paths take care of it.
Assigning to __debug__ wouldn't do anything to the compiler as it never actually reads the variable, so assignment would just cause weirdness from other use
Yeah it was a blast to make. If memory serves, a significant piece was over a very lovely and piney joint with Kieran Prasch at an airbnb in... Seattle?
Hmm. Having read your post, surely the bug is having a function where
set(foo, false)
removes foo entirely. What if you want foo to have the value False? Even besides the unintended behaviour where 0 is coerced to a boolean value, this function seems poorly designed.
Misery is trying to retrofit "bool", True/False, and nil/null to a language. C had to do that. Python had to do that. Getting those wrong is one of the classic language design mistakes. It seems like treating "True" as a value that equates to 1 will work, but then the special cases get you. Like being able to perform arithmetic on True.
Common language design boners:
- Not building in strings. That's now in the past. Everybody has strings. (Well, C...)
- Not building in multidimensional arrays of the numeric types. Everything that number-crunches needs them, and having multiple definitions is Not Fun and may lead to expensive re-copying between different libraries.
This is an enormous blind spot in language design. It's one of the reasons FORTRAN, which has good multidimensional numeric arrays, is still often used for number-crunching.
- Not standardizing the small vectors (vec2, vec3, vec4) and their matrix friends. Graphics code depends on these, and it's really annoying if there are multiple slightly incompatible implementations. Especially since GPUs have hardware for those types, and you want CPU and GPU to use the same representations.
- Not having arrays of bits. Pascal had PACKED ARRAY[0..N] of BOOLEAN but that was lost in later languages. It's useful to have that as a language construct, because most modern CPUs have good hardware for dealing with bit strings, and you'd like the compiler to use it.
Most useful languages acquire these features, but, when they come in late, there are multiple similar implementations, and libraries made incompatible by depending on different implementations.
(Amusingly, when Second Life switched from Linden Scripting Language to Luau, they initially had True, TRUE, and true all in use, as different types with different semantics. I was able to persuade the devs to unify the boolean types.)
> Not having arrays of bits. Pascal had PACKED ARRAY[0..N] of BOOLEAN but that was lost in later languages. It's useful to have that as a language construct, because most modern CPUs have good hardware for dealing with bit strings, and you'd like the compiler to use it.
I'm not sure exactly which features are responsible (I'm inclined to blame templates), but C++'s std::vector<bool> is a rough edge. For those unfamiliar, the standard specifies this vector template in a way that's not compatible with other vectors.
They should have made `std::bitset<std::dynamic_extent>` what todays `std::vector<bool>` is (actually maybe not, `std::bitset` is fixed sized, just compile time fixed size). While at it also make `std::array<std::dynamic_extent>` a runtime fixed size array.
Vectors & multidimesional arrays are something I'm 100% adding to my language's core.
It kind of started with vectors as the very first feature (I was sick & tired of libraries reinventing their own `Point`/`VectorN` in incompatible ways).
Strings are a really weird data type. I'm not sure you can do much better than C strings without implicitly requiring dynamic memory allocation, which C deliberately does not do.
Definitely agree on multidimensional arrays. I feel like efficient arrays in general are underrated in high-level language design.
> Strings are a really weird data type. I'm not sure you can do much better than C strings without implicitly requiring dynamic memory allocation, which C deliberately does not do.
The thing you want is what Rust delivers in the box, &str a string slice reference type, in Rust's case the "string" is UTF-8 encoded text. On the bare metal the way to represent this type is as a "fat pointer" typically a pair of registers, one with the address of the first byte of the string and the other with a length.
C should have fat pointers, they were proposed, for IIRC C89 but the proposal was rejected. That's pretty sad, the fat pointer is expensive to the point of maybe feeling extravagant on a PDP-11, but by 1989 that's long gone.
More ridiculously C++ didn't get this type (which it eventually called std::string_view and provides in its standard library not as a built-in) until 2017, years after Rust 1.0 shipped. In the meanwhile C++ just did not have a sensible way to do this, strings are hard apparently.
The string buffer feature, allowing you to actually make strings is less important, as you say it will need an allocator and so on very bare metal you might not have this - but the string slice reference doesn't need an allocator.
I think it's worth delivering the basic "it's a growable array type, duh" implemenation of the string buffer type, which is what Rust's String type is, but C++ chooses to ship an oddly specific small-string optimized version as std::string right from the offset.
Oh wow, I had no idea that SL did another language change after migrating LSL to Mono. Surprising considering that happened late '00s/early '10s?
I'd consider LSL to have been foundational in my ultimate interest/career in software engineering. The strict typing, very usable compile/runtime errors, and good documentation/examples made it so easy to pick up as a teen. Not to mention as long as you didn't edit/save a script again it would always run the same regardless of updates.
You list a few absences but absences aren't the end of the world, I say it's worse when designers make a booboo where the language semantics are wrong. In C++ there are so many of these it's not sporting but a recurring example from the garbage collected languages would be the for-each loop mistake.
Several times now†, people make a language where the way a for-each loop (for each Goose in Geese ...) works is that there's a single variable Goose and each time around the loop we change which value is referred to by the Goose variable. This seems intuitively like a reasonable way to do this. But it's wrong and eventually your programmers will get nasty surprises. What you actually should deliver is an implementation where each time around the loop there's a new variable named Goose, that variable goes away at the end of that iteration and will be replaced by the next one, with the same exact name.
> What you actually should deliver is an implementation where each time around the loop there's a new variable named Goose, that variable goes away at the end of that iteration and will be replaced by the next one, with the same exact name.
Is this because a closure inside a loop will capture a reference to `Goose`?
I think that this is a capture problem not a variable problem. The closure should always do the right thing and capture the value of all variables (not just ones inside the loop), instead of capturing the reference to the variables.
Then the general problem is fixed to match what developers expect, instead of a specific instance of that class of problems being fixed and working differently to how other captured variables work.
> I think that this is a capture problem not a variable problem. The closure should always do the right thing and capture the value of all variables (not just ones inside the loop), instead of capturing the reference to the variables.
Now your "lalanthran closures" can't mutate the world because they work exclusively with copies not references, if they try to mutate something then whatever they're touching was just a copy not the real thing.
> Now your "lalanthran closures" can't mutate the world because they work exclusively with copies not references, if they try to mutate something then whatever they're touching was just a copy not the real thing.
That is true. I still don't like the idea of "Here is a general rule. It applies everywhere but $HERE." Whether that general rule is "All captures are by value" or "All captures are by reference", the rule should not have exceptions based on context in the code.
A better tradeoff would be to have the general rule (whatever it is) apply everywhere, along with syntax for capturing (or not, depending what the default is). I'd rather have it grab everything by value, and for those things that are susceptible to race conditions (because more than one closure is modifying it), explicitly annotate it with a sigil (`&`, or a keyword, or similar).
I mean, in pseudocode, when I see:
... variables x, y and z are declared and used in this scope ...
return (x, y, x) => { ... }
I don't want to have to examine the surrounding scope to know whether or not `y` is susceptible to a race. I'd rather just see:
... variables x, y and z are declared and used in this scope ...
return (x, &y, x) => { ... }
An alternative viewpoint is that many languages have immutable variables and they seem to be getting along just fine without needing mutation on variables, shared or otherwise.
> the rule should not have exceptions based on context in the code.
But the rules didn't and still don't have any such exceptions.
> A better tradeoff would be to have the general rule (whatever it is) apply everywhere, along with syntax for capturing (or not, depending what the default is)
This "solution" is how it works in C++. We can thus castigate the programmer for writing the wrong runes in their captures list and never for a moment doubt that we got it right when we introduced so very many footguns...
Tony Hoare's observation applies "One way is to make the program so simple, there are obviously no errors. The other is to make it so complicated, there are no obvious errors."
> An alternative viewpoint is that many languages have immutable variables and they seem to be getting along just fine without needing mutation on variables, shared or otherwise.
Sure, and one of the astonishing things in C# or Go before they fixed this is that you can indeed have immutable variables which change, even though that's silly - the language can decide that you mustn't change Goose, but it doesn't need to obey its own rules because it will change it for each loop iteration.
The interior of a for-loop is only a scope, not a closure. In most non-dynamic languages you can't package up the state and hold onto it beyond the life of the loop, which is what closures are for.
Most trouble in this area came from the iteration variable outliving the loop. That's not good when the iteration variable is a pointer. In C, it often is, and at the end of the loop, it points to an invalid address. It was a change to C (when?) to make the iteration variable go out of scope before code after the loop could get at it.
Yes I remember a friend doing university marking for a beginners programming course years ago and some student had managed to swap True and False making their assignment very wonky.
I don't. I have no interest in trying to assign a value to something that's built-in and not meant to be written to. Like who fucking cares that the boolean constants are actually weird little structure that sometimes let you mess with them and other times your edits are ignored? Maybe this is helpful for writing an entry for an obfuscated code challenge but I'm not doing weird shit like that with the code I expect to work between various Python versions and implementations, especially when I'm getting paid to do it.
Please. Such a charged response wasn’t justified at all. Different people are curious about different things. You ask “who fucking cares?”. The answer? You don’t, and the person you’re replying to does.
Python is awful. There are so many one offs in libraries, none agree on a style, it’s slow, and it’s way too easy to do the wrong thing. I often work with data scientists and have to productionize their jupyter notebooks which is pure suboptimal hell. I guess it must be a good easy learning curve for research/scratchpad
> I often work with data scientists and have to productionize their jupyter notebooks
I'm not a huge Python fan, despite working with it fulltime, but this feels like mixing correlation and causation. Data scientists would not be writing good, optimized code in any language.
Python is amazing compared to writing sh/bat scripts. Different languages are for different purposes, using eg. Rust to write system scripts would just be mental. Whether people abuse those languages for purposes they were not intended for is another story, but that doesn't mean the language is inherently bad.
C'mon man, I don't know any mid and above python developer who seriously has ever considered programming in Jupiter Notebooks. Python is not slow, it's you being the issue. If you are an amateur then it's easy to do the wrong thing, that's true.
The performance hell thing is also also kind of a virtue, though. The language is awful, but the libraries you need to use to do any real compute (numpy, torch, sympy, etc) keep you in a few pretty well-constrained patterns that are easy enough to translate.
If you've ever read through FORTRAN code from a mathematics department or MATLAB/C/C++ from (non-software) engineering disciplines, then you can kind of understand why productionizing a jupyter notebook is not the worst of all possible worlds.
I've never become a fan of the language syntax, but otherwise I've become quite smitten with the total Python ecosystem. The Agents/LLMs + uv combo have made Python so useful and productive for me.
My CLI tools publish from Github to PyPI so that I can run tools with just `uvx sql-agent-cli` or `uvx dlna-here. Nothing for me to handle downloading (directly myself), no environment to manually setup, portable (Linux, Windows, Mac, ARM, x86). Easy for agents to run from a skill.md file without any other prereq than uv.
Really useful library ecosystem to leverage. No more shell scripts, or TS/JS/PHP backend services. I've even used Python on devices I've built around Raspberry Pi Zero 2 boards.
The imports/packages situation is terrible in general. This was basically broken until uv, and uv is still not the default.
And it's weird how you import files. They're dot-separated packages that resemble file structure but not exactly. NodeJS has a self-explanatory require("./foo.js") or "../foo.js". The newer JS `import` syntax is annoyingly different from `require` but not terrible.
Relatively new, scripts can define dependency metadata now adays to achieve that to some degree. Any pip style versioning including an exact match works there.
I used to build quant investment notebooks that had to be deployed in production. Lots of problems with that. Mine were: Notebook cells run out of order, so you often have something that works in a session, but not in a fresh run. Developing against limited datasets, so you fail against things you didn’t know to test for. Small adaptions that have to be made every time the notebook is translated into a code file.
We streamlined it by making a graph-structured Computation a first class object that tracked staleness as code or data was updated. Then that class could be directly published, and when failures happened in production, the graph could be serialized with the inputs and intermediate calculation data that caused failure, for investigation in a notebook.
Python is a language for "consenting adults". It doesn't try to prevent you from doing awful things so you can do great things. People who can't program well are given plenty of rope to hang themselves. It shares that with Perl and Ruby.
That said, I find it the nicest, cleanest option of the three. I still wouldn't use it for large and complex projects. I really like it for stuff where one might otherwise use shellscript. It's way way better than shellscript... except if it's all about files and running external commands.
>language for "consenting adults". It doesn't try to prevent you from doing awful things so you can do great things. People who can't program well are given plenty of rope to hang themselves.
This is exactly what I remember being said about C (which I agree with) and often given as a reason why higher level languages like Python or Java have so many protections against things C/C++ allowed (memory management being the biggest one of course). Very funny, and I assume not coincidental, to read this about Python in the modern programming landscape.
Well, there are plenty of safety mechanisms that Python doesn't have and dangerous (usually powerful, occasionally badly designed) mechanisms that it does have.
Why wouldn't we? It's still a very relevant distinction, even if the terminology is a bit weird (since scripting is by definition programming). A programmer has very different needs when he writes a script to automate some server tasks versus a complex piece of software. It makes perfect sense that different tools will be more or less effective at meeting those different needs.
There is no reason why automating a server task cannot be a complex piece of software. I think you’re not aware of just what server automation is used for nowadays in countless cases. Also, Python is more popularly used in ML and web-development areas compared to server automation, so that would mean it’s not a scripting language by your logic.
You don’t like working with data scientists. The data science Python ecosystem is really a separate beast that’ll have “normal” coders scratching their heads at the best of times, some of the most popular packages do all sorts of metaprogramming, and the standards for code quality are very different. Don’t blame the language. Well, blame it only in that it allows such things in the first place, which does have some very nice precipitations now and again, as well as some very bad ones.
In an age where people are still standing by C over memory-safe systems programming languages, I feel quite comfortable depending Python for the great many things that Python is good at.
How is that a language problem? Data scientists are not engineers. No matter what language you give them, they will hand you something you are going to have to polish for production.
The fact that Python has become the language of choice for machine learning and data science is not a language issue.
I’ll never not be bitter than Python “won” the scripting language war over Ruby, more or less just because someone did a bit of AI work in it first and it took over that space by default.
Ruby has such a nice holistic consistency to it. With a few exceptions, it feels like it was conceived of by one person with a core idea in mind. Python feels like a mess.
Long before AI work, numerical data crunching is what Python became popular for among non-computer scientists and this led directly to the AI use cases. The reason was obviously the lower barrier to entry without having a software engineering background. I also share with you that feeling about Ruby in particular.
Python's strength is that it's easy to make C libs work in it. That's also why CPython is de facto the only Python implementation and stuff like PyPy never took off.
I like Python but this was always actually one of my pain points. The CPython C API is full of foot guns, the API surface is expansive, and writing against it requires a lot of careful care. Anyone who's ever written C modules for both languages would be able to attest how much more pleasant the experience was for Ruby than Python.
> I’ll never not be bitter than Python “won” the scripting language war over Ruby, more or less just because someone did a bit of AI work in it first and it took over that space by default.
This is just my personal opinion with no data to back it up, but I suspect that Python "won" because it has excellent Windows support, while Ruby doesn't. Even a decade ago, Python's website offered an official native Windows installer [0], while Ruby's website [1] still points you to a third-party installer, which doesn't even have native support since it uses MSYS2 [2].
Most non-developers use Windows, so if you're choosing the first language to teach a large group of people, good Windows support is fairly important. Python being the "default" introductory language gave it a huge number of users, then I suspect that everything flowed down from there.
> but I suspect that Python "won" because it has excellent Windows support, while Ruby doesn't. Even a decade ago, Python's website offered an official native Windows installer
I think you might be able to go an additional decade backwards. Back in college most of my friends were on windows and one of them was using python for class projects.
This could be it. The classic student with a Windows laptop and git-bash installed, where they think git and bash are the same thing, also PuTTY.
I also wonder how many people gave up on Python just because the installer doesn't put it in your PATH. You'd install Python then no python, wtf. Ok so https://discuss.python.org/t/python-command-not-found/22255 ... Then you fix it and it runs the wrong version of Python.
Did you mean to say that Ruby doesn't link name spaces to file system paths? Ruby has namespaces and they're far more flexible than Python's... too flexible in my opinion, making it harder to find things.
Ruby is beautiful in its design, but it made imports and namespaces (a.k.a. modules) separate concepts. This is so flexible it became hard to ever find anything easily in practice.
Likewise, it made classes incredibly easy to extend, which led to a monkey patching bonanza and far too much magic everywhere (Rails being by far the worst offender but not the only one). It meant having to keep too much stuff in your head and needing deep framework/library familiarity just to be able to understand basic code.
In the end, I feel like the incredible flexibility was Ruby's undoing, not just lack of library availability for a specific popular application. I was a Ruby zealot at one point but it began to lose its lustre not because it wasn't beautiful in theory, but because it was inconvenient in practice. In some ways, Python's restrictiveness became its greatest attribute. Then Python 3 helped to fix a lot of the inconsistency.
Both Python and Ruby have a consistent logic to them, just not consistent with each other. Just as all attributes are methods in Ruby, so all methods area attributes in Python, etc. Things get much easier in either language if you stop fighting their internal logic.
When one writes jupyter notebooks for DS you are not writing python. If you ask 10 DSs explain to me what python's attribute lookup model is and why is it different from other OO languages like say Java or C++, they would not care about it. The only thing DSs care about is the rich DS Library support and fast speed of protoyping. To a DS using jupyter this is almost the same feedback loop as a type system at compile time.
Have you tried using `uv`'s newer tools? They help a lot e.g. with linting speed, lock management, package dependency separation, correct python version mgmt and no need to fudge with venv.
Spoken like someone who hasn't tried to get data scientists to use other languages productively, where they'll be missing half the libraries, will have to triple their dependency count because you can't count on large common libraries and will have to dig to the ends of GitHub to find random functionality etc.
A lot of Python design decisions have felt weird and off to me but they’ve long justified it by saying that it’s those little ugly design choices that make the language so usable and effective in practice compared to more well-designed languages that hardly anybody uses. I’m not enough of an expert to clearly say if that’s really true, but imo, there’s a repeated pattern of slightly weirdly designed languages becoming super popular: Python, Javascript, perhaps C as well. Or, maybe we only notice the weirdness because these languages are used so much and get nitpicked to no end.
It's not the ugly design choices that make it popular. It's just that it was the first that did a decent job of being easy to read/write. Even if you write an objectively better language, you can't replace the ecosystem of libraries, education material and human support that comes with popularity. Popularity begets popularity, and since ease of use is the selling point, it's probably never going to practical to replace it as a general purpose go-to for beginners. And where people begin, they tend to stay. So instead we slowpy march on, bearing the burdens of original sin.
Python has some absolutely kick-ass libraries, even without C. It has Django, for those of us who like developing web apps but never could fall in love with Ruby on Rails. And Django is amazing. I've also yet to see a better language for writing quick ETL scripts and pipelines. Also, an 'I need a script for $SYSADMIN_TASK but I want to be able to read it later.' Anything dominated by external latencies (web, databases, etc) will be fast enough for many uses in Python.
Sure, it's not a language to write a web browser or game engine in. And it is slow. But it has some very strong niches outside of ML/Data science. Personally, I love it. To each their own.
It also helps prevent people from ever redefining them. I mean, if you try to redefine False (in a version of Python that allows it) then you deserve everything that's about to happen to your code... but at the same time, it could possibly lead to a security attack. Redefine False then import some module and get unexpected behavior that you can manipulate to your advantage, somehow. I don't know how that would work, it probably wouldn't... but there's also no reason not to lock those names in and prevent them from ever being redefined.
Python is just such a weird language in general despite its popularity that I honestly cannot recommend anyone who starts programming to choose Python as their first language, contrary to popular sentiments. I mean, I was one of the first person to start using Python when I was in grad school almost a decade ago when everybody else in my field was still using Matlab for their lab code, for the simply reason that Numpy was less awful than Matlab and I needed something that can easily print graphs to PDFs.
The only thing good I can say about Python nowadays is that it's easy to get started for the first five minutes, and then you'll have to deal with all of its weirdness: significant whitespace, truthiness, duck typing, GIL, distribution/packaging, etc, etc.
I was a big fan of Julia as the potential replacement for Python for science for such a long time and I had evangelized it a lot previously, but recently I've been more and more convinced that JIT/multiple dispatch was only good if you already know how to program well to begin with, which for a lot of academics who are not working in computer science, they write quite horrific code. I think it may be better off to skip Python altogether and write your code in a statically typed language to begin with.
Are academics who are not working in computer science interested in learning statically typed programming languages?
In the past, the usual answer to people who need a programming language but did not want to learn programming was to give them a domain specific language that focused on solving the specific problem they wanted to solve.
Well, many times in my field of mechanical engineering, they have to, because CFD and FEA are very performance sensitive. The professors I had were still writing FORTRAN and C++ before, but maybe they've switched to Rust now.
> I honestly cannot recommend anyone who starts programming to choose Python as their first language, contrary to popular sentiments
> I think it may be better off to skip Python altogether and write your code in a statically typed language to begin with
Having a good REPL is a huge advantage for beginners (and expert users too), but I'm not aware of any (popular) statically-typed languages with a good REPL.
Despite Python's many faults, it's easy to install (especially on Windows), it has a large standard library, there are third-party packages available for essentially everything, it comes with a user-friendly REPL out-of-the-box, and it gives comprehensible error messages. I'm not really aware of any other (popular) languages with all these attributes.
> Having a good REPL is a huge advantage for beginners (and expert users too), but I'm not aware of any (popular) statically-typed languages with a good REPL.
scala's repl is decent. It has its annoyances, but so does python's (white space sensitivity + repl + terminal emulators stuck in the late mid century don't mix).
I hate Python and I'm onboard with putting it down in many ways, but significant whitespace isn't weird in a first programming language. It's only weird if you've absorbed from some other language the convention that whitespace shouldn't be significant.
Haskell, Lean, and Agda too. But even if Python were the only one, it wouldn't be weird to someone for whom this was their first programming language. It would just seem the way programming languages are. There's nothing intrinsically weird about indentation being significant. It's quite visibly part of the code you write and read.
> There's nothing intrinsically weird about indentation being significant. It's quite visibly part of the code you write and read.
The idea is fine, I guess, although I certainly don't care for it. Where it gets most nasty is that whitespace that looks the same (in your editor) might not be equal and will cause you pain.
> Where it gets most nasty is that whitespace that looks the same (in your editor) might not be equal and will cause you pain.
First off, what editor could I end up using in 2026 that makes this a realistic problem? Even the most basic editors I know of have options to convert tabs to spaces automatically (and any responsible teacher will tell the student to indent with spaces), and to continue the previous line's indentation automatically.
Second, modern Python is stricter about this, and also gives clear error messages.
> and then you'll have to deal with all of its weirdness: significant whitespace
By "weirdness" here you apparently mean not having to worry about matching up curly braces, and getting what you want automatically just for indenting your code the way you're supposed to indent it anyway... ?
> truthiness
Which is different from how it works in other similar languages, how exactly?
> duck typing
Which is weird, how exactly?
> GIL
You can go a lot further than five minutes in Python without having to worry about threads at all, and if you do attempt threading, unless you're writing C extensions, the worst thing the GIL does is deny you the multi-core processing you thought you were going to get.
> distribution/packaging
Tons better now, but it was honestly never difficult, people just didn't care.
If you count pre-release versions, there are actually 7 pre-declared constants, since Python 3.15 (planned for release in November [0]) adds a new constant "TYPE_CHECKING" that should behave like "Ellipsis" and "NotImplemented" do right now [1].
you could shadow True for twenty years and Python just shrugged, then one day it's a SyntaxError and every tutorial you ever wrote breaks. peak Python, honestly.
152 comments
[ 0.26 ms ] story [ 16.7 ms ] threadNice to see it get attention this time.
It would be nice if, just once, we could have a thread about interesting esoteric Python behaviours without people taking it as an invitation to dump their laundry list of things they personally don't (or do, for that matter) like about Python, or to make inane comparisons to other languages (especially JavaScript, for very unclear reasons) while being simply uninformed.
Now it feels like a weird PHP itself that is slow, brittle, and dangerous to write code at scale in.
The loose typing, potluck standard library, and horrible package manager (insofar as the community does not know how to package code) all feel so dated.
The Python community has spent the last 15 years refusing to improve in any meaningful way, or to learn anything from their peers. As someone who used to choose only jobs that would let me work with Python, I’ve gone through every phase of grief, and now just try to forget that it exists.
But to describe the Python community as "spen[ding] the last 15 years refusing to improve in any meaningful way" is just laughably wrong. I can't give details as I haven't been doing much Python work, but even so I know of multiple changes, such as the typing system, or packaging improvements, which have significantly improved the language AFAICT. If there's a reason why you would not consider those to be "improv[ing] in any meaningful way", please enlighten me.
But reading through a Python script that I had Claude Code write for me taught me another one: apparently there's now a / operator on strings, because Claude wrote `path = "some" / "dir" / "filename.txt"` without importing anything outside of the stdlib. I presume it is shorthand for calling os.path.join and will therefore apply the correct path separator on Linux vs Windows.
Which I will say is overstated but I can see where they are coming from even when you bring into scope things like types, async, etc.
First types. This is what the type hints in Python allow you to do:
We can say that this is great Python has type hints, but at the same time the lesson they learned is wrong because the feature to have isn't type hints but actually enforcing them so the above code cannot be written. In my eye, this is an anti-feature.Second async, this is also the wrong lesson to have learned from other languages. Adding async/await is a bandaid over the problem that the synchronous imperative model clashes with asynchronous distributed semantics. The async/await keyword are a way to try to bridge between the two, but it creates a "function coloring" problem that all these languages which added async/await have.
The lesson Python should have learned is to not add these keywords and go back to its glue language root, allowing actually natively asynchronous languages to coordinate asynchronous processes, while Python code handles the synchronous core. Python doesn't have to be everything, the wrong lesson was to try to be the one language to rule them all.
You also brought up the packaging improvements, which I feel were the wrong lesson learned. The problem with the Python packaging ecosystem is well known since it's been expressed in the XKCD comic. The lesson from other languages is: one blessed compiler toolchain integrated into the packaging story makes for a better user experience. This is the npm, cargo lesson. For Python to really learn it, uv or equivalent would be the blessed way of managing Python projects. Instead it's still a very fragmented landscape with many competing solutions, which goes against Python's own zen.
Moving on to pattern matching, again I feel the wrong lesson was learned. Pattern matching is a feature from the functional paradigm that in my opinion became more popular with developers when they became exposed to it in Rust, and so Python joined in and added the feature as well. But the reason it's such a nice feature in Rust is because it will refuse to compile any code that does not do exhaustive matching on all variants. This is great because it catches problems early and forces you to consider the non happy path where things can error. So pattern matching alone isn't the feature it's pattern matching PLUS structured Enums and exhaustive patterns.
So in Python you can do this:
Output will be "done" rather than an error on the match, which is what should happen if they had learned the right lesson, because this is no better than an if or switch. Again I consider this an anti-feature -- better to not have at all if it doesn't work as expected.Anyway, I'm not trying to say Python is bad, I'm just saying I get where the other poster is coming from when they say Python has refused to learn the right lessons. Although I would not agree they haven't improved a lot over 15 years.
Many people like them however. But you’re being way too charitable to that dumbass comment.
It sounds like scoffing at a Silver medalist to me.
In fact, Python is better at what it does than almost anything from the era. That’s why we use it. Nothing is ever going to be perfect because better solutions are emergent, and reverse time travel does not exist.
Okay that's where you lost me. You can talk about the language from your perspective, and like I said we can agree to disagree, but it crosses a line when you want to comment on others' experience and put them down just because they disagree with you. No on is scoffing here, what I wrote was a considered criticism. Have a nice day.
I'm with Conal Elliot when he said on Type Theory for All that it is sooo much harder to understand a program in Python.
I have used all three languages; and you clearly have no idea of the notion of usability of a language. So many things contradict this, let me list them off the top of my head
- Getting a running toolchain working: Prexisting (most OSes bundle a Python interpreter) or a package install away for Python. Scheme / Racket is some odd mix of custom IDEs with Dr. in the name, or someone's 20 page essay on how SLIME is the best thing ever. Haskell gets into odd stuff with ghci, cabal, and stack, and all of them are extremely slow.
- Tutorials: Python has a ton of them, they all get you printing to stdout and calculating things in about 10 minutes. Scheme / Racket typically spends multiple chapters navel-gazing about lists, cons, and such. Haskell is actually better in terms of the Hello World stuff, but ghci v/s ghc bites you again; and no one has a clear idea of which one to use.
- Advanced concepts: Python has mainstream but halfhearted OOP; and things like decorators and metaprogramming. Quickly intelligible if you learned something else like Java or C++. Or if you learned shell scripts you can get quite a bit done with just imperative. Racket/Scheme: 3 chapters in and you're still trying to figure out tail recursion. Haskell: Instead of just doing fun things with take and foldl you're being hit with trivia about typeclasses.
You’re replying to a post making assertions about beginners.
That doesn’t usually mean people with 4 years programming experience picking up a new language.
Racket: criticizing for having a beginner-friendly IDE doesn’t make a lot of sense. There’s always Magic Racket for VSCode for the others.
I guess you’re not starting people with “How to Design Programs” because that’s pictures and animations for ages.
Haskell: that was funny but an absurd criticism ghc vs ghci?? Nobody has that problem. The other stuff - valid but lead with it instead of trolling.
Back when I was a beginner actually interested in getting out of the beginner step of Haskell, this was an issue for me times. So there's at least one person :)
Also anecdotally I have seen people ask this in Freenode #haskell as well (the “Freenode” probably tells you how long ago this was) ; and there a few issues [1] and [2] where I see beginners having the same/similar issue. The second one is particularly funny, 4 people give 5 solutions and no one seems to know what the actual fix is. Instead you have people arguing whether a repeated do works or not. This would never happen with Python, just saying :)
> Racket: criticizing for having a beginner-friendly IDE doesn’t make a lot of sense.
Sorry, perhaps too harsh but I don't think it's as beginner-friendly as you think. I think it would probably help if they made the design more modern and welcoming. All these details about you can rewrite entire languages in Lisp and we can't even at least get a GUI that looks like it was written after 2007?
---------
[1] https://www.reddit.com/r/haskell/comments/1kfym5s/difference...
[2] https://www.reddit.com/r/haskell/comments/18yj7i5/i_have_jus...
But to _learn_ programming, I really, really don't see how using Haskell would be simpler than Python. Perhaps if you have a specific background (e.g., math), but else python is almost pseudo code already. You'll really have to convince me that a more abstract language is better...
Python: errors based on incorrect indentation (many beginners don’t use nice IDEs), or don’t understand the meaning of the hints) and scope (don’t forget your “global” if you’re hacking in PyGame) are challenges.
Starting with 3.13, even the REPL is a sufficiently nice IDE to avoid any careless indentation errors from poor formatting (as opposed to ones caused by actually not understanding how many levels of indentation the current line of code should have).
Consider appending two strings in Scheme:
Appending two strings in Scheme: Since the type is present in the function name, it would be redundant to include it in the variable name.Can’t take credit for the quote, read it somewhere.
The whole language changed when they kicked what’s-his-name out, and it’s a tool I almost never reach for anymore, whereas 15 years ago it was my Swiss Army knife.
I still really enjoy using python though. It's not really a fair comparison because I hadn't used PHP and Perl for as long but I just don't hit some mystifying issue every single session like I did with those languages when I'm using python. I honestly have never even read about that __debug__ constant. It's fun to hear about it but it's just not something that's comes up much.
I’ve of course certainly heard of, seen, and used `assert`, but more often than not, outside of pytest, I see its use way more in potential footgun scenarios—I doubt that many people know that assertions can be silenced, and that they’d probably be better off raising exceptions in many cases where they’re using `assert`.
I would not be surprised in the least if that pattern existed in the wild. In fact, it's quite common to see this in C/C++ codebases too: people will use assert() to check a security-relevant property, and then disable those checks in their release builds "because it can't happen".
The most direct way to find out about `__debug__` is to read `python -h` (or the usage message, which is not all that easy to trigger) in full, and then head over to the documentation.
> Does it actually get used that often in real world code?
https://github.com/search?q=language%3APython+%2F%28%3F-i%29...
Assigning to __debug__ wouldn't do anything to the compiler as it never actually reads the variable, so assignment would just cause weirdness from other use
https://github.com/nucypher/constantSorrow/blob/master/tests...
True = 1
False = 0
then later these got added to the language. In Python 2 you could still reassign them.
True, False = False, True
Python 3 you could no longer reassign them.
Common language design boners:
- Not building in strings. That's now in the past. Everybody has strings. (Well, C...)
- Not building in multidimensional arrays of the numeric types. Everything that number-crunches needs them, and having multiple definitions is Not Fun and may lead to expensive re-copying between different libraries. This is an enormous blind spot in language design. It's one of the reasons FORTRAN, which has good multidimensional numeric arrays, is still often used for number-crunching.
- Not standardizing the small vectors (vec2, vec3, vec4) and their matrix friends. Graphics code depends on these, and it's really annoying if there are multiple slightly incompatible implementations. Especially since GPUs have hardware for those types, and you want CPU and GPU to use the same representations.
- Not having arrays of bits. Pascal had PACKED ARRAY[0..N] of BOOLEAN but that was lost in later languages. It's useful to have that as a language construct, because most modern CPUs have good hardware for dealing with bit strings, and you'd like the compiler to use it.
Most useful languages acquire these features, but, when they come in late, there are multiple similar implementations, and libraries made incompatible by depending on different implementations.
(Amusingly, when Second Life switched from Linden Scripting Language to Luau, they initially had True, TRUE, and true all in use, as different types with different semantics. I was able to persuade the devs to unify the boolean types.)
I'm not sure exactly which features are responsible (I'm inclined to blame templates), but C++'s std::vector<bool> is a rough edge. For those unfamiliar, the standard specifies this vector template in a way that's not compatible with other vectors.
It kind of started with vectors as the very first feature (I was sick & tired of libraries reinventing their own `Point`/`VectorN` in incompatible ways).
Definitely agree on multidimensional arrays. I feel like efficient arrays in general are underrated in high-level language design.
The thing you want is what Rust delivers in the box, &str a string slice reference type, in Rust's case the "string" is UTF-8 encoded text. On the bare metal the way to represent this type is as a "fat pointer" typically a pair of registers, one with the address of the first byte of the string and the other with a length.
C should have fat pointers, they were proposed, for IIRC C89 but the proposal was rejected. That's pretty sad, the fat pointer is expensive to the point of maybe feeling extravagant on a PDP-11, but by 1989 that's long gone.
More ridiculously C++ didn't get this type (which it eventually called std::string_view and provides in its standard library not as a built-in) until 2017, years after Rust 1.0 shipped. In the meanwhile C++ just did not have a sensible way to do this, strings are hard apparently.
The string buffer feature, allowing you to actually make strings is less important, as you say it will need an allocator and so on very bare metal you might not have this - but the string slice reference doesn't need an allocator.
I think it's worth delivering the basic "it's a growable array type, duh" implemenation of the string buffer type, which is what Rust's String type is, but C++ chooses to ship an oddly specific small-string optimized version as std::string right from the offset.
I'd consider LSL to have been foundational in my ultimate interest/career in software engineering. The strict typing, very usable compile/runtime errors, and good documentation/examples made it so easy to pick up as a teen. Not to mention as long as you didn't edit/save a script again it would always run the same regardless of updates.
Several times now†, people make a language where the way a for-each loop (for each Goose in Geese ...) works is that there's a single variable Goose and each time around the loop we change which value is referred to by the Goose variable. This seems intuitively like a reasonable way to do this. But it's wrong and eventually your programmers will get nasty surprises. What you actually should deliver is an implementation where each time around the loop there's a new variable named Goose, that variable goes away at the end of that iteration and will be replaced by the next one, with the same exact name.
† At least Go and C#, I think there are others
Is this because a closure inside a loop will capture a reference to `Goose`?
I think that this is a capture problem not a variable problem. The closure should always do the right thing and capture the value of all variables (not just ones inside the loop), instead of capturing the reference to the variables.
Then the general problem is fixed to match what developers expect, instead of a specific instance of that class of problems being fixed and working differently to how other captured variables work.
Now your "lalanthran closures" can't mutate the world because they work exclusively with copies not references, if they try to mutate something then whatever they're touching was just a copy not the real thing.
That is true. I still don't like the idea of "Here is a general rule. It applies everywhere but $HERE." Whether that general rule is "All captures are by value" or "All captures are by reference", the rule should not have exceptions based on context in the code.
A better tradeoff would be to have the general rule (whatever it is) apply everywhere, along with syntax for capturing (or not, depending what the default is). I'd rather have it grab everything by value, and for those things that are susceptible to race conditions (because more than one closure is modifying it), explicitly annotate it with a sigil (`&`, or a keyword, or similar).
I mean, in pseudocode, when I see:
I don't want to have to examine the surrounding scope to know whether or not `y` is susceptible to a race. I'd rather just see: An alternative viewpoint is that many languages have immutable variables and they seem to be getting along just fine without needing mutation on variables, shared or otherwise.But the rules didn't and still don't have any such exceptions.
> A better tradeoff would be to have the general rule (whatever it is) apply everywhere, along with syntax for capturing (or not, depending what the default is)
This "solution" is how it works in C++. We can thus castigate the programmer for writing the wrong runes in their captures list and never for a moment doubt that we got it right when we introduced so very many footguns...
Tony Hoare's observation applies "One way is to make the program so simple, there are obviously no errors. The other is to make it so complicated, there are no obvious errors."
> An alternative viewpoint is that many languages have immutable variables and they seem to be getting along just fine without needing mutation on variables, shared or otherwise.
Sure, and one of the astonishing things in C# or Go before they fixed this is that you can indeed have immutable variables which change, even though that's silly - the language can decide that you mustn't change Goose, but it doesn't need to obey its own rules because it will change it for each loop iteration.
Most trouble in this area came from the iteration variable outliving the loop. That's not good when the iteration variable is a pointer. In C, it often is, and at the end of the loop, it points to an invalid address. It was a change to C (when?) to make the iteration variable go out of scope before code after the loop could get at it.
for all the hate js used to get, py is at least a few magnitudes worse.
my opinion ofc. don’t get mad xD
Did you encounter JS first?
> I often work with data scientists and have to productionize their jupyter notebooks
I'm not a huge Python fan, despite working with it fulltime, but this feels like mixing correlation and causation. Data scientists would not be writing good, optimized code in any language.
For little utilities, it’s faster than a lot of alternatives - just start the interpreter, no compilation needed.
It’s all relative, but if you view it as replacing bash scripts for renaming files or running other tools, it’s 100x better.
If you've ever read through FORTRAN code from a mathematics department or MATLAB/C/C++ from (non-software) engineering disciplines, then you can kind of understand why productionizing a jupyter notebook is not the worst of all possible worlds.
My CLI tools publish from Github to PyPI so that I can run tools with just `uvx sql-agent-cli` or `uvx dlna-here. Nothing for me to handle downloading (directly myself), no environment to manually setup, portable (Linux, Windows, Mac, ARM, x86). Easy for agents to run from a skill.md file without any other prereq than uv.
Really useful library ecosystem to leverage. No more shell scripts, or TS/JS/PHP backend services. I've even used Python on devices I've built around Raspberry Pi Zero 2 boards.
And it's weird how you import files. They're dot-separated packages that resemble file structure but not exactly. NodeJS has a self-explanatory require("./foo.js") or "../foo.js". The newer JS `import` syntax is annoyingly different from `require` but not terrible.
https://packaging.python.org/en/latest/specifications/inline...
We open sourced the implementation https://github.com/janushendersonassetallocation/loman
That said, I find it the nicest, cleanest option of the three. I still wouldn't use it for large and complex projects. I really like it for stuff where one might otherwise use shellscript. It's way way better than shellscript... except if it's all about files and running external commands.
This is exactly what I remember being said about C (which I agree with) and often given as a reason why higher level languages like Python or Java have so many protections against things C/C++ allowed (memory management being the biggest one of course). Very funny, and I assume not coincidental, to read this about Python in the modern programming landscape.
I'm bringing Y2K back next!
In an age where people are still standing by C over memory-safe systems programming languages, I feel quite comfortable depending Python for the great many things that Python is good at.
The fact that Python has become the language of choice for machine learning and data science is not a language issue.
Ruby has such a nice holistic consistency to it. With a few exceptions, it feels like it was conceived of by one person with a core idea in mind. Python feels like a mess.
This is just my personal opinion with no data to back it up, but I suspect that Python "won" because it has excellent Windows support, while Ruby doesn't. Even a decade ago, Python's website offered an official native Windows installer [0], while Ruby's website [1] still points you to a third-party installer, which doesn't even have native support since it uses MSYS2 [2].
Most non-developers use Windows, so if you're choosing the first language to teach a large group of people, good Windows support is fairly important. Python being the "default" introductory language gave it a huge number of users, then I suspect that everything flowed down from there.
[0]: https://web.archive.org/web/20160824235759/https://www.pytho...
[1]: https://www.ruby-lang.org/en/downloads/
[2]: https://rubyinstaller.org/
I think you might be able to go an additional decade backwards. Back in college most of my friends were on windows and one of them was using python for class projects.
Yeah, Python has had good Windows support at least 15 [0] or 25 years [1], depending on how you count it.
> Back in college most of my friends were on windows and one of them was using python for class projects.
Well it's always been possible to install Ruby on Windows too, it's just that Python supports it so much better.
[0]: https://peps.python.org/pep-0397/
[1]: https://peps.python.org/pep-0277/
I also wonder how many people gave up on Python just because the installer doesn't put it in your PATH. You'd install Python then no python, wtf. Ok so https://discuss.python.org/t/python-command-not-found/22255 ... Then you fix it and it runs the wrong version of Python.
Ruby's most important error is that it does not support namespaces. This by itself makes it a far less scalable language than Python.
Likewise, it made classes incredibly easy to extend, which led to a monkey patching bonanza and far too much magic everywhere (Rails being by far the worst offender but not the only one). It meant having to keep too much stuff in your head and needing deep framework/library familiarity just to be able to understand basic code.
In the end, I feel like the incredible flexibility was Ruby's undoing, not just lack of library availability for a specific popular application. I was a Ruby zealot at one point but it began to lose its lustre not because it wasn't beautiful in theory, but because it was inconvenient in practice. In some ways, Python's restrictiveness became its greatest attribute. Then Python 3 helped to fix a lot of the inconsistency.
Both Python and Ruby have a consistent logic to them, just not consistent with each other. Just as all attributes are methods in Ruby, so all methods area attributes in Python, etc. Things get much easier in either language if you stop fighting their internal logic.
Have you tried using `uv`'s newer tools? They help a lot e.g. with linting speed, lock management, package dependency separation, correct python version mgmt and no need to fudge with venv.
At least it’s Python/Jupyter and not R, SAS, or MATLAB.
Sure, it's not a language to write a web browser or game engine in. And it is slow. But it has some very strong niches outside of ML/Data science. Personally, I love it. To each their own.
uv + PEP723 make this even better.
Could this be so that the interpreter don't inadvertently manipulate them or pass them to a function? param=None and param="" can be very different.
Took me down some rabbit holes, but interesting to see the chatter about the fix here:
https://github.com/python/cpython/issues/80233
Initially you could reassign True,False but that was verboten with the switch to python 3! The walrus operator was the one simply an oversight:
https://python-history.blogspot.com/2013/11/story-of-none-tr...
Explanation from Guido himself
*hordes
The only thing good I can say about Python nowadays is that it's easy to get started for the first five minutes, and then you'll have to deal with all of its weirdness: significant whitespace, truthiness, duck typing, GIL, distribution/packaging, etc, etc.
I was a big fan of Julia as the potential replacement for Python for science for such a long time and I had evangelized it a lot previously, but recently I've been more and more convinced that JIT/multiple dispatch was only good if you already know how to program well to begin with, which for a lot of academics who are not working in computer science, they write quite horrific code. I think it may be better off to skip Python altogether and write your code in a statically typed language to begin with.
In the past, the usual answer to people who need a programming language but did not want to learn programming was to give them a domain specific language that focused on solving the specific problem they wanted to solve.
> I think it may be better off to skip Python altogether and write your code in a statically typed language to begin with
Having a good REPL is a huge advantage for beginners (and expert users too), but I'm not aware of any (popular) statically-typed languages with a good REPL.
Despite Python's many faults, it's easy to install (especially on Windows), it has a large standard library, there are third-party packages available for essentially everything, it comes with a user-friendly REPL out-of-the-box, and it gives comprehensible error messages. I'm not really aware of any other (popular) languages with all these attributes.
scala's repl is decent. It has its annoyances, but so does python's (white space sensitivity + repl + terminal emulators stuck in the late mid century don't mix).
The idea is fine, I guess, although I certainly don't care for it. Where it gets most nasty is that whitespace that looks the same (in your editor) might not be equal and will cause you pain.
This is why idiomatic Python only uses spaces for indentation.
First off, what editor could I end up using in 2026 that makes this a realistic problem? Even the most basic editors I know of have options to convert tabs to spaces automatically (and any responsible teacher will tell the student to indent with spaces), and to continue the previous line's indentation automatically.
Second, modern Python is stricter about this, and also gives clear error messages.
By "weirdness" here you apparently mean not having to worry about matching up curly braces, and getting what you want automatically just for indenting your code the way you're supposed to indent it anyway... ?
> truthiness
Which is different from how it works in other similar languages, how exactly?
> duck typing
Which is weird, how exactly?
> GIL
You can go a lot further than five minutes in Python without having to worry about threads at all, and if you do attempt threading, unless you're writing C extensions, the worst thing the GIL does is deny you the multi-core processing you thought you were going to get.
> distribution/packaging
Tons better now, but it was honestly never difficult, people just didn't care.
They are contrasting in languages where true is true and 1 is 1; but true is not 1, and 1 is not true.
> Which is weird, how exactly?
Sometimes if it looks like a duck and quacks like a duck it can still not be a duck.
> You can go a lot further than five minutes in Python without having to worry about threads at all
Unless you specifically want to do multithreading.
> the worst thing the GIL does is deny you the multi-core processing you thought you were going to get.
That is the worst thing because I wanted that multi-core processing, that's the whole reason I wanted to do multithreading.
[0]: https://peps.python.org/pep-0790/#schedule
[1]: https://peps.python.org/pep-0781/#backwards-compatibility