Python really needs to take the Typescript approach of "all valid Python4 is valid Python3". And then add value types so we can have int64 etc. And allow object refs to be frozen after instantiation to avoid the indirection tax.
Sensible type-annotated python code could be so much faster if it didn't have to assume everything could change at any time. Most things don't change, and if they do they change on startup (e.g. ORM bindings).
To clarify, it is nuts that in an object method, there is a performance enhancement through caching a member value.
class SomeClass
def init(self)
self.x = 0
def SomeMethod(self)
q = self.x
## do stuff with q, because otherwise you're dereferencing self.x all the damn time
Any decent JIT compiler (and HotSpot's is world class) will optimize this out. Likely this was done very early on in development, or was just to reduce bytecode size to promote inlining heuristics that use it
If what you call gets inlined, then the compiler can see that it either does or doesn't modify the attribute and optimize it accordingly. Even virtual calls can often be inlined via, e.g., class hierarchy analysis and inline caches.
If these analyses don't apply and the callee could do anything, then of course the compiler can't keep the value hoisted. But a function call has to occur anyway, so the hoisted value will be pushed/popped from the stack and you might as well reload it from the object's field anyway, rather than waste a stack slot.
That was a niche optimization primarily targeting code at intepretor. Even the most basic optimizing compiler in HotSpot tiered compilation chain at that time (the client compiler or C1) would be able to optimize that into the register. Since String is such an important class, even small stuffs like this is done.
> it is nuts that in an object method, there is a performance enhancement through caching a member value
i don't understand what you think is nuts about this. it's an interpreted language and the word `self` is not special in any way (it's just convention - you can call the first param to a method anything you want). so there's no way for the interpreter/compiler/runtime to know you're accessing a field of the class itself (let alone that that field isn't a computed property or something like that).
lots of hottakes that people have (like this one) are rooted in just a fundamental misunderstanding of the language and programming languages in general <shrugs>.
What's nuts is that the language doesn't guarantee that successive references to the same member value within the same function body are stable. You can look it up once, go off and do something else, and look it up again and it's changed. It's dynamism taken to an unnecessary extreme. Nobody in the real world expects this behaviour. Making it just a bit less dynamic wouldn't change the fundamentals of the language but it would make it a lot more tractable.
And yes, it often is obvious to humans that’s not intended to happen, and almost never what happens, but proving that is often hard or even impossible.
> Nobody in the real world expects this behaviour.
For example, numbers and strings are immutable objects in Python. If self.x is a number and its numeric value is changed by a method call, self.x will be a different object after that. I'd dare say people expect this to work.
basically all object oriented languages work like that. You access a member; you call a method which changes that member; you expect that change is visible lower in the code, and there're no statically computable guarantees that particular member is not touched in the called method (which is potentially shadowed in a subclass). It's not dynamism, even c++ works the same, it's an inherent tax on OOP. All you can do is try to minimize cost of that additional dereference.
I'm not even touching threads here.
now, functional languages don't have this problem at all.
OOP has nothing to do with it. In your C++ example, foo(bar const&); is basically the same as bar.foo();. At the end of the day, whether passing it in as an argument or accessing this via the method call syntax it's just a pointer to a struct. Not to mention, a C++ compiler can, and often does, choose to put even references to member variables in registers and access them that way within the method call.
This is a Python specific problem caused by everything being boxed by default and the interpreter does not even know what's in the box until it dereferences it, which is a problem that extends to the "self" object. In contrast in C++ the compiler knows everything there's to know about the type of this which avoids the issue.
> This is a Python specific problem caused by everything being boxed
I would say it is part python being highly dynamic and part C++ being full of undefined behavior.
A c++ compiler will only optimize member access if it can prove that the member isn't overwritten in the same thread. Compatible pointers, opaque method calls, ... the list of reasons why that optimization can fail is near endless, C even added the restrict keyword because just having write access to two pointers of compatible types can force the compiler to reload values constantly. In python anything is a function call to some unknown code and any function could get access to any variable on the stack (manipulating python stack frames is fun).
Then there is the fun thing the C++ compiler gets up to with varibles that are modified by different threads, while(!done) turning into while(true) because you didn't tell the compiler that done needs to be threadsafe is always fun.
That's not true. I mean: it's true that it has little to do with OOP, but most imperative languages (only exception I know is Rust) have the issue, it's not "Python specific". For example (https://godbolt.org/z/aobz9q7Y9):
struct S {
const int x;
int f() const;
};
int S::f() const {
int a = x;
printf("hello\n");
int b = x;
return a-b;
}
The compiler can't reuse 'x' unless it's able to prove that it definitely couldn't have changed during the `printf()` call - and it's unable to prove it. The member is loaded twice. C++ compilers can usually only prove it for trivial code with completely inlined functions that doesn't mutate any external state, or mutates in a definitely-not-aliasing way (strict aliasing). (and the `const` don't do any difference here at all)
In Python the difference is that it can basically never prove it at all.
In Python attribute access aren't stable! `self.x` where `x` is a property is not guaranteed to refer to the same thing.
And getting rid of descriptors would be a _fundamental change to the language_. An immeense one. Loads of features are built off of descriptors or descriptor-like things.
And what you're complaining about is also not true in Javascript world either... I believe you can build descriptor-like things in JS now as well.
_But_ if you want that you can use stuff like mypyc + annotations to get that for you. There are tools that let you get to where you want. Just not out of the box because Python isn't that language.
Remember, this is a scripting language, not a compiled language. Every optimization for things you talk about would be paid on program load (you have pyc stuff but still..)
Gotta show up with proof that what you're saying is verifiable and works well. Up until ~6 or 7 years ago CPython had a concept of being easy to onboard onto. Dataflow analyses make the codebase harder to deal with.
Having said all of that.... would be nice to just inline RPython-y code and have it all work nicely. I don't need it on everything and proving safety is probably non-trivial but I feel like we've got to be closer to doing this than in the past.
I ... think in theory the JIT can solve for that too. In theory
> same member value within the same function body are stable
Did you miss the part where I explained to you there's no way to identify that it's a member variable?
> Nobody in the real world expects this behaviour
As has already been explained to you by a sibling comment you are in fact wrong and there are in fact plenty of people in the real world who do actually expect this behavior.
So I'll repeat myself: lots of hottakes from just pure. Unadulterated, possibly willful, ignorance.
The above is a very thick response that doesn't address the parent's points, just sweeps them under the rag with "that's just how it was designed/it works".
"Did you miss the part where I explained to you there's no way to identify that it's a member variable?"
No, you you did miss the case where that in itself can be considered nuts - or at least an unfortunate early decision.
"this just how things are dunn around diz here parts" is not an argument.
> the word `self` is not special in any way (it's just convention - you can call the first param to a method anything you want).
The name `self` is a convention, yes, but interestingly in python methods the first parameter is special beyond the standard "bound method" stuff. See for example PEP 367 (New Super) for how `super()` resolution works (TL;DR the super function is a special builtin that generates extra code referencing the first parameter and the lexically defining class)
That was how the Mojo language started. And then soon after the hype they said that being a superset of Python was no longer the goal. Probably because being a superset of Python is not a guarantee for performance either.
Maybe, but I quoted specific part I was replying to. TS has no impact on runtime performance of JS. Type hints in Python have no impact on runtime performance of Python (unless you try things like mypyc etc; actually, mypy provides `from mypy_extensions import i64`)
Therefore Python has no use for TS-like superset, because it already has facilities for static analysis with no bearing on runtime, which is what TS provides.
Performance is one part of the discussion, but cleanliness is another. A Python4 that actually used typing in the interpreter, had value types, had a comptime phase to allow most metaprogramming to work (like monkey patching for tests) would be great! It would be faster, cleaner, easier to reason about, and still retain the great syntax and flexibility of the language.
I too see potential in this - it started feeling a bit weird in recent years switching between Go, Python and Rust codebases with Python code looking more and more like a traditional statically typed language and not getting the performance benefits.
I know I know, there are libraries and frameworks which make heavy use of fun stuff you can do with strings (leading to the breakdown of even the latest and greatest IDE tooling and red squiggly lines all over you code) and don’t get me started on async etc.
Funnily enough I’ve found Python to be excellent for modelling my problem domain with Pydantic (so far basically unparalleled, open for suggestions in Go/Rust), while the language also gets out of my way when I get creative with list expressions and the like. So overall, still it is extremely productive for the work I’m doing, I just need to spin up more containers in prod.
Isn't rpython doing that, allowing changes on startup and then it's basically statically typed? Does it still exist? Was it ever production ready? I only once read a paper about it decades ago.
RPython is great, but it changes semantics in all sorts of ways. No sets for example. WTF? The native Set type is one of the best features of Python. Tuples also get mangled in RPython.
I think sadly a lot of Python in the wild relies heavily, somewhere, on the crazy unoptimisable stuff. For example pytest monkey patches everything everywhere all the time.
You could make this clean break and call it Python 4 but frankly I fear it won't be Python anymore.
As a person who has spent a lot of time with pytest, I'm ready for testing framework that doesn't do any of that non-obvious stuff.
Generally use unittest as much as I can these days, so much less _wierd_ about how it does things. Like jeeze pytest, do you _really_ need to stress test every obscure language feature? Your job is to call tests.
Allowing metaprogramming at module import (or another defined phase) would cover most monkey patching use cases. From __future__ import python4 would allow developers to declare their code optimisable.
Oh, and while we're at it, fix the "empty array is instantiated at parse time so all your functions with a default empty array argument share the same object" bullshit.
It has nothing to do with whether the list is empty. It has nothing to do with lists at all. It's the behaviour of default arguments.
It happens at the time that the function object is created, which is during runtime.
You only notice because lists are mutable. You should already prefer not to mutate parameters, and it especially doesn't make sense to mutate a parameter that has a default value because the point of mutating parameters is that the change can be seen by the caller, but a caller that uses a default value can't see the default value.
The behaviour can be used intentionally. (I would argue that it's overused intentionally; people use it to "bind" loop variables to lambdas when they should be using `functools.partial`.)
If you're getting got by this, you're fundamentally expecting Python to work in a way that Pythonistas consider not to make sense.
Execution time, not parse time. It's a side effect of function declarations being statements that are executed, not the list/dict itself. It would happen with any object.
It's still ridiculous. A hypothetical Python4 would treat function declarations as declarations not executable statements, with no impact on real world code except to remove all the boilerplate checks.
There is no such thing as a "function declaration" in Python. The keyword is "def", which is the first three letters of the word "define" (and not a prefix of "declare"), for a reason.
The entire point of it being an executable statement is to let you change things on the fly. This is key to how the REPL works. If I have `def foo(): ...` twice, the second one overwrites the first. There's no need to do any checks ahead of time, and it works the same way in the REPL as in a source file, without any special logic, for the exact same reason that `foo = 1` works when done twice. It's actually very elegant.
People who don't like these decisions have plenty of other options for languages they can use. Only Python is Python. Python should not become not-Python in order to satisfy people who don't like Python and don't understand what Python is trying to be.
You think so but then you write a function with a default argument pointing to some variable that is a list and now suddenly the semantics of that are... what?
you could just treat argument initialization as an executable expression which is called every time you call a function. If you have a=[], then it's a new [] every time. If a=MYLIST then it's a reference to the same MYLIST. Simple. And most sane languages do it this way, I really don't know why python has (and maintain) this quirk.
What realistic use case do you have for caring about whether two integers of the same value are distinct objects? Modern versions of Python warn about doing unpredicatble things with `is` exactly because you are not supposed to do those things. Valid use cases for `is` at all are rare.
There might not be that many of them, depending on how you count, but they're not rare in the slightest. For example, you have to use `is` in the common case where you want the default value of a function argument to be an empty list.
> Python really needs to take the Typescript approach of "all valid Python4 is valid Python3
Great idea, but I'm not convinced that they learned anything from the Python 2 to 3 transition, so I wouldn't hold my breath.
If you want a language system without contempt for backward compatibility, you're probably better off with Java/C++/JavaScript/etc. (though using JS libraries is like building on quicksand.) Bit of a shame since I want to like Python/Rust/Swift/other modern-ish languages, but it turns out that formal language specifications were actually a pretty good idea. API stability is another.
I'm been occasionally glancing at PR/issue tracker to keep up to date with things happening with the JIT, but I've never seen where the high level discussions were happening; the issues and PRs always jumped right to the gritty details. Is there anywhere a high-level introduction/example of how trace projection vs recording work and differ? Googling for the terms often returns CPython issue tracker as the first result, and repo's jit.md is relatively barebones and rarely updated :(
Similarly, I don't entirely understand refcount elimination; I've seen the codegen difference, but since the codegen happens at build time, does this mean each opcode is possibly split into two (or more?) stencils, with and without removed increfs/decrefs? With so many opcodes and their specialized variants, how many stencils are there now?
I think CPython already had tier2 and some tracing infrastructure when the copy-and-patch JIT backend was added; it's the "JIT frontend" that's more obscure to me.
UPDATE: I misunderstood the question :-/ You can ignore this.
I love playing with compilers for fun, so maybe I can shed some light. I’ll explain it in a simplified way for everyone’s benefit (going to ignore the stack):
When an object is passed between functions in Python, it doesn’t get copied. Instead, a reference to the object’s memory address is sent. This reference acts as a pointer to the object’s data. Think of it like a sticky note with the object’s memory address written on it. Now, imagine throwing away one sticky note every time a function that used a reference returns.
When an object has zero references, it can be freed from memory and reused. Ensuring the number of references, or the “reference count” is always accurate is therefore a big deal. It is often the source of memory leaks, but I wouldn’t attribute it to a speed up (only if it replaces GC, then yes).
Oh man, Python 2 > 3 was such a massive shift. Took almost half a decade if not more and yet it mainly changing superficial syntax stuff. They should have allowed ABIs to break and get these internal things done. Probably came up with a new, tighter API for integrating with other lower level languages so going forward Python internals can be changed more freely without breaking everything.
The text encoding stuff wasn't a small change considering what it could break, at least. And remember we're sometimes talking about software that would cost a lot of money to migrate or upgrade. I still maintain some 2.x python code-bases that will be very expensive to migrate and the customer is not willing to invest that money.
Although your general sentiment is something I agree with(if it's going to be painful do it and get it over with), I don't believe anybody knew or could've guessed what the reaction of the ecosystem would be.
Your last point about being able to change internals more freely is also great in theory but very difficult(if not impossible) to achieve in practice.
I don't know. Having maintained some small projects that were free and open source, I saw the hostility and entitlement that can come from that position. And those projects were a spec of dust next to something like Python. So I think the core team is doing the best they can. It was always going to be damned if you do, damned if you don't.
> I still maintain some 2.x python code-bases that will be very expensive to migrate and the customer is not willing to invest that money.
Slight tangent: if Claude can decimate IBM stock price by migrating off Cobol for cheap, surely we can do Python 2 to 3 now, too?
About the internals: we sort of missed an opportunity there, but back then there also didn't quite know what they were doing (or at least we have better ideas of what's useful today). And making the step from 2 to 3 even bigger might have been a bad idea?
I wasn't aware that migrating projects off Cobol has become cheap and it would only take a Claude subscription.
In my experience, the problem had always been maintaining the business logic and any integrations with third-party software that also may be running legacy code-bases or have been abandoned. It can get quite complicated, from what I've seen. Now of course if you're talking about well maintained code-bases with 100%, or close to 100% test coverage, and that includes the integration part along with having the ability to maintain the user experience and/or user interface then yes it becomes a relatively easy process of "just write the code". But, in my experience, this has never been the case.
For the 2.x code-bases I maintain, the customers simply doesn't want to pay for any of it. They might choose to at a later time, but so far it has been more cost effective for them to pay me to maintain that legacy code than pay to have it migrated. Other customers have different needs and thus budget differently.
I'll refrain from judging if 2 to 3 was a missed opportunity or not. I believe the core team does actually know what they're doing and that any decision would've been criticized.
I skip news like that. It's an AI business hyping one of their tools in a major AI hype-cycle. Shares can go up and down based on sentiment. My point still stands.
To me, there's a big difference between saying that migration projects can now be assisted with some AI tooling and saying that it is cheap and to just get Claude to do it.
Maybe I am out of touch but the former is realistic and the latter is just magical hand-waving.
> I believe the core team does actually know what they're doing and that any decision would've been criticized.
I agree with the latter. About the former: they probably made a good decisions given the information available at the time. I mean that nowadays they know more than they did in the past.
Absoultely, I had a 2 -> 3 code base I'd mostly given up on, and Claude was amazing. It even re-wrote some libraries I used without py3 versions, decided to just write the parts of the libraries I needed.
It does much better with good tests. In my case the output was a statically generated website, so I could just say 'make the same website, given these inputs'.
I cannot believe people are still acting like Python 2->3 was a huge fuck-up and an enormous missed opportunity. When in reality Python is by most measures the most popular language and became so AFTER that switch.
Since the switch we have seen enormous companies being built from scratch. There is no reason for anyone to be complaining about it being too hard to upgrade in 2026
Living through it... Python 3 made a lot of changes for the better but 3.0 in particular included a bunch of unforced errors that made it too hard for people to upgrade in one go.
It wasn't until much later (I would say 3.4 or 3.5?) that we had good tooling to allow for migrating from Python 2 to Python 3 gradually, which is what most tools needed to do.
The final thing that made Python upgrading easy was making a bunch of changes (along with stuff like six) so that you could write code that would run identically in Python 2 and Python 3. That lets you do refactors over time, little cleanups, and not have the huge "move to Python 3" commit.
> Python is by most measures the most popular language and became so AFTER that switch
The switch had nothing to do with Python's rise in popularity though, it was because of NumPy and later PyTorch being adopted by data scientist and later machine learning tasks that themselves became very popular. Python's popularity rose alongside those.
> There is no reason for anyone to be complaining about it being too hard to upgrade in 2026
The "complaints" are about unnecessary and pointless breakage, that was very difficult for many codebases to upgrade for years. That by now most of these codebases have been either abandoned, upgraded or decided to stick with Python2 until the end of time doesn't mean these pains didn't happen nor that the language's developers inflicting them to their users were a good idea because some largely unrelated external factors made the language popular several years later.
> that was very difficult for many codebases to upgrade for years.
In case people have forgotten: python 3.3 through 3.5 (and 3.6 I think) each had to reintroduce something that was removed to make the upgrade easier. Jumping from 2.7 to 3.3 (or higher depending on what you needed) was the recommended route because of this, it was less work than going to 3.0, 3.1, or 3.2
It took a long time for python 3 to add the necessary backwards compatibility features to allow people to switch over. Once they did it was fine, but it was a massive fuck up until then. The migration took far longer than it should have done
Its widely regarded as a disaster for good reason, that forced some corrections in python to fix it. Just because its fine now, does not mean it was always fine
The Python devs didn’t want to make huge changes because they were worried Python 3 would end up taking forever like Perl 6. Instead they went to the other extreme and broke everyone’s code for trivial reasons and minimal benefit, which meant no-one wanted to upgrade.
Even the main driver for Python 3, the bytes-Unicode split, has unfortunately turned out to be sub-optimal. Python essentially bet on UTF-32 (with space-saving optimisations), while everyone else has chosen UTF-8.
> Python essentially bet on UTF-32 (with space-saving optimisations)
How so? Python3 strings are unicode and all the encoding/decoding functions default to utf-8. In practice this means all the python I write is utf-8 compatible unicode and I don't ever have to think about it.
UTF-32 allows for constant time character accesses, which means that mystr[i] isn't O(n). Most other languages can only provide constant time access for code units.
UTF-32 allows for constant time access to code points. Neither UTF-8 nor UTF-16 can do the same (there are 2 to the power of 20 valid code points, though not all are in use).
While most characters might be encodable as a single code point, Python does not normalize strings, so there is no guarantee that even relatively normal characters are actually stored as single code points.
> all the encoding/decoding functions default to utf-8
Languages that use UTF-8 natively don't need those functions at all. And the ones in Python aren't trivial - see, for example, `surrogateescape`.
As the sibling comment says, the only benefit of all this encoding/decoding is that it allows strings to support constant-time indexing of code points, which isn't something that's commonly needed.
Internally Python holds a string as an array of uint32. A utf-8 representation is created on demand from it (and cached). So pansa2 is basically correct [^1].
IMO, while this may not be optimal, it's far better than the more arcane choice made by other systems. For example, due to reasons only Microsoft can understand, Windows is stuck with UTF-16.
[1] Actually it's more intelligent. For example, Python automatically uses uint8 instead of uint32 for ASCII strings.
There is no caching of a "utf-8 representation". You may check for example:
>>> x = '日本語'*100000000
>>> import time
>>> t = time.time(); y = x.encode(); time.time() - t # takes nontrivial time
>>> t = time.time(); y = x.encode(); time.time() - t # not cached; not any faster
Generally, the only reason this would happen implicitly is for I/O; actual operations on the string operate directly on the internal representation.
Python uses either 8, 16 or 32 bits per character according to the maximum code point found in the string; uint8 is thus used for all strings representable in Latin-1, not just "ASCII". (It does have other optimizations for ASCII strings.)
The reason for Windows being stuck with UTF-16 is quite easy to understand: backwards compatibility. Those APIs were introduced before there supplementary Unicode planes, such that "UTF-16" could be equated with UCS-2; then the surrogate-pair logic was bolted on top of that. Basically the same thing that happened in Java.
> There is no caching of a "utf-8 representation".
No there certainly is. This is documented in the official API documentation:
UTF-8 representation is created on demand and cached in the Unicode object.
https://docs.python.org/3/c-api/unicode.html#unicode-objects
In particular, Python's Unicode object (PyUnicodeObject) contains a field named utf8. This field is populated when PyUnicode_AsUTF8AndSize() is first called and reused thereafter. You can check the exact code I'm talking about here:
> Python essentially bet on UTF-32 (with space-saving optimisations), while everyone else has chosen UTF-8.
It did nothing of the sort. UTF-8 is the default source file encoding and has been the target for many APIs. It likely would have been the default for all I/O stuff if we lived in a world where Windows had functioning Unicode in the terminal the whole time and didn't base all its internal APIs on UTF-16.
I assume you're referring to the internal representation of strings. Describing it as "UTF-32 with space-saving optimizations" is missing the point, and also a contradiction in terms. Yes, it is a system that uses the same number of bytes per character within a given string (and chooses that width according to the string contents). This makes random access possible. Doing anything else would have broken historical expectations about string slicing. There are good arguments that one shouldn't write code like that anyway, but it's hard to identify anything "sub-optimal" about the result except that strings like "I'm learning 日本語" use more memory than they might be able to get away with. (But there are other strings, like "ℍℯℓ℗", that can use a 2-byte width while the UTF-8 encoding would add 3 bytes per character.)
It's wrong. Python3 eliminated mountains of annoying bugs that happened all over the code base because of mixing of unicode strings and byte strings. Python2 was an absolute mess.
I'm curious is the JIT developers could mention any Python features that prevent promising JIT features. An earlier Ken Jin blog [1], mentions how __del__ complicates reference counting optimization.
There is a story that Python is harder to optimize than, say, Typescript, with Python flexibility and the C API getting mentioned. Maybe, if the list of troublesome Python features was out there, programmers could know to avoid those features with the promise of activating the JIT when it can prove the feature is not in use. This could provide a way out of the current Python hard-to-JIT trap. It's just a gist of an idea, but certainly an interesting first step would be to hear from the JIT people which Python features they find troublesome.
It's interesting you mention __del__ because Javascript not only doesn't have destructors but for security reasons (that are above my pay grade) but the spec _explicitly prohibits_ implementations from allowing visibility into garbage collection state, meaning that code cannot have any visibility into deallocations.
I think __del__ is tricky though. In theory __del__ is not meant to be reliable. In practice CPython reliably calls it cuz it reference counts. So people know about it and use it (though I've only really seen it used for best effort cleanup checks)
In a world where more people were using PyPy we could have pressure from that perspective to avoid leaning into it. And that would also generate more pressure to implement code that is performant in "any" system.
> In practice CPython reliably calls it cuz it reference counts ... In a world where more people were using PyPy we could have pressure from that perspective to avoid leaning into it
A big part of the problem is that much of the power of the Python ecosystem comes specifically from extensions/bindings written in languages with manual (C) or RAII/ref-counted (C++, Rust) memory management, and having predictable Python-level cleanup behavior can be pretty necessary to making cleanup behavior in bound C/C++/Rust objects work. Breaking this behavior or causing too much of a performance hit is basically a non-starter for a lot of Python users, even if doing so would improve the performance of "pure" Python programs.
That link itself calls out that conformant implementations can’t be relied on to call callbacks.
> A conforming JavaScript implementation, even one that does garbage collection, is not required to call cleanup callbacks. When and whether it does so is entirely down to the implementation of the JavaScript engine. When a registered object is reclaimed, any cleanup callbacks for it may be called then, or some time later, or not at all.
It's likely that major implementations will call cleanup callbacks at some point during execution, but those calls may be substantially after the related object was reclaimed. Furthermore, if there is an object registered in two registries, there is no guarantee that the two callbacks are called next to each other — one may be called and the other never called, or the other may be called much later.
There are also situations where even implementations that normally call cleanup callbacks are unlikely to call them:
It's supported in all of the major engines. And you also can't rely on the garbage collector to run at a predictable time (or at all!), so the engine never calling finalizers is functionally the same as the garbage collector being unusual.
The only (other) visible effect of GC not running is memory exhaustion. WeakRef/FinalizationGroup not getting triggered can have lots of script-visible effects, so can be much much worse. I wouldn't describe that as "functionally the same".
Oh! While this one does mention that you don't have visibility, this + weak refs seem to change the game
I remember a couple of years ago (well probably around 2021) reading about GC exposure concerns and seeing some line in some TC39 doc like "users should not have visibility into collection" but if we've shipped weakrefs sounds like we're not thinking about that anymore
We still try to limit any additional exposure as much as possible, and WR/FG are specced to keep the visibility as coarse as possible. (Collections won't be visible until the current script execution finishes, though async adds a lot more places where that can happen.)
A proposal to add new ways of observing garbage collection will still be shot down immediately without a damn good justification.
> meaning that code cannot have any visibility into deallocations.
This is more pedantry than a serious question. JavaScript has WeakReference, sure it'd be cumbersome and inefficient because you'd need to manually make and poll each thing you wanted to observe, but could it not be said that it does provide a view on deallocations?
Yes, WeakRef and FinalizationGroup both make GC visible (the latter removes the need to poll in your example). So not pedantic at all. They were eventually added after much reluctance from the language designers and implementers, partly because they can lead to code being broken by (valid & correct) engine optimizations, which is a big no-no on the web. But some things simply cannot be implemented without them.
Note that 90% of the uses for them actually shouldn't be using them, usually for subtle reasons. It's always a big cause for debate.
> However, I misunderstood and came up with an even more extreme version: instead of tracing versions of normal instructions, I had only one instruction responsible for tracing, and all instructions in the second table point to that. Yes I know this part is confusing, I’ll hopefully try to explain better one day. This turned out to be a really really good choice. I found that the initial dual table approach was so much slower due to a doubling of the size of the interpreter, causing huge compiled code bloat, and naturally a slowdown.
> By using only a single instruction and two tables, we only increase the interpreter by a size of 1 instruction, and also keep the base interpreter ultra fast. I affectionally call this mechanism dual dispatch.
I really do hope they'll write that better explanation one day because this sounds pretty intriguing all on its own.
Great to see this going, Python also deserves a JIT, and given that only few bother with PyPy or GraalPy, shipping into the CPYthon is the only way to have less "rewrite into XYZ".
The guy said he hopes the free-threaded build'll be the only one in "3.16 or 3.17", I wonder if that should apply to the JIT too or how the JIT and interpreter interact.
I continue to believe that free-threading hurts performance more than it helps and Python should abandon it.
Having to have thread safe code all over the place just for the 1% of users who need to have multi-threading in Python and can't use subinterpreters for some reason is nuts.
> Having to have thread safe code all over the place just for the 1% of users who need to have multi-threading in Python and can't use subinterpreters for some reason is nuts.
Way more than 1% of the community, particularly of the community actively developing Python, wants free-threaded. The problem here is that the Python community consists of several different groups:
1. Basically pure Python code with no threading
2. Basically pure Python with appropriate thread safety
3. Basically pure Python code with already broken threaded code, just getting lucky for now
4. Mixed Python and C/C++/Rust code, with appropriate threading behavior in the C or C++ components
5. Mixed Python and C or C++ code, with C and C++ components depending on GIL behavior
Group 1 gets a slightly reduced performance. Groups 2 and 4 get a major win with free-threaded Python, being able to use threading through their interfaces to C/C++/Rust components. Group 3 is already writing buggy code and will probably see worse consequences from their existing bugs. Group 5 will have to either avoid threading in their Python code or rewrite their C/C++ components.
Right now, a big portion of the Python language developer base consists of Groups 2 and 4. Group 5 is basically perceived as holding Python-the-language and Python-the-implementations back.
Where is the major win? Sorry but I just don't see the use case for free-threading.
Native code can already be multi-threaded so if you are using Python to drive parallelized native code, there's no win there. If your Python code is the bottleneck, well then you could have subinterpreters with shared buffers and locks. If you really need to have shared objects, do you actually need to mutate them from multiple interpreters? If not, what about exploring language support for frozen objects or proxies?
The only thing that free threading gives you is concurrent mutations to Python objects, which is like, whatever. In all my years of writing Python I have never once found myself thinking "I wish I could mutate the same object from two different threads".
PHP does this as well. Most distributions ship PHP without thread safety, but it's seeing more use now that FrankenPHP uses it. Speaking of which, it would be nice if PHP's JIT got a little love: it's never eked out more than marginal gains in heavily-numeric code.
Pure Python code always needed mutexes for thread safety with or without ol' GIL. I thought the difficulty with removing the GIL instead had to do with C extensions that rely on it.
I don't want to go too heavy on the negatives, but what's nuts is Python going for trust-the-programmer style multithreading. The risk is that extension modules could cause a lot of crashes.
My understanding is that many extension modules are already written to take advantage of multithreading by releasing the GIL when calling into C code. This allows true concurrency in the extension, and also invites all the hazards of multithreading. I wonder how many bugs will be uncovered in such extensions by the free threaded builds, but it seems like the “nuts” choice actually happened a long time ago.
As far as I know, PyPy doesn't support all CPython extensions, so pure Python code will probably (very likely) run fine but for other things most bets are off. I believe PyPy also only supports up to 3.11?
A lot of Python code still leans on CPython internals, C extensions, debuggers, or odd platform behavior, so PyPy works until some dependency or tool turns that gap into a support problem.
The JIT helps on hot loops, but for mixed workloads the warmup cost and compatibility tax are enough to keep most teams on the interpreter their deps target first.
Why shouldn't the reference implementation get JIT? Just because some other implementations already have it is no reason not to. That'd be like skipping list comprehensions because they already exist in CPython.
PyPy is limited to maintenance mode due to a lack of funding/contributors. In the past, I think a few contributors or funding is what helped push "minor" PyPy versions. It's too bad PyPy couldn't take the federal funding the PSF threw away.
Because the same people who made a big deal about supporting PyPy and PEP 399 when it was fashionable to do so are now told by their corporations that PyPy does not matter. CPython only moves with what is currently fashionable, employer mandated and profitable.
It is exactly what I'm referring to. I didn't say there aren't still people around. But they're far enough behind CPython that folks like NumPy are dropping support. Unless they get a substantial injection of new people and new energy, they're likely to continue falling behind.
Thanks for all the amazing work! I have Noob question. Wouldn't this get the funding back? Or would that not be preferable way to continue(as opposed to just volunteer driven)?
Like this is a big deal to get a project to a state where volunteers are spun up and actively breaking tasks and getting work done, no? It's a python JIT something I know next to nothing about — as do most application developers — which tells one how difficult this must have been.
The funding was Microsoft employing most of the team. They were laid off (or at least, moved onto different projects), apparently because they weren't working on AI.
With Python being the main language for AI, isn't like more important to be more performant?
I kinda don't get Microsoft reasoning, maybe they're just tight in money
Python is pretty big as glue in the AI ecosystem as far as I can tell. It also seems to be most agent's 'preferred' language to write code in, when you don't specify anything.
(The latter is probably more to do with the preferences they give it in the re-inforcement learning phase than anything technical, though.)
I always wanted this for Python but now that machines write code instead of humans I feel like languages like Python will not be needed as much anymore. They're made for humans, not machines. If a machine is going to do the dirty work I want it to produce something lean, fast, and strictly verified.
We got daguerrotypes, and then photographic film, and then digital cameras, along with image editing software, and now AI image generation systems; yet there are still people who go out and apply oil paints to a canvas with natural hair brushes. I'm not willing to lose that.
Pretty much my thoughts the other day... now that Codex does the writing, maybe I can finally switch to Go for the web backend stuff without being annoyed by some of its archaisms and gain significant execution performance, while still having a relatively easy to read language.
You ask a machine to write your code and you still care about being easy to read?
In my experience the people who care the most about code readability tend to be the people most opinionated on having the right abstractions, which are historically not available in Go.
Nah all the `if err != nil` is just so much noise they obscures the real logic. And for the longest time it didn’t have generics to write map/filter/reduce on slices, forcing people to use loops where the intention is less clear.
Ideally, the errors shouldn't be returned as-is, but wrapped with context instead. If that context doesn't matter for you, you can have your editor wrap the if instead, which helps a lot.
What is wrong with the Python code base that makes this so much harder to implement than seemingly all other code bases? Ruby, PHP, JS. They all seemed to add JITs in significantly less time. A Python JIT has been asked for for like 2 decades at this point.
I can't really talk about Ruby. But PHP is much more static and surface of things you have to care about at runtime is like magnitude smaller and there already was opache as a starting point.
And speaking of something like JIT in V8 is of the most sophisticated and complicated ever built. There hasn't been near enough man hours and funding to cpython to make it fair comparison
The Python C api leaks its guts. Too much of the internal representation was made available for extensions and now basically any change would be guaranteed to break backwards compatibility with something.
I would argue that the libraries, and specifically NumPy, are the reason Python is still in the picture today.
It will be interesting to see, moving forward, what languages survive. A 15% perf increase seems nice, until you realize that you get a 10x increase porting to Rust (and the AI does it for you).
Maybe library use/popularity is somewhat related to backwards compatibility.
Python it's a language that really good libraries for different domains. like web: django/flask AI numpy pytorch and more. All the ecosystem for scripting and being already installed in most linux distros and on macs. For GUI it has really good bindings for the major frameworks QT,GTK.
Python’s backward compatibility story still isn’t great compared to things like the Go 1.x compatibility promise, and languages with formal specs like JS and C.
The Python devs still make breaking changes, they’ve just learned not to update the major version number when they do so.
Indeed, Python's version format is semver but it's just aesthetics, they remove stuff in most (every?) minor version. Just yesterday I wasted hours trying to figure out a bug before realizing my colleague hadn't read the patch notes.
Python does not take backwards compatibility seriously. 2 to 3 is a big compatibility break. But things like `map(None, seq1, seq2)` also broke; such deliberate compatibility break is motivated by no more than aesthetic purity.
Python does not take backwards compatibility very seriously at all. Take a look at all the deprecated APIs.
I would say it's probably worth it to clean up all the junk that Python has accumulated... But it's definitely not very high up the list of languages in terms of backwards compatibility. In fact I'm struggling to think of other languages that are worse. Typescript probably? Certainly Go, C++ and Rust are significantly better.
For what it’s worth Ruby’s JIT took several different implementations, definitely struggled with Rails compatibility and literally used some people’s PhD research. It wasn’t a trivial affair
Some languages are much harder to compile well to machine code. Some big factors (for any languages) are things like: lack of static types and high "type uncertainty", other dynamic language features, established inefficient extension interfaces that have to be maintained, unusual threading models...
That makes sense if you're comparing with Java or C#, but not Ruby, which is way more dynamic than Python.
The more likely reason is that there simply hasn't been that big a push for it. Ruby was dog slow before the JIT and Rails was very popular, so there was a lot of demand and room for improvement. PHP was the primary language used by Facebook for a long time, and they had deep pockets. JS powers the web, so there's a huge incentive for companies like Google to make it faster. Python never really had that same level of investment, at least from a performance standpoint.
To your point, though, the C API has made certain types of optimizations extremely difficult, as the PyPy team has figured out.
Google, Dropbox, and Microsoft from what I can recall all tried to make Python fast so I don’t buy the “hasn’t seen a huge amount of investment”. For a long time Guido was opposed to any changes and that ossified the ecosystem.
But the main problem was actually that pypy was never adopted as “the JIT” mechanism. That would have made a huge difference a long time ago and made sure they evolved in lock step.
Microsoft is the one the TFA refers to cryptically when it says "the Faster CPython team lost its main sponsor in 2025".
AFAIK it was not driven by anything on the tech side. It was simply unlucky timing, the project getting in the middle of Microsoft's heavy handed push to cut everything. So much so that the people who were hired by MS to work on this found out they were laid off in a middle of a conference where they were giving talks on it.
> Python never really had that same level of investment, at least from a performance standpoint.
Or lack of incentive?
Alot of big python projects that does machine learning and data processing offloads the heavy data processing from pure python code to libraries like numpy and pandas that take advantage of C api binding to do native execution.
The simplest JIT just generates the machine code instructions that the interpreter loop would execute anyway. It’s not an extremely difficult thing, but it also doesn’t give you much benefit.
A worthwhile JIT is a fully optimizing compiler, and that is the hard part. Language semantics are much less important - dynamic languages aren’t particularly harder here, but the performance roof is obviously just much lower.
I think that it's just that python people took the problem different, they made working with c and other languages better, and just made bindings for python and offloaded the performant code to these libraries. Ex: numpy
For better or for worse they have been very consistent throughout the years that they don't want want to degrade existing performance. It is why the GIL existed for so long
That's a completely separate codebase that purposefully breaks backwards compatibility in specific areas to achieve their goals. That's not the same as having a first-class JIT in CPython, the actual Python implementation that ~everyone uses.
Yes, the graphs are incomprehensible because those are not defined in the article. They turn out to be different physical machines with different architectures: https://doesjitgobrrr.com/about
So the biggest gains so far are on Windows 11 Pro of (x86_64) ~20%? Is that because Windows was bad as a baseline (promethius)? It doesn't seem like the x86_64/Linux has improved as dramatically ~5% (ripley). I'm just surprised OS has that much of an effect that can be attributed to JIT vs other OS issues.
It's hard to say whether it's Windows related since the two x86_64 machines don't just run different OSes, they also have different processors, from different manufacturers. I don't know whether an AMD Ryzen 5 3600X versus Intel i5-8400 have dramatically different features, but unlike a generic static binary for x86_64, a JIT could in principle exploit features specific to a given manufacturer.
The immediate question has been answered, but what about the names? The latter three are obvious references to the Alien universe, but what relationship does blueberry have to them?
They are all JIT on different architectures, measured relative to CPython. https://doesjitgobrrr.com/about: blueberry is aarch64 Raspberry Pi, ripley is x86_64 Intel, jones is aarch64 M3 Pro, prometheus is x86_64 AMD.
I am trying to push back. I don't care if other people think the tools make them faster, I did not sign up to be a guinea pig for my employer or their AI-corp partner.
Sensible type-annotated python code could be so much faster if it didn't have to assume everything could change at any time. Most things don't change, and if they do they change on startup (e.g. ORM bindings).
If these analyses don't apply and the callee could do anything, then of course the compiler can't keep the value hoisted. But a function call has to occur anyway, so the hoisted value will be pushed/popped from the stack and you might as well reload it from the object's field anyway, rather than waste a stack slot.
i don't understand what you think is nuts about this. it's an interpreted language and the word `self` is not special in any way (it's just convention - you can call the first param to a method anything you want). so there's no way for the interpreter/compiler/runtime to know you're accessing a field of the class itself (let alone that that field isn't a computed property or something like that).
lots of hottakes that people have (like this one) are rooted in just a fundamental misunderstanding of the language and programming languages in general <shrugs>.
The language supports multiple threads and doesn’t have private fields (https://docs.python.org/3/tutorial/classes.html#private-vari...), so the runtime cannot rule out that the value gets changed in-between.
And yes, it often is obvious to humans that’s not intended to happen, and almost never what happens, but proving that is often hard or even impossible.
For example, numbers and strings are immutable objects in Python. If self.x is a number and its numeric value is changed by a method call, self.x will be a different object after that. I'd dare say people expect this to work.
now, functional languages don't have this problem at all.
This is a Python specific problem caused by everything being boxed by default and the interpreter does not even know what's in the box until it dereferences it, which is a problem that extends to the "self" object. In contrast in C++ the compiler knows everything there's to know about the type of this which avoids the issue.
I would say it is part python being highly dynamic and part C++ being full of undefined behavior.
A c++ compiler will only optimize member access if it can prove that the member isn't overwritten in the same thread. Compatible pointers, opaque method calls, ... the list of reasons why that optimization can fail is near endless, C even added the restrict keyword because just having write access to two pointers of compatible types can force the compiler to reload values constantly. In python anything is a function call to some unknown code and any function could get access to any variable on the stack (manipulating python stack frames is fun).
Then there is the fun thing the C++ compiler gets up to with varibles that are modified by different threads, while(!done) turning into while(true) because you didn't tell the compiler that done needs to be threadsafe is always fun.
struct S { const int x; int f() const; }; int S::f() const { int a = x; printf("hello\n"); int b = x; return a-b; }
The compiler can't reuse 'x' unless it's able to prove that it definitely couldn't have changed during the `printf()` call - and it's unable to prove it. The member is loaded twice. C++ compilers can usually only prove it for trivial code with completely inlined functions that doesn't mutate any external state, or mutates in a definitely-not-aliasing way (strict aliasing). (and the `const` don't do any difference here at all)
In Python the difference is that it can basically never prove it at all.
And getting rid of descriptors would be a _fundamental change to the language_. An immeense one. Loads of features are built off of descriptors or descriptor-like things.
And what you're complaining about is also not true in Javascript world either... I believe you can build descriptor-like things in JS now as well.
_But_ if you want that you can use stuff like mypyc + annotations to get that for you. There are tools that let you get to where you want. Just not out of the box because Python isn't that language.
Remember, this is a scripting language, not a compiled language. Every optimization for things you talk about would be paid on program load (you have pyc stuff but still..)
Gotta show up with proof that what you're saying is verifiable and works well. Up until ~6 or 7 years ago CPython had a concept of being easy to onboard onto. Dataflow analyses make the codebase harder to deal with.
Having said all of that.... would be nice to just inline RPython-y code and have it all work nicely. I don't need it on everything and proving safety is probably non-trivial but I feel like we've got to be closer to doing this than in the past.
I ... think in theory the JIT can solve for that too. In theory
Did you miss the part where I explained to you there's no way to identify that it's a member variable?
> Nobody in the real world expects this behaviour
As has already been explained to you by a sibling comment you are in fact wrong and there are in fact plenty of people in the real world who do actually expect this behavior.
So I'll repeat myself: lots of hottakes from just pure. Unadulterated, possibly willful, ignorance.
"Did you miss the part where I explained to you there's no way to identify that it's a member variable?"
No, you you did miss the case where that in itself can be considered nuts - or at least an unfortunate early decision.
"this just how things are dunn around diz here parts" is not an argument.
The name `self` is a convention, yes, but interestingly in python methods the first parameter is special beyond the standard "bound method" stuff. See for example PEP 367 (New Super) for how `super()` resolution works (TL;DR the super function is a special builtin that generates extra code referencing the first parameter and the lexically defining class)
It is called type hints, and is already there. TS typing doesn't bring any perf benefits over plain JS.
* a user can define subtype `class my_int(int)`, so you cannot optimize the layout of `class Foo`
* the builtin `int` and `float` are big-int like numbers, so operations on them are branchy and allocating.
and the fact that Foo is mutable and that `id(foo.a)` has to produce something complicates things further.
Therefore Python has no use for TS-like superset, because it already has facilities for static analysis with no bearing on runtime, which is what TS provides.
1) Add TS like language on top of Python in backwards compatible way
2) Introduce frozen/final runtime types
3) Use 1 and 2 to drive runtime optimizations
Funnily enough I’ve found Python to be excellent for modelling my problem domain with Pydantic (so far basically unparalleled, open for suggestions in Go/Rust), while the language also gets out of my way when I get creative with list expressions and the like. So overall, still it is extremely productive for the work I’m doing, I just need to spin up more containers in prod.
You could make this clean break and call it Python 4 but frankly I fear it won't be Python anymore.
It has nothing to do with whether the list is empty. It has nothing to do with lists at all. It's the behaviour of default arguments.
It happens at the time that the function object is created, which is during runtime.
You only notice because lists are mutable. You should already prefer not to mutate parameters, and it especially doesn't make sense to mutate a parameter that has a default value because the point of mutating parameters is that the change can be seen by the caller, but a caller that uses a default value can't see the default value.
The behaviour can be used intentionally. (I would argue that it's overused intentionally; people use it to "bind" loop variables to lambdas when they should be using `functools.partial`.)
If you're getting got by this, you're fundamentally expecting Python to work in a way that Pythonistas consider not to make sense.
It's just slightly annoying having to work around this by defaulting to None.
The entire point of it being an executable statement is to let you change things on the fly. This is key to how the REPL works. If I have `def foo(): ...` twice, the second one overwrites the first. There's no need to do any checks ahead of time, and it works the same way in the REPL as in a source file, without any special logic, for the exact same reason that `foo = 1` works when done twice. It's actually very elegant.
People who don't like these decisions have plenty of other options for languages they can use. Only Python is Python. Python should not become not-Python in order to satisfy people who don't like Python and don't understand what Python is trying to be.
There might not be that many of them, depending on how you count, but they're not rare in the slightest. For example, you have to use `is` in the common case where you want the default value of a function argument to be an empty list.
Great idea, but I'm not convinced that they learned anything from the Python 2 to 3 transition, so I wouldn't hold my breath.
If you want a language system without contempt for backward compatibility, you're probably better off with Java/C++/JavaScript/etc. (though using JS libraries is like building on quicksand.) Bit of a shame since I want to like Python/Rust/Swift/other modern-ish languages, but it turns out that formal language specifications were actually a pretty good idea. API stability is another.
Similarly, I don't entirely understand refcount elimination; I've seen the codegen difference, but since the codegen happens at build time, does this mean each opcode is possibly split into two (or more?) stencils, with and without removed increfs/decrefs? With so many opcodes and their specialized variants, how many stencils are there now?
https://discuss.python.org/t/pep-744-jit-compilation/50756/8... here's one thing
I do think you can also just outright ask questions about it on the forums and you'll get some answers.
At the end of the day there's only so many people working on this though.
I love playing with compilers for fun, so maybe I can shed some light. I’ll explain it in a simplified way for everyone’s benefit (going to ignore the stack):
When an object is passed between functions in Python, it doesn’t get copied. Instead, a reference to the object’s memory address is sent. This reference acts as a pointer to the object’s data. Think of it like a sticky note with the object’s memory address written on it. Now, imagine throwing away one sticky note every time a function that used a reference returns.
When an object has zero references, it can be freed from memory and reused. Ensuring the number of references, or the “reference count” is always accurate is therefore a big deal. It is often the source of memory leaks, but I wouldn’t attribute it to a speed up (only if it replaces GC, then yes).
Although your general sentiment is something I agree with(if it's going to be painful do it and get it over with), I don't believe anybody knew or could've guessed what the reaction of the ecosystem would be.
Your last point about being able to change internals more freely is also great in theory but very difficult(if not impossible) to achieve in practice.
I don't know. Having maintained some small projects that were free and open source, I saw the hostility and entitlement that can come from that position. And those projects were a spec of dust next to something like Python. So I think the core team is doing the best they can. It was always going to be damned if you do, damned if you don't.
Slight tangent: if Claude can decimate IBM stock price by migrating off Cobol for cheap, surely we can do Python 2 to 3 now, too?
About the internals: we sort of missed an opportunity there, but back then there also didn't quite know what they were doing (or at least we have better ideas of what's useful today). And making the step from 2 to 3 even bigger might have been a bad idea?
In my experience, the problem had always been maintaining the business logic and any integrations with third-party software that also may be running legacy code-bases or have been abandoned. It can get quite complicated, from what I've seen. Now of course if you're talking about well maintained code-bases with 100%, or close to 100% test coverage, and that includes the integration part along with having the ability to maintain the user experience and/or user interface then yes it becomes a relatively easy process of "just write the code". But, in my experience, this has never been the case.
For the 2.x code-bases I maintain, the customers simply doesn't want to pay for any of it. They might choose to at a later time, but so far it has been more cost effective for them to pay me to maintain that legacy code than pay to have it migrated. Other customers have different needs and thus budget differently.
I'll refrain from judging if 2 to 3 was a missed opportunity or not. I believe the core team does actually know what they're doing and that any decision would've been criticized.
"IBM Sinks Most Since 2000 as Anthropic Touts Cobol Tool"
https://finance.yahoo.com/news/ibm-sinks-most-since-2000-210...
It may not be "cheap", but possibly cheaper than IBM's consulting.
To me, there's a big difference between saying that migration projects can now be assisted with some AI tooling and saying that it is cheap and to just get Claude to do it.
Maybe I am out of touch but the former is realistic and the latter is just magical hand-waving.
I agree with the latter. About the former: they probably made a good decisions given the information available at the time. I mean that nowadays they know more than they did in the past.
It does much better with good tests. In my case the output was a statically generated website, so I could just say 'make the same website, given these inputs'.
Since the switch we have seen enormous companies being built from scratch. There is no reason for anyone to be complaining about it being too hard to upgrade in 2026
It wasn't until much later (I would say 3.4 or 3.5?) that we had good tooling to allow for migrating from Python 2 to Python 3 gradually, which is what most tools needed to do.
The final thing that made Python upgrading easy was making a bunch of changes (along with stuff like six) so that you could write code that would run identically in Python 2 and Python 3. That lets you do refactors over time, little cleanups, and not have the huge "move to Python 3" commit.
The switch had nothing to do with Python's rise in popularity though, it was because of NumPy and later PyTorch being adopted by data scientist and later machine learning tasks that themselves became very popular. Python's popularity rose alongside those.
> There is no reason for anyone to be complaining about it being too hard to upgrade in 2026
The "complaints" are about unnecessary and pointless breakage, that was very difficult for many codebases to upgrade for years. That by now most of these codebases have been either abandoned, upgraded or decided to stick with Python2 until the end of time doesn't mean these pains didn't happen nor that the language's developers inflicting them to their users were a good idea because some largely unrelated external factors made the language popular several years later.
In case people have forgotten: python 3.3 through 3.5 (and 3.6 I think) each had to reintroduce something that was removed to make the upgrade easier. Jumping from 2.7 to 3.3 (or higher depending on what you needed) was the recommended route because of this, it was less work than going to 3.0, 3.1, or 3.2
Its widely regarded as a disaster for good reason, that forced some corrections in python to fix it. Just because its fine now, does not mean it was always fine
Even the main driver for Python 3, the bytes-Unicode split, has unfortunately turned out to be sub-optimal. Python essentially bet on UTF-32 (with space-saving optimisations), while everyone else has chosen UTF-8.
How so? Python3 strings are unicode and all the encoding/decoding functions default to utf-8. In practice this means all the python I write is utf-8 compatible unicode and I don't ever have to think about it.
While most characters might be encodable as a single code point, Python does not normalize strings, so there is no guarantee that even relatively normal characters are actually stored as single code points.
Try this in Python:
You will see:Languages that use UTF-8 natively don't need those functions at all. And the ones in Python aren't trivial - see, for example, `surrogateescape`.
As the sibling comment says, the only benefit of all this encoding/decoding is that it allows strings to support constant-time indexing of code points, which isn't something that's commonly needed.
IMO, while this may not be optimal, it's far better than the more arcane choice made by other systems. For example, due to reasons only Microsoft can understand, Windows is stuck with UTF-16.
[1] Actually it's more intelligent. For example, Python automatically uses uint8 instead of uint32 for ASCII strings.
Python uses either 8, 16 or 32 bits per character according to the maximum code point found in the string; uint8 is thus used for all strings representable in Latin-1, not just "ASCII". (It does have other optimizations for ASCII strings.)
The reason for Windows being stuck with UTF-16 is quite easy to understand: backwards compatibility. Those APIs were introduced before there supplementary Unicode planes, such that "UTF-16" could be equated with UCS-2; then the surrogate-pair logic was bolted on top of that. Basically the same thing that happened in Java.
No there certainly is. This is documented in the official API documentation:
In particular, Python's Unicode object (PyUnicodeObject) contains a field named utf8. This field is populated when PyUnicode_AsUTF8AndSize() is first called and reused thereafter. You can check the exact code I'm talking about here:https://github.com/python/cpython/blob/main/Objects/unicodeo...
Is it clear enough?
It did nothing of the sort. UTF-8 is the default source file encoding and has been the target for many APIs. It likely would have been the default for all I/O stuff if we lived in a world where Windows had functioning Unicode in the terminal the whole time and didn't base all its internal APIs on UTF-16.
I assume you're referring to the internal representation of strings. Describing it as "UTF-32 with space-saving optimizations" is missing the point, and also a contradiction in terms. Yes, it is a system that uses the same number of bytes per character within a given string (and chooses that width according to the string contents). This makes random access possible. Doing anything else would have broken historical expectations about string slicing. There are good arguments that one shouldn't write code like that anyway, but it's hard to identify anything "sub-optimal" about the result except that strings like "I'm learning 日本語" use more memory than they might be able to get away with. (But there are other strings, like "ℍℯℓ℗", that can use a 2-byte width while the UTF-8 encoding would add 3 bytes per character.)
There is a story that Python is harder to optimize than, say, Typescript, with Python flexibility and the C API getting mentioned. Maybe, if the list of troublesome Python features was out there, programmers could know to avoid those features with the promise of activating the JIT when it can prove the feature is not in use. This could provide a way out of the current Python hard-to-JIT trap. It's just a gist of an idea, but certainly an interesting first step would be to hear from the JIT people which Python features they find troublesome.
[1] https://fidget-spinner.github.io/posts/faster-jit-plan.html
I think __del__ is tricky though. In theory __del__ is not meant to be reliable. In practice CPython reliably calls it cuz it reference counts. So people know about it and use it (though I've only really seen it used for best effort cleanup checks)
In a world where more people were using PyPy we could have pressure from that perspective to avoid leaning into it. And that would also generate more pressure to implement code that is performant in "any" system.
A big part of the problem is that much of the power of the Python ecosystem comes specifically from extensions/bindings written in languages with manual (C) or RAII/ref-counted (C++, Rust) memory management, and having predictable Python-level cleanup behavior can be pretty necessary to making cleanup behavior in bound C/C++/Rust objects work. Breaking this behavior or causing too much of a performance hit is basically a non-starter for a lot of Python users, even if doing so would improve the performance of "pure" Python programs.
Doesn't FinalizationRegistry let you do exactly that?
https://developer.mozilla.org/en-US/docs/Web/JavaScript/Refe...
> A conforming JavaScript implementation, even one that does garbage collection, is not required to call cleanup callbacks. When and whether it does so is entirely down to the implementation of the JavaScript engine. When a registered object is reclaimed, any cleanup callbacks for it may be called then, or some time later, or not at all. It's likely that major implementations will call cleanup callbacks at some point during execution, but those calls may be substantially after the related object was reclaimed. Furthermore, if there is an object registered in two registries, there is no guarantee that the two callbacks are called next to each other — one may be called and the other never called, or the other may be called much later. There are also situations where even implementations that normally call cleanup callbacks are unlikely to call them:
I remember a couple of years ago (well probably around 2021) reading about GC exposure concerns and seeing some line in some TC39 doc like "users should not have visibility into collection" but if we've shipped weakrefs sounds like we're not thinking about that anymore
A proposal to add new ways of observing garbage collection will still be shot down immediately without a damn good justification.
This is more pedantry than a serious question. JavaScript has WeakReference, sure it'd be cumbersome and inefficient because you'd need to manually make and poll each thing you wanted to observe, but could it not be said that it does provide a view on deallocations?
Note that 90% of the uses for them actually shouldn't be using them, usually for subtle reasons. It's always a big cause for debate.
> Using str.frobnicate prevents TurboJit on line 63
> By using only a single instruction and two tables, we only increase the interpreter by a size of 1 instruction, and also keep the base interpreter ultra fast. I affectionally call this mechanism dual dispatch.
I really do hope they'll write that better explanation one day because this sounds pretty intriguing all on its own.
Kudos to those involved into making it happen.
I recently read an interview about implementing free-threading and getting modifications through the ecosystem to really enable it: https://alexalejandre.com/programming/interview-with-ngoldba...
The guy said he hopes the free-threaded build'll be the only one in "3.16 or 3.17", I wonder if that should apply to the JIT too or how the JIT and interpreter interact.
Having to have thread safe code all over the place just for the 1% of users who need to have multi-threading in Python and can't use subinterpreters for some reason is nuts.
Way more than 1% of the community, particularly of the community actively developing Python, wants free-threaded. The problem here is that the Python community consists of several different groups:
1. Basically pure Python code with no threading
2. Basically pure Python with appropriate thread safety
3. Basically pure Python code with already broken threaded code, just getting lucky for now
4. Mixed Python and C/C++/Rust code, with appropriate threading behavior in the C or C++ components
5. Mixed Python and C or C++ code, with C and C++ components depending on GIL behavior
Group 1 gets a slightly reduced performance. Groups 2 and 4 get a major win with free-threaded Python, being able to use threading through their interfaces to C/C++/Rust components. Group 3 is already writing buggy code and will probably see worse consequences from their existing bugs. Group 5 will have to either avoid threading in their Python code or rewrite their C/C++ components.
Right now, a big portion of the Python language developer base consists of Groups 2 and 4. Group 5 is basically perceived as holding Python-the-language and Python-the-implementations back.
Native code can already be multi-threaded so if you are using Python to drive parallelized native code, there's no win there. If your Python code is the bottleneck, well then you could have subinterpreters with shared buffers and locks. If you really need to have shared objects, do you actually need to mutate them from multiple interpreters? If not, what about exploring language support for frozen objects or proxies?
The only thing that free threading gives you is concurrent mutations to Python objects, which is like, whatever. In all my years of writing Python I have never once found myself thinking "I wish I could mutate the same object from two different threads".
Microsoft used to do this for their C runtime library.
A lot of Python code still leans on CPython internals, C extensions, debuggers, or odd platform behavior, so PyPy works until some dependency or tool turns that gap into a support problem.
The JIT helps on hot loops, but for mixed workloads the warmup cost and compatibility tax are enough to keep most teams on the interpreter their deps target first.
See https://github.com/numpy/numpy/issues/30416 for example. It's not being updated for compatibility with new versions of Python.
So it’s not unmaintained, no. But the project is currently under resourced to keep up with the latest Python spec.
You said it was defunct, which would mean there aren't still people working on it.
Also CPython 3.10 is not EOL so library authors won't be using anything from 3.11 anyway.
Like this is a big deal to get a project to a state where volunteers are spun up and actively breaking tasks and getting work done, no? It's a python JIT something I know next to nothing about — as do most application developers — which tells one how difficult this must have been.
The funding was Microsoft employing most of the team. They were laid off (or at least, moved onto different projects), apparently because they weren't working on AI.
(The latter is probably more to do with the preferences they give it in the re-inforcement learning phase than anything technical, though.)
That is not remotely the case for anyone who produces quality work.
In my experience the people who care the most about code readability tend to be the people most opinionated on having the right abstractions, which are historically not available in Go.
This would be a potential case for a new major version number.
It will be interesting to see, moving forward, what languages survive. A 15% perf increase seems nice, until you realize that you get a 10x increase porting to Rust (and the AI does it for you).
Maybe library use/popularity is somewhat related to backwards compatibility.
Disclaimer: I teach Python for a living.
So, you keep reading/writing Python and push a button to get binary executables through whatever hoops are best today ?
(I haven't seen the "fits your brain" tagline in the recent past ...)
> taking backwards compatibility so seriously
Python’s backward compatibility story still isn’t great compared to things like the Go 1.x compatibility promise, and languages with formal specs like JS and C.
The Python devs still make breaking changes, they’ve just learned not to update the major version number when they do so.
I would say it's probably worth it to clean up all the junk that Python has accumulated... But it's definitely not very high up the list of languages in terms of backwards compatibility. In fact I'm struggling to think of other languages that are worse. Typescript probably? Certainly Go, C++ and Rust are significantly better.
The more likely reason is that there simply hasn't been that big a push for it. Ruby was dog slow before the JIT and Rails was very popular, so there was a lot of demand and room for improvement. PHP was the primary language used by Facebook for a long time, and they had deep pockets. JS powers the web, so there's a huge incentive for companies like Google to make it faster. Python never really had that same level of investment, at least from a performance standpoint.
To your point, though, the C API has made certain types of optimizations extremely difficult, as the PyPy team has figured out.
But the main problem was actually that pypy was never adopted as “the JIT” mechanism. That would have made a huge difference a long time ago and made sure they evolved in lock step.
AFAIK it was not driven by anything on the tech side. It was simply unlucky timing, the project getting in the middle of Microsoft's heavy handed push to cut everything. So much so that the people who were hired by MS to work on this found out they were laid off in a middle of a conference where they were giving talks on it.
Or lack of incentive?
Alot of big python projects that does machine learning and data processing offloads the heavy data processing from pure python code to libraries like numpy and pandas that take advantage of C api binding to do native execution.
A worthwhile JIT is a fully optimizing compiler, and that is the hard part. Language semantics are much less important - dynamic languages aren’t particularly harder here, but the performance roof is obviously just much lower.
Including simply implementing the slow parts in C, such as the high performance machine learning ecosystem that exists in Python.