Following system colour scheme Selected dark colour scheme Selected light colour scheme

Python Enhancement Proposals

PEP 554 – Multiple Interpreters in the Stdlib

Author:
Eric Snow <ericsnowcurrently at gmail.com>
Discussions-To:
Discourse thread
Status:
Superseded
Type:
Standards Track
Created:
05-Sep-2017
Python-Version:
3.13
Post-History:
07-Sep-2017, 08-Sep-2017, 13-Sep-2017, 05-Dec-2017, 04-May-2020, 14-Mar-2023, 01-Nov-2023
Superseded-By:
734

Table of Contents

Note

This PEP effectively continues in a cleaner form in PEP 734. This PEP is kept as-is for the sake of the various sections of background information and deferred/rejected ideas that have been stripped from PEP 734.

Abstract

CPython has supported multiple interpreters in the same process (AKA “subinterpreters”) since version 1.5 (1997). The feature has been available via the C-API. [c-api] Multiple interpreters operate in relative isolation from one another, which facilitates novel alternative approaches to concurrency.

This proposal introduces the stdlib interpreters module. It exposes the basic functionality of multiple interpreters already provided by the C-API, along with basic support for communicating between interpreters. This module is especially relevant since PEP 684 introduced a per-interpreter GIL in Python 3.12.

Proposal

Summary:

  • add a new stdlib module: “interpreters”
  • add concurrent.futures.InterpreterPoolExecutor
  • help for extension module maintainers

The “interpreters” Module

The interpreters module will provide a high-level interface to the multiple interpreter functionality, and wrap a new low-level _interpreters (in the same way as the threading module). See the Examples section for concrete usage and use cases.

Along with exposing the existing (in CPython) multiple interpreter support, the module will also support a basic mechanism for passing data between interpreters. That involves setting “shareable” objects in the __main__ module of a target subinterpreter. Some such objects, like os.pipe(), may be used to communicate further. The module will also provide a minimal implementation of “channels” as a demonstration of cross-interpreter communication.

Note that objects are not shared between interpreters since they are tied to the interpreter in which they were created. Instead, the objects’ data is passed between interpreters. See the Shared Data and API For Communication sections for more details about sharing/communicating between interpreters.

API summary for interpreters module

Here is a summary of the API for the interpreters module. For a more in-depth explanation of the proposed classes and functions, see the “interpreters” Module API section below.

For creating and using interpreters:

signature description
list_all() -> [Interpreter] Get all existing interpreters.
get_current() -> Interpreter Get the currently running interpreter.
get_main() -> Interpreter Get the main interpreter.
create() -> Interpreter Initialize a new (idle) Python interpreter.

signature description
class Interpreter A single interpreter.
.id The interpreter’s ID (read-only).
.is_running() -> bool Is the interpreter currently executing code?
.close() Finalize and destroy the interpreter.
.set_main_attrs(**kwargs) Bind “shareable” objects in __main__.
.get_main_attr(name) Get a “shareable” object from __main__.
.exec(src_str, /)
Run the given source code in the interpreter
(in the current thread).

For communicating between interpreters:

signature description
is_shareable(obj) -> Bool
Can the object’s data be passed
between interpreters?
create_channel() -> (RecvChannel, SendChannel)
Create a new channel for passing
data between interpreters.

concurrent.futures.InterpreterPoolExecutor

An executor will be added that extends ThreadPoolExecutor to run per-thread tasks in subinterpreters. Initially, the only supported tasks will be whatever Interpreter.exec() takes (e.g. a str script). However, we may also support some functions, as well as eventually a separate method for pickling the task and arguments, to reduce friction (at the expense of performance for short-running tasks).

Help for Extension Module Maintainers

In practice, an extension that implements multi-phase init (PEP 489) is considered isolated and thus compatible with multiple interpreters. Otherwise it is “incompatible”.

Many extension modules are still incompatible. The maintainers and users of such extension modules will both benefit when they are updated to support multiple interpreters. In the meantime, users may become confused by failures when using multiple interpreters, which could negatively impact extension maintainers. See Concerns below.

To mitigate that impact and accelerate compatibility, we will do the following:

  • be clear that extension modules are not required to support use in multiple interpreters
  • raise ImportError when an incompatible module is imported in a subinterpreter
  • provide resources (e.g. docs) to help maintainers reach compatibility
  • reach out to the maintainers of Cython and of the most used extension modules (on PyPI) to get feedback and possibly provide assistance

Examples

Run isolated code in current OS thread

interp = interpreters.create()
print('before')
interp.exec('print("during")')
print('after')

Run in a different thread

interp = interpreters.create()
def run():
    interp.exec('print("during")')
t = threading.Thread(target=run)
print('before')
t.start()
t.join()
print('after')

Pre-populate an interpreter

interp = interpreters.create()
interp.exec(tw.dedent("""
    import some_lib
    import an_expensive_module
    some_lib.set_up()
    """))
wait_for_request()
interp.exec(tw.dedent("""
    some_lib.handle_request()
    """))

Handling an exception

interp = interpreters.create()
try:
    interp.exec(tw.dedent("""
        raise KeyError
        """))
except interpreters.RunFailedError as exc:
    print(f"got the error from the subinterpreter: {exc}")

Re-raising an exception

interp = interpreters.create()
try:
    try:
        interp.exec(tw.dedent("""
            raise KeyError
            """))
    except interpreters.RunFailedError as exc:
        raise exc.__cause__
except KeyError:
    print("got a KeyError from the subinterpreter")

Note that this pattern is a candidate for later improvement.

Interact with the __main__ namespace

interp = interpreters.create()
interp.set_main_attrs(a=1, b=2)
interp.exec(tw.dedent("""
    res = do_something(a, b)
    """))
res = interp.get_main_attr('res')

Synchronize using an OS pipe

interp = interpreters.create()
r1, s1 = os.pipe()
r2, s2 = os.pipe()

def task():
    interp.exec(tw.dedent(f"""
        import os
        os.read({r1}, 1)
        print('during B')
        os.write({s2}, '')
        """))

t = threading.thread(target=task)
t.start()
print('before')
os.write(s1, '')
print('during A')
os.read(r2, 1)
print('after')
t.join()

Sharing a file descriptor

interp = interpreters.create()
with open('spamspamspam') as infile:
    interp.set_main_attrs(fd=infile.fileno())
    interp.exec(tw.dedent(f"""
        import os
        for line in os.fdopen(fd):
            print(line)
        """))

Passing objects via pickle

interp = interpreters.create()
r, s = os.pipe()
interp.exec(tw.dedent(f"""
    import os
    import pickle
    reader = {r}
    """))
interp.exec(tw.dedent("""
        data = b''
        c = os.read(reader, 1)
        while c != b'\x00':
            while c != b'\x00':
                data += c
                c = os.read(reader, 1)
            obj = pickle.loads(data)
            do_something(obj)
            c = os.read(reader, 1)
        """))
for obj in input:
    data = pickle.dumps(obj)
    os.write(s, data)
    os.write(s, b'\x00')
os.write(s, b'\x00')

Capturing an interpreter’s stdout

interp = interpreters.create()
stdout = io.StringIO()
with contextlib.redirect_stdout(stdout):
    interp.exec(tw.dedent("""
        print('spam!')
        """))
assert(stdout.getvalue() == 'spam!')

# alternately:
interp.exec(tw.dedent("""
    import contextlib, io
    stdout = io.StringIO()
    with contextlib.redirect_stdout(stdout):
        print('spam!')
    captured = stdout.getvalue()
    """))
captured = interp.get_main_attr('captured')
assert(captured == 'spam!')

A pipe (os.pipe()) could be used similarly.

Running a module

interp = interpreters.create()
main_module = mod_name
interp.exec(f'import runpy; runpy.run_module({main_module!r})')

Running as script (including zip archives & directories)

interp = interpreters.create()
main_script = path_name
interp.exec(f"import runpy; runpy.run_path({main_script!r})")

Using a channel to communicate

tasks_recv, tasks = interpreters.create_channel()
results, results_send = interpreters.create_channel()

def worker():
    interp = interpreters.create()
    interp.set_main_attrs(tasks=tasks_recv, results=results_send)
    interp.exec(tw.dedent("""
        def handle_request(req):
            ...

        def capture_exception(exc):
            ...

        while True:
            try:
                req = tasks.recv()
            except Exception:
                # channel closed
                break
            try:
                res = handle_request(req)
            except Exception as exc:
                res = capture_exception(exc)
            results.send_nowait(res)
        """))
threads = [threading.Thread(target=worker) for _ in range(20)]
for t in threads:
    t.start()

requests = ...
for req in requests:
    tasks.send(req)
tasks.close()

for t in threads:
    t.join()

Sharing a memoryview (imagine map-reduce)

data, chunksize = read_large_data_set()
buf = memoryview(data)
numchunks = (len(buf) + 1) / chunksize
results = memoryview(b'\0' * numchunks)

tasks_recv, tasks = interpreters.create_channel()

def worker():
    interp = interpreters.create()
    interp.set_main_attrs(data=buf, results=results, tasks=tasks_recv)
    interp.exec(tw.dedent("""
        while True:
            try:
                req = tasks.recv()
            except Exception:
                # channel closed
                break
            resindex, start, end = req
            chunk = data[start: end]
            res = reduce_chunk(chunk)
            results[resindex] = res
        """))
t = threading.Thread(target=worker)
t.start()

for i in range(numchunks):
    if not workers_running():
        raise ...
    start = i * chunksize
    end = start + chunksize
    if end > len(buf):
        end = len(buf)
    tasks.send((start, end, i))
tasks.close()
t.join()

use_results(results)

Rationale

Running code in multiple interpreters provides a useful level of isolation within the same process. This can be leveraged in a number of ways. Furthermore, subinterpreters provide a well-defined framework in which such isolation may extended. (See PEP 684.)

Alyssa (Nick) Coghlan explained some of the benefits through a comparison with multi-processing [benefits]:

[I] expect that communicating between subinterpreters is going
to end up looking an awful lot like communicating between
subprocesses via shared memory.

The trade-off between the two models will then be that one still
just looks like a single process from the point of view of the
outside world, and hence doesn't place any extra demands on the
underlying OS beyond those required to run CPython with a single
interpreter, while the other gives much stricter isolation
(including isolating C globals in extension modules), but also
demands much more from the OS when it comes to its IPC
capabilities.

The security risk profiles of the two approaches will also be quite
different, since using subinterpreters won't require deliberately
poking holes in the process isolation that operating systems give
you by default.

CPython has supported multiple interpreters, with increasing levels of support, since version 1.5. While the feature has the potential to be a powerful tool, it has suffered from neglect because the multiple interpreter capabilities are not readily available directly from Python. Exposing the existing functionality in the stdlib will help reverse the situation.

This proposal is focused on enabling the fundamental capability of multiple interpreters, isolated from each other, in the same Python process. This is a new area for Python so there is relative uncertainly about the best tools to provide as companions to interpreters. Thus we minimize the functionality we add in the proposal as much as possible.

Concerns

  • “subinterpreters are not worth the trouble”

Some have argued that subinterpreters do not add sufficient benefit to justify making them an official part of Python. Adding features to the language (or stdlib) has a cost in increasing the size of the language. So an addition must pay for itself.

In this case, multiple interpreter support provide a novel concurrency model focused on isolated threads of execution. Furthermore, they provide an opportunity for changes in CPython that will allow simultaneous use of multiple CPU cores (currently prevented by the GIL–see PEP 684).

Alternatives to subinterpreters include threading, async, and multiprocessing. Threading is limited by the GIL and async isn’t the right solution for every problem (nor for every person). Multiprocessing is likewise valuable in some but not all situations. Direct IPC (rather than via the multiprocessing module) provides similar benefits but with the same caveat.

Notably, subinterpreters are not intended as a replacement for any of the above. Certainly they overlap in some areas, but the benefits of subinterpreters include isolation and (potentially) performance. In particular, subinterpreters provide a direct route to an alternate concurrency model (e.g. CSP) which has found success elsewhere and will appeal to some Python users. That is the core value that the interpreters module will provide.

  • “stdlib support for multiple interpreters adds extra burden on C extension authors”

In the Interpreter Isolation section below we identify ways in which isolation in CPython’s subinterpreters is incomplete. Most notable is extension modules that use C globals to store internal state. (PEP 3121 and PEP 489 provide a solution to that problem, followed by some extra APIs that improve efficiency, e.g. PEP 573).

Consequently, projects that publish extension modules may face an increased maintenance burden as their users start using subinterpreters, where their modules may break. This situation is limited to modules that use C globals (or use libraries that use C globals) to store internal state. For numpy, the reported-bug rate is one every 6 months. [bug-rate]

Ultimately this comes down to a question of how often it will be a problem in practice: how many projects would be affected, how often their users will be affected, what the additional maintenance burden will be for projects, and what the overall benefit of subinterpreters is to offset those costs. The position of this PEP is that the actual extra maintenance burden will be small and well below the threshold at which subinterpreters are worth it.

  • “creating a new concurrency API deserves much more thought and experimentation, so the new module shouldn’t go into the stdlib right away, if ever”

Introducing an API for a new concurrency model, like happened with asyncio, is an extremely large project that requires a lot of careful consideration. It is not something that can be done as simply as this PEP proposes and likely deserves significant time on PyPI to mature. (See Nathaniel’s post on python-dev.)

However, this PEP does not propose any new concurrency API. At most it exposes minimal tools (e.g. subinterpreters, channels) which may be used to write code that follows patterns associated with (relatively) new-to-Python concurrency models. Those tools could also be used as the basis for APIs for such concurrency models. Again, this PEP does not propose any such API.

  • “there is no point to exposing subinterpreters if they still share the GIL”
  • “the effort to make the GIL per-interpreter is disruptive and risky”

A common misconception is that this PEP also includes a promise that interpreters will no longer share the GIL. When that is clarified, the next question is “what is the point?”. This is already answered at length in this PEP. Just to be clear, the value lies in:

* increase exposure of the existing feature, which helps improve
  the code health of the entire CPython runtime
* expose the (mostly) isolated execution of interpreters
* preparation for per-interpreter GIL
* encourage experimentation
  • “data sharing can have a negative impact on cache performance in multi-core scenarios”

(See [cache-line-ping-pong].)

This shouldn’t be a problem for now as we have no immediate plans to actually share data between interpreters, instead focusing on copying.

About Subinterpreters

Concurrency

Concurrency is a challenging area of software development. Decades of research and practice have led to a wide variety of concurrency models, each with different goals. Most center on correctness and usability.

One class of concurrency models focuses on isolated threads of execution that interoperate through some message passing scheme. A notable example is Communicating Sequential Processes [CSP] (upon which Go’s concurrency is roughly based). The intended isolation inherent to CPython’s interpreters makes them well-suited to this approach.

Shared Data

CPython’s interpreters are inherently isolated (with caveats explained below), in contrast to threads. So the same communicate-via-shared-memory approach doesn’t work. Without an alternative, effective use of concurrency via multiple interpreters is significantly limited.

The key challenge here is that sharing objects between interpreters faces complexity due to various constraints on object ownership, visibility, and mutability. At a conceptual level it’s easier to reason about concurrency when objects only exist in one interpreter at a time. At a technical level, CPython’s current memory model limits how Python objects may be shared safely between interpreters; effectively, objects are bound to the interpreter in which they were created. Furthermore, the complexity of object sharing increases as interpreters become more isolated, e.g. after GIL removal (though this is mitigated somewhat for some “immortal” objects (see PEP 683).

Consequently, the mechanism for sharing needs to be carefully considered. There are a number of valid solutions, several of which may be appropriate to support in Python’s stdlib and C-API. Any such solution is likely to share many characteristics with the others.

In the meantime, we propose here a minimal solution (Interpreter.set_main_attrs()), which sets some precedent for how objects are shared. More importantly, it facilitates the introduction of more advanced approaches later and allows them to coexist and cooperate. In part to demonstrate that, we will provide a basic implementation of “channels”, as a somewhat more advanced sharing solution.

Separate proposals may cover:

  • the addition of a public C-API based on the implementation Interpreter.set_main_attrs()
  • the addition of other sharing approaches to the “interpreters” module

The fundamental enabling feature for communication is that most objects can be converted to some encoding of underlying raw data, which is safe to be passed between interpreters. For example, an int object can be turned into a C long value, sent to another interpreter, and turned back into an int object there. As another example, None may be passed as-is.

Regardless, the effort to determine the best way forward here is mostly outside the scope of this PEP. In the meantime, this proposal describes a basic interim solution using pipes (os.pipe()), as well as providing a dedicated capability (“channels”). See API For Communication below.

Interpreter Isolation

CPython’s interpreters are intended to be strictly isolated from each other. Each interpreter has its own copy of all modules, classes, functions, and variables. The same applies to state in C, including in extension modules. The CPython C-API docs explain more. [caveats]

However, there are ways in which interpreters do share some state. First of all, some process-global state remains shared:

  • file descriptors
  • low-level env vars
  • process memory (though allocators are isolated)
  • builtin types (e.g. dict, bytes)
  • singletons (e.g. None)
  • underlying static module data (e.g. functions) for builtin/extension/frozen modules

There are no plans to change this.

Second, some isolation is faulty due to bugs or implementations that did not take subinterpreters into account. This includes things like extension modules that rely on C globals. [cryptography] In these cases bugs should be opened (some are already):

Finally, some potential isolation is missing due to the current design of CPython. Improvements are currently going on to address gaps in this area:

  • extensions using the PyGILState_* API are somewhat incompatible [gilstate]

Existing Usage

Multiple interpreter support has not been a widely used feature. In fact, there have been only a handful of documented cases of widespread usage, including mod_wsgi, OpenStack Ceph, and JEP. On the one hand, these cases provide confidence that existing multiple interpreter support is relatively stable. On the other hand, there isn’t much of a sample size from which to judge the utility of the feature.

Alternate Python Implementations

I’ve solicited feedback from various Python implementors about support for subinterpreters. Each has indicated that they would be able to support multiple interpreters in the same process (if they choose to) without a lot of trouble. Here are the projects I contacted:

  • jython ([jython])
  • ironpython (personal correspondence)
  • pypy (personal correspondence)
  • micropython (personal correspondence)

“interpreters” Module API

The module provides the following functions:

list_all() -> [Interpreter]

   Return a list of all existing interpreters.

get_current() => Interpreter

   Return the currently running interpreter.

get_main() => Interpreter

   Return the main interpreter.  If the Python implementation
   has no concept of a main interpreter then return None.

create() -> Interpreter

   Initialize a new Python interpreter and return it.
   It will remain idle until something is run in it and always
   run in its own thread.

is_shareable(obj) -> bool:

   Return True if the object may be "shared" between interpreters.
   This does not necessarily mean that the actual objects will be
   shared.  Instead, it means that the objects' underlying data will
   be shared in a cross-interpreter way, whether via a proxy, a
   copy, or some other means.

The module also provides the following class:

class Interpreter(id):

   id -> int:

      The interpreter's ID. (read-only)

   is_running() -> bool:

      Return whether or not the interpreter's "exec()" is currently
      executing code.  Code running in subthreads is ignored.
      Calling this on the current interpreter will always return True.

   close():

      Finalize and destroy the interpreter.

      This may not be called on an already running interpreter.
      Doing so results in a RuntimeError.

   set_main_attrs(iterable_or_mapping, /):
   set_main_attrs(**kwargs):

      Set attributes in the interpreter's __main__ module
      corresponding to the given name-value pairs.  Each value
      must be a "shareable" object and will be converted to a new
      object (e.g. copy, proxy) in whatever way that object's type
      defines.  If an attribute with the same name is already set,
      it will be overwritten.

      This method is helpful for setting up an interpreter before
      calling exec().

   get_main_attr(name, default=None, /):

      Return the value of the corresponding attribute of the
      interpreter's __main__ module.  If the attribute isn't set
      then the default is returned.  If it is set, but the value
      isn't "shareable" then a ValueError is raised.

      This may be used to introspect the __main__ module, as well
      as a very basic mechanism for "returning" one or more results
      from Interpreter.exec().

   exec(source_str, /):

      Run the provided Python source code in the interpreter,
      in its __main__ module.

      This may not be called on an already running interpreter.
      Doing so results in a RuntimeError.

      An "interp.exec()" call is similar to a builtin exec() call
      (or to calling a function that returns None).  Once
      "interp.exec()" completes, the code that called "exec()"
      continues executing (in the original interpreter).  Likewise,
      if there is any uncaught exception then it effectively
      (see below) propagates into the code where ``interp.exec()``
      was called.  Like exec() (and threads), but unlike function
      calls, there is no return value.  If any "return" value from
      the code is needed, send the data out via a pipe (os.pipe())
      or channel or other cross-interpreter communication mechanism.

      The big difference from exec() or functions is that
      "interp.exec()" executes the code in an entirely different
      interpreter, with entirely separate state.  The interpreters
      are completely isolated from each other, so the state of the
      original interpreter (including the code it was executing in
      the current OS thread) does not affect the state of the target
      interpreter (the one that will execute the code).  Likewise,
      the target does not affect the original, nor any of its other
      threads.

      Instead, the state of the original interpreter (for this thread)
      is frozen, and the code it's executing code completely blocks.
      At that point, the target interpreter is given control of the
      OS thread.  Then, when it finishes executing, the original
      interpreter gets control back and continues executing.

      So calling "interp.exec()" will effectively cause the current
      Python thread to completely pause.  Sometimes you won't want
      that pause, in which case you should make the "exec()" call in
      another thread.  To do so, add a function that calls
      "interp.exec()" and then run that function in a normal
      "threading.Thread".

      Note that the interpreter's state is never reset, neither
      before "interp.exec()" executes the code nor after.  Thus the
      interpreter state is preserved between calls to
      "interp.exec()".  This includes "sys.modules", the "builtins"
      module, and the internal state of C extension modules.

      Also note that "interp.exec()" executes in the namespace of the
      "__main__" module, just like scripts, the REPL, "-m", and
      "-c".  Just as the interpreter's state is not ever reset, the
      "__main__" module is never reset.  You can imagine
      concatenating the code from each "interp.exec()" call into one
      long script.  This is the same as how the REPL operates.

      Supported code: source text.

In addition to the functionality of Interpreter.set_main_attrs(), the module provides a related way to pass data between interpreters: channels. See Channels below.

Uncaught Exceptions

Regarding uncaught exceptions in Interpreter.exec(), we noted that they are “effectively” propagated into the code where interp.exec() was called. To prevent leaking exceptions (and tracebacks) between interpreters, we create a surrogate of the exception and its traceback (see traceback.TracebackException), set it to __cause__ on a new interpreters.RunFailedError, and raise that.

Directly raising (a proxy of) the exception is problematic since it’s harder to distinguish between an error in the interp.exec() call and an uncaught exception from the subinterpreter.

Interpreter Restrictions

Every new interpreter created by interpreters.create() now has specific restrictions on any code it runs. This includes the following:

  • importing an extension module fails if it does not implement multi-phase init
  • daemon threads may not be created
  • os.fork() is not allowed (so no multiprocessing)
  • os.exec*() is not allowed (but “fork+exec”, a la subprocess is okay)

Note that interpreters created with the existing C-API do not have these restrictions. The same is true for the “main” interpreter, so existing use of Python will not change.

We may choose to later loosen some of the above restrictions or provide a way to enable/disable granular restrictions individually. Regardless, requiring multi-phase init from extension modules will always be a default restriction.

API For Communication

As discussed in Shared Data above, multiple interpreter support is less useful without a mechanism for sharing data (communicating) between them. Sharing actual Python objects between interpreters, however, has enough potential problems that we are avoiding support for that in this proposal. Nor, as mentioned earlier, are we adding anything more than a basic mechanism for communication.

That mechanism is the Interpreter.set_main_attrs() method. It may be used to set up global variables before Interpreter.exec() is called. The name-value pairs passed to set_main_attrs() are bound as attributes of the interpreter’s __main__ module. The values must be “shareable”. See Shareable Types below.

Additional approaches to communicating and sharing objects are enabled through Interpreter.set_main_attrs(). A shareable object could be implemented which works like a queue, but with cross-interpreter safety. In fact, this PEP does include an example of such an approach: channels.

Shareable Types

An object is “shareable” if its type supports shareable instances. The type must implement a new internal protocol, which is used to convert an object to interpreter-independent data and then converted back to an object on the other side. Also see is_shareable() above.

A minimal set of simple, immutable builtin types will be supported initially, including:

  • None
  • bool
  • bytes
  • str
  • int
  • float

We will also support a small number of complex types initially:

Further builtin types may be supported later, complex or not. Limiting the initial shareable types is a practical matter, reducing the potential complexity of the initial implementation. There are a number of strategies we may pursue in the future to expand supported objects, once we have more experience with interpreter isolation.

In the meantime, a separate proposal will discuss making the internal protocol (and C-API) used by Interpreter.set_main_attrs() public. With that protocol, support for other types could be added by extension modules.

Communicating Through OS Pipes

Even without a dedicated object for communication, users may already use existing tools. For example, one basic approach for sending data between interpreters is to use a pipe (see os.pipe()):

  1. interpreter A calls os.pipe() to get a read/write pair of file descriptors (both int objects)
  2. interpreter A calls interp.set_main_attrs(), binding the read FD (or embeds it using string formatting)
  3. interpreter A calls interp.exec() on interpreter B
  4. interpreter A writes some bytes to the write FD
  5. interpreter B reads those bytes

Several of the earlier examples demonstrate this, such as Synchronize using an OS pipe.

Channels

The interpreters module will include a dedicated solution for passing object data between interpreters: channels. They are included in the module in part to provide an easier mechanism than using os.pipe() and in part to demonstrate how libraries may take advantage of Interpreter.set_main_attrs() and the protocol it uses.

A channel is a simplex FIFO. It is a basic, opt-in data sharing mechanism that draws inspiration from pipes, queues, and CSP’s channels. [fifo] The main difference from pipes is that channels can be associated with zero or more interpreters on either end. Like queues, which are also many-to-many, channels are buffered (though they also offer methods with unbuffered semantics).

Channels have two operations: send and receive. A key characteristic of those operations is that channels transmit data derived from Python objects rather than the objects themselves. When objects are sent, their data is extracted. When the “object” is received in the other interpreter, the data is converted back into an object owned by that interpreter.

To make this work, the mutable shared state will be managed by the Python runtime, not by any of the interpreters. Initially we will support only one type of objects for shared state: the channels provided by interpreters.create_channel(). Channels, in turn, will carefully manage passing objects between interpreters.

This approach, including keeping the API minimal, helps us avoid further exposing any underlying complexity to Python users.

The interpreters module provides the following function related to channels:

create_channel() -> (RecvChannel, SendChannel):

   Create a new channel and return (recv, send), the RecvChannel
   and SendChannel corresponding to the ends of the channel.

   Both ends of the channel are supported "shared" objects (i.e.
   may be safely shared by different interpreters.  Thus they
   may be set using "Interpreter.set_main_attrs()".

The module also provides the following channel-related classes:

class RecvChannel(id):

   The receiving end of a channel.  An interpreter may use this to
   receive objects from another interpreter.  Any type supported by
   Interpreter.set_main_attrs() will be supported here, though at
   first only a few of the simple, immutable builtin types
   will be supported.

   id -> int:

      The channel's unique ID.  The "send" end has the same one.

   recv(*, timeout=None):

      Return the next object from the channel.  If none have been
      sent then wait until the next send (or until the timeout is hit).

      At the least, the object will be equivalent to the sent object.
      That will almost always mean the same type with the same data,
      though it could also be a compatible proxy.  Regardless, it may
      use a copy of that data or actually share the data.  That's up
      to the object's type.

   recv_nowait(default=None):

      Return the next object from the channel.  If none have been
      sent then return the default.  Otherwise, this is the same
      as the "recv()" method.


class SendChannel(id):

   The sending end of a channel.  An interpreter may use this to
   send objects to another interpreter.  Any type supported by
   Interpreter.set_main_attrs() will be supported here, though
   at first only a few of the simple, immutable builtin types
   will be supported.

   id -> int:

      The channel's unique ID.  The "recv" end has the same one.

   send(obj, *, timeout=None):

      Send the object (i.e. its data) to the "recv" end of the
      channel.  Wait until the object is received.  If the object
      is not shareable then ValueError is raised.

      The builtin memoryview is supported, so sending a buffer
      across involves first wrapping the object in a memoryview
      and then sending that.

   send_nowait(obj):

      Send the object to the "recv" end of the channel.  This
      behaves the same as "send()", except for the waiting part.
      If no interpreter is currently receiving (waiting on the
      other end) then queue the object and return False.  Otherwise
      return True.

Caveats For Shared Objects

Again, Python objects are not shared between interpreters. However, in some cases data those objects wrap is actually shared and not just copied. One example might be PEP 3118 buffers.

In those cases the object in the original interpreter is kept alive until the shared data in the other interpreter is no longer used. Then object destruction can happen like normal in the original interpreter, along with the previously shared data.

Documentation

The new stdlib docs page for the interpreters module will include the following:

  • (at the top) a clear note that support for multiple interpreters is not required from extension modules
  • some explanation about what subinterpreters are
  • brief examples of how to use multiple interpreters (and communicating between them)
  • a summary of the limitations of using multiple interpreters
  • (for extension maintainers) a link to the resources for ensuring multiple interpreters compatibility
  • much of the API information in this PEP

Docs about resources for extension maintainers already exist on the Isolating Extension Modules howto page. Any extra help will be added there. For example, it may prove helpful to discuss strategies for dealing with linked libraries that keep their own subinterpreter-incompatible global state.

Note that the documentation will play a large part in mitigating any negative impact that the new interpreters module might have on extension module maintainers.

Also, the ImportError for incompatible extension modules will be updated to clearly say it is due to missing multiple interpreters compatibility and that extensions are not required to provide it. This will help set user expectations properly.

Alternative Solutions

One possible alternative to a new module is to add support for interpreters to concurrent.futures. There are several reasons why that wouldn’t work:

  • the obvious place to look for multiple interpreters support is an “interpreters” module, much as with “threading”, etc.
  • concurrent.futures is all about executing functions but currently we don’t have a good way to run a function from one interpreter in another

Similar reasoning applies for support in the multiprocessing module.

Open Questions

  • will is be too confusing that interp.exec() runs in the current thread?
  • should we add pickling fallbacks right now for interp.exec(), and/or Interpreter.set_main_attrs() and Interpreter.get_main_attr()?
  • should we support (limited) functions in interp.exec() right now?
  • rename Interpreter.close() to Interpreter.destroy()?
  • drop Interpreter.get_main_attr(), since we have channels?
  • should channels be its own PEP?

Deferred Functionality

In the interest of keeping this proposal minimal, the following functionality has been left out for future consideration. Note that this is not a judgement against any of said capability, but rather a deferment. That said, each is arguably valid.

Add convenience API

There are a number of things I can imagine would smooth out hypothetical rough edges with the new module:

  • add something like Interpreter.run() or Interpreter.call() that calls interp.exec() and falls back to pickle
  • fall back to pickle in Interpreter.set_main_attrs() and Interpreter.get_main_attr()

These would be easy to do if this proves to be a pain point.

Avoid possible confusion about interpreters running in the current thread

One regular point of confusion has been that Interpreter.exec() executes in the current OS thread, temporarily blocking the current Python thread. It may be worth doing something to avoid that confusion.

Some possible solutions for this hypothetical problem:

  • by default, run in a new thread?
  • add Interpreter.exec_in_thread()?
  • add Interpreter.exec_in_current_thread()?

In earlier versions of this PEP the method was interp.run(). The simple change to interp.exec() alone will probably reduce confusion sufficiently, when coupled with educating users via the docs. It it turns out to be a real problem, we can pursue one of the alternatives at that point.

Clarify “running” vs. “has threads”

Interpreter.is_running() refers specifically to whether or not Interpreter.exec() (or similar) is running somewhere. It does not say anything about if the interpreter has any subthreads running. That information might be helpful.

Some things we could do:

  • rename Interpreter.is_running() to Interpreter.is_running_main()
  • add Interpreter.has_threads(), to complement Interpreter.is_running()
  • expand to Interpreter.is_running(main=True, threads=False)

None of these are urgent and any could be done later, if desired.

A Dunder Method For Sharing

We could add a special method, like __xid__ to correspond to tp_xid. At the very least, it would allow Python types to convert their instances to some other type that implements tp_xid.

The problem is that exposing this capability to Python code presents a degree of complixity that hasn’t been explored yet, nor is there a compelling case to investigate that complexity.

Interpreter.call()

It would be convenient to run existing functions in subinterpreters directly. Interpreter.exec() could be adjusted to support this or a call() method could be added:

Interpreter.call(f, *args, **kwargs)

This suffers from the same problem as sharing objects between interpreters via queues. The minimal solution (running a source string) is sufficient for us to get the feature out where it can be explored.

Interpreter.run_in_thread()

This method would make a interp.exec() call for you in a thread. Doing this using only threading.Thread and interp.exec() is relatively trivial so we’ve left it out.

Synchronization Primitives

The threading module provides a number of synchronization primitives for coordinating concurrent operations. This is especially necessary due to the shared-state nature of threading. In contrast, interpreters do not share state. Data sharing is restricted to the runtime’s shareable objects capability, which does away with the need for explicit synchronization. If any sort of opt-in shared state support is added to CPython’s interpreters in the future, that same effort can introduce synchronization primitives to meet that need.

CSP Library

A csp module would not be a large step away from the functionality provided by this PEP. However, adding such a module is outside the minimalist goals of this proposal.

Syntactic Support

The Go language provides a concurrency model based on CSP, so it’s similar to the concurrency model that multiple interpreters support. However, Go also provides syntactic support, as well as several builtin concurrency primitives, to make concurrency a first-class feature. Conceivably, similar syntactic (and builtin) support could be added to Python using interpreters. However, that is way outside the scope of this PEP!

Multiprocessing

The multiprocessing module could support interpreters in the same way it supports threads and processes. In fact, the module’s maintainer, Davin Potts, has indicated this is a reasonable feature request. However, it is outside the narrow scope of this PEP.

C-extension opt-in/opt-out

By using the PyModuleDef_Slot introduced by PEP 489, we could easily add a mechanism by which C-extension modules could opt out of multiple interpreter support. Then the import machinery, when operating in a subinterpreter, would need to check the module for support. It would raise an ImportError if unsupported.

Alternately we could support opting in to multiple interpreters support. However, that would probably exclude many more modules (unnecessarily) than the opt-out approach. Also, note that PEP 489 defined that an extension’s use of the PEP’s machinery implies multiple interpreters support.

The scope of adding the ModuleDef slot and fixing up the import machinery is non-trivial, but could be worth it. It all depends on how many extension modules break under subinterpreters. Given that there are relatively few cases we know of through mod_wsgi, we can leave this for later.

Poisoning channels

CSP has the concept of poisoning a channel. Once a channel has been poisoned, any send() or recv() call on it would raise a special exception, effectively ending execution in the interpreter that tried to use the poisoned channel.

This could be accomplished by adding a poison() method to both ends of the channel. The close() method can be used in this way (mostly), but these semantics are relatively specialized and can wait.

Resetting __main__

As proposed, every call to Interpreter.exec() will execute in the namespace of the interpreter’s existing __main__ module. This means that data persists there between interp.exec() calls. Sometimes this isn’t desirable and you want to execute in a fresh __main__. Also, you don’t necessarily want to leak objects there that you aren’t using any more.

Note that the following won’t work right because it will clear too much (e.g. __name__ and the other “__dunder__” attributes:

interp.exec('globals().clear()')

Possible solutions include:

  • a create() arg to indicate resetting __main__ after each interp.exec() call
  • an Interpreter.reset_main flag to support opting in or out after the fact
  • an Interpreter.reset_main() method to opt in when desired
  • importlib.util.reset_globals() [reset_globals]

Also note that resetting __main__ does nothing about state stored in other modules. So any solution would have to be clear about the scope of what is being reset. Conceivably we could invent a mechanism by which any (or every) module could be reset, unlike reload() which does not clear the module before loading into it.

Regardless, since __main__ is the execution namespace of the interpreter, resetting it has a much more direct correlation to interpreters and their dynamic state than does resetting other modules. So a more generic module reset mechanism may prove unnecessary.

This isn’t a critical feature initially. It can wait until later if desirable.

Resetting an interpreter’s state

It may be nice to re-use an existing subinterpreter instead of spinning up a new one. Since an interpreter has substantially more state than just the __main__ module, it isn’t so easy to put an interpreter back into a pristine/fresh state. In fact, there may be parts of the state that cannot be reset from Python code.

A possible solution is to add an Interpreter.reset() method. This would put the interpreter back into the state it was in when newly created. If called on a running interpreter it would fail (hence the main interpreter could never be reset). This would likely be more efficient than creating a new interpreter, though that depends on what optimizations will be made later to interpreter creation.

While this would potentially provide functionality that is not otherwise available from Python code, it isn’t a fundamental functionality. So in the spirit of minimalism here, this can wait. Regardless, I doubt it would be controversial to add it post-PEP.

Copy an existing interpreter’s state

Relatedly, it may be useful to support creating a new interpreter based on an existing one, e.g. Interpreter.copy(). This ties into the idea that a snapshot could be made of an interpreter’s memory, which would make starting up CPython, or creating new interpreters, faster in general. The same mechanism could be used for a hypothetical Interpreter.reset(), as described previously.

Shareable file descriptors and sockets

Given that file descriptors and sockets are process-global resources, making them shareable is a reasonable idea. They would be a good candidate for the first effort at expanding the supported shareable types. They aren’t strictly necessary for the initial API.

Integration with async

Per Antoine Pitrou [async]:

Has any thought been given to how FIFOs could integrate with async
code driven by an event loop (e.g. asyncio)?  I think the model of
executing several asyncio (or Tornado) applications each in their
own subinterpreter may prove quite interesting to reconcile multi-
core concurrency with ease of programming.  That would require the
FIFOs to be able to synchronize on something an event loop can wait
on (probably a file descriptor?).

The basic functionality of multiple interpreters support does not depend on async and can be added later.

A possible solution is to provide async implementations of the blocking channel methods (recv(), and send()).

Alternately, “readiness callbacks” could be used to simplify use in async scenarios. This would mean adding an optional callback (kw-only) parameter to the recv_nowait() and send_nowait() channel methods. The callback would be called once the object was sent or received (respectively).

(Note that making channels buffered makes readiness callbacks less important.)

Support for iteration

Supporting iteration on RecvChannel (via __iter__() or _next__()) may be useful. A trivial implementation would use the recv() method, similar to how files do iteration. Since this isn’t a fundamental capability and has a simple analog, adding iteration support can wait until later.

Channel context managers

Context manager support on RecvChannel and SendChannel may be helpful. The implementation would be simple, wrapping a call to close() (or maybe release()) like files do. As with iteration, this can wait.

Pipes and Queues

With the proposed object passing mechanism of “os.pipe()”, other similar basic types aren’t strictly required to achieve the minimal useful functionality of multiple interpreters. Such types include pipes (like unbuffered channels, but one-to-one) and queues (like channels, but more generic). See below in Rejected Ideas for more information.

Even though these types aren’t part of this proposal, they may still be useful in the context of concurrency. Adding them later is entirely reasonable. The could be trivially implemented as wrappers around channels. Alternatively they could be implemented for efficiency at the same low level as channels.

Return a lock from send()

When sending an object through a channel, you don’t have a way of knowing when the object gets received on the other end. One way to work around this is to return a locked threading.Lock from SendChannel.send() that unlocks once the object is received.

Alternately, the proposed SendChannel.send() (blocking) and SendChannel.send_nowait() provide an explicit distinction that is less likely to confuse users.

Note that returning a lock would matter for buffered channels (i.e. queues). For unbuffered channels it is a non-issue.

Support prioritization in channels

A simple example is queue.PriorityQueue in the stdlib.

Support inheriting settings (and more?)

Folks might find it useful, when creating a new interpreter, to be able to indicate that they would like some things “inherited” by the new interpreter. The mechanism could be a strict copy or it could be copy-on-write. The motivating example is with the warnings module (e.g. copy the filters).

The feature isn’t critical, nor would it be widely useful, so it can wait until there’s interest. Notably, both suggested solutions will require significant work, especially when it comes to complex objects and most especially for mutable containers of mutable complex objects.

Make exceptions shareable

Exceptions are propagated out of run() calls, so it isn’t a big leap to make them shareable. However, as noted elsewhere, it isn’t essential or (particularly common) so we can wait on doing that.

Make everything shareable through serialization

We could use pickle (or marshal) to serialize everything and thus make them shareable. Doing this is potentially inefficient, but it may be a matter of convenience in the end. We can add it later, but trying to remove it later would be significantly more painful.

Make RunFailedError.__cause__ lazy

An uncaught exception in a subinterpreter (from interp.exec()) is copied to the calling interpreter and set as __cause__ on a RunFailedError which is then raised. That copying part involves some sort of deserialization in the calling interpreter, which can be expensive (e.g. due to imports) yet is not always necessary.

So it may be useful to use an ExceptionProxy type to wrap the serialized exception and only deserialize it when needed. That could be via ExceptionProxy__getattribute__() or perhaps through RunFailedError.resolve() (which would raise the deserialized exception and set RunFailedError.__cause__ to the exception.

It may also make sense to have RunFailedError.__cause__ be a descriptor that does the lazy deserialization (and set __cause__) on the RunFailedError instance.

Return a value from interp.exec()

Currently interp.exec() always returns None. One idea is to return the return value from whatever the subinterpreter ran. However, for now it doesn’t make sense. The only thing folks can run is a string of code (i.e. a script). This is equivalent to PyRun_StringFlags(), exec(), or a module body. None of those “return” anything. We can revisit this once interp.exec() supports functions, etc.

Add a shareable synchronization primitive

This would be _threading.Lock (or something like it) where interpreters would actually share the underlying mutex. The main concern is that locks and isolated interpreters may not mix well (as learned in Go).

We can add this later if it proves desirable without much trouble.

Propagate SystemExit and KeyboardInterrupt Differently

The exception types that inherit from BaseException (aside from Exception) are usually treated specially. These types are: KeyboardInterrupt, SystemExit, and GeneratorExit. It may make sense to treat them specially when it comes to propagation from interp.exec(). Here are some options:

* propagate like normal via RunFailedError
* do not propagate (handle them somehow in the subinterpreter)
* propagate them directly (avoid RunFailedError)
* propagate them directly (set RunFailedError as __cause__)

We aren’t going to worry about handling them differently. Threads already ignore SystemExit, so for now we will follow that pattern.

Add an explicit release() and close() to channel end classes

It can be convenient to have an explicit way to close a channel against further global use. Likewise it could be useful to have an explicit way to release one of the channel ends relative to the current interpreter. Among other reasons, such a mechanism is useful for communicating overall state between interpreters without the extra boilerplate that passing objects through a channel directly would require.

The challenge is getting automatic release/close right without making it hard to understand. This is especially true when dealing with a non-empty channel. We should be able to get by without release/close for now.

Add SendChannel.send_buffer()

This method would allow no-copy sending of an object through a channel if it supports the PEP 3118 buffer protocol (e.g. memoryview).

Support for this is not fundamental to channels and can be added on later without much disruption.

Auto-run in a thread

The PEP proposes a hard separation between subinterpreters and threads: if you want to run in a thread you must create the thread yourself and call interp.exec() in it. However, it might be convenient if interp.exec() could do that for you, meaning there would be less boilerplate.

Furthermore, we anticipate that users will want to run in a thread much more often than not. So it would make sense to make this the default behavior. We would add a kw-only param “threaded” (default True) to interp.exec() to allow the run-in-the-current-thread operation.

Rejected Ideas

Explicit channel association

Interpreters are implicitly associated with channels upon recv() and send() calls. They are de-associated with release() calls. The alternative would be explicit methods. It would be either add_channel() and remove_channel() methods on Interpreter objects or something similar on channel objects.

In practice, this level of management shouldn’t be necessary for users. So adding more explicit support would only add clutter to the API.

Add an API based on pipes

A pipe would be a simplex FIFO between exactly two interpreters. For most use cases this would be sufficient. It could potentially simplify the implementation as well. However, it isn’t a big step to supporting a many-to-many simplex FIFO via channels. Also, with pipes the API ends up being slightly more complicated, requiring naming the pipes.

Add an API based on queues

Queues and buffered channels are almost the same thing. The main difference is that channels have a stronger relationship with context (i.e. the associated interpreter).

The name “Channel” was used instead of “Queue” to avoid confusion with the stdlib queue.Queue.

“enumerate”

The list_all() function provides the list of all interpreters. In the threading module, which partly inspired the proposed API, the function is called enumerate(). The name is different here to avoid confusing Python users that are not already familiar with the threading API. For them “enumerate” is rather unclear, whereas “list_all” is clear.

Alternate solutions to prevent leaking exceptions across interpreters

In function calls, uncaught exceptions propagate to the calling frame. The same approach could be taken with interp.exec(). However, this would mean that exception objects would leak across the inter-interpreter boundary. Likewise, the frames in the traceback would potentially leak.

While that might not be a problem currently, it would be a problem once interpreters get better isolation relative to memory management (which is necessary to stop sharing the GIL between interpreters). We’ve resolved the semantics of how the exceptions propagate by raising a RunFailedError instead, for which __cause__ wraps a safe proxy for the original exception and traceback.

Rejected possible solutions:

  • reproduce the exception and traceback in the original interpreter and raise that.
  • raise a subclass of RunFailedError that proxies the original exception and traceback.
  • raise RuntimeError instead of RunFailedError
  • convert at the boundary (a la subprocess.CalledProcessError) (requires a cross-interpreter representation)
  • support customization via Interpreter.excepthook (requires a cross-interpreter representation)
  • wrap in a proxy at the boundary (including with support for something like err.raise() to propagate the traceback).
  • return the exception (or its proxy) from interp.exec() instead of raising it
  • return a result object (like subprocess does) [result-object] (unnecessary complexity?)
  • throw the exception away and expect users to deal with unhandled exceptions explicitly in the script they pass to interp.exec() (they can pass error info out via channels); with threads you have to do something similar

Always associate each new interpreter with its own thread

As implemented in the C-API, an interpreter is not inherently tied to any thread. Furthermore, it will run in any existing thread, whether created by Python or not. You only have to activate one of its thread states (PyThreadState) in the thread first. This means that the same thread may run more than one interpreter (though obviously not at the same time).

The proposed module maintains this behavior. Interpreters are not tied to threads. Only calls to Interpreter.exec() are. However, one of the key objectives of this PEP is to provide a more human-centric concurrency model. With that in mind, from a conceptual standpoint the module might be easier to understand if each interpreter were associated with its own thread.

That would mean interpreters.create() would create a new thread and Interpreter.exec() would only execute in that thread (and nothing else would). The benefit is that users would not have to wrap Interpreter.exec() calls in a new threading.Thread. Nor would they be in a position to accidentally pause the current interpreter (in the current thread) while their interpreter executes.

The idea is rejected because the benefit is small and the cost is high. The difference from the capability in the C-API would be potentially confusing. The implicit creation of threads is magical. The early creation of threads is potentially wasteful. The inability to run arbitrary interpreters in an existing thread would prevent some valid use cases, frustrating users. Tying interpreters to threads would require extra runtime modifications. It would also make the module’s implementation overly complicated. Finally, it might not even make the module easier to understand.

Only associate interpreters upon use

Associate interpreters with channel ends only once recv(), send(), etc. are called.

Doing this is potentially confusing and also can lead to unexpected races where a channel is auto-closed before it can be used in the original (creating) interpreter.

Allow multiple simultaneous calls to Interpreter.exec()

This would make sense especially if Interpreter.exec() were to manage new threads for you (which we’ve rejected). Essentially, each call would run independently, which would be mostly fine from a narrow technical standpoint, since each interpreter can have multiple threads.

The problem is that the interpreter has only one __main__ module and simultaneous Interpreter.exec() calls would have to sort out sharing __main__ or we’d have to invent a new mechanism. Neither would be simple enough to be worth doing.

Add a “reraise” method to RunFailedError

While having __cause__ set on RunFailedError helps produce a more useful traceback, it’s less helpful when handling the original error. To help facilitate this, we could add RunFailedError.reraise(). This method would enable the following pattern:

try:
    try:
        interp.exec(script)
    except RunFailedError as exc:
        exc.reraise()
except MyException:
    ...

This would be made even simpler if there existed a __reraise__ protocol.

All that said, this is completely unnecessary. Using __cause__ is good enough:

try:
    try:
        interp.exec(script)
    except RunFailedError as exc:
        raise exc.__cause__
except MyException:
    ...

Note that in extreme cases it may require a little extra boilerplate:

try:
    try:
        interp.exec(script)
    except RunFailedError as exc:
        if exc.__cause__ is not None:
            raise exc.__cause__
        raise  # re-raise
except MyException:
    ...

Implementation

The implementation of the PEP has 4 parts:

  • the high-level module described in this PEP (mostly a light wrapper around a low-level C extension
  • the low-level C extension module
  • additions to the internal C-API needed by the low-level module
  • secondary fixes/changes in the CPython runtime that facilitate the low-level module (among other benefits)

These are at various levels of completion, with more done the lower you go:

  • the high-level module has been, at best, roughly implemented. However, fully implementing it will be almost trivial.
  • the low-level module is mostly complete. The bulk of the implementation was merged into master in December 2018 as the “_xxsubinterpreters” module (for the sake of testing multiple interpreters functionality). Only the exception propagation implementation remains to be finished, which will not require extensive work.
  • all necessary C-API work has been finished
  • all anticipated work in the runtime has been finished

The implementation effort for PEP 554 is being tracked as part of a larger project aimed at improving multi-core support in CPython. [multi-core-project]

References


Source: https://github.com/python/peps/blob/main/peps/pep-0554.rst

Last modified: 2023-11-28 02:32:35 GMT