Use Argument Clinic in sqlite3.

What's the purpose of using AC, did the change improve performance?
It can make hard to track the code history.

The primary reason is that it will be provide easy access to module state.

The first step in making sqlite3 support multiphase init was to create heap types. The second step is argument clinic. The third will be to use AC for module state. The last step will then be final multiphase support.

Also, IMHO, AC greatly improves the body of methods (readability => maintainability, hardened parsing of arguments)

Also, see the comment from Victor here: https://github.com/python/cpython/pull/22478#issuecomment-702201260

> Also, see the comment from Victor here: https://github.com/python/cpython/pull/22478#issuecomment-702201260

Okay got it

By the way, what's the preferred way to benchmark performance?

New changeset 7d210271579ae31f43b32f73c2aff5bc4fe0d27f by Erlend Egeberg Aasland in branch 'master':
bpo-40956: Convert _sqlite3 module level functions to Argument Clinic (GH-22484)
https://github.com/python/cpython/commit/7d210271579ae31f43b32f73c2aff5bc4fe0d27f

@erlendaasland

Thank you erlendaasland for working on GH-22484.
This time was a good time to review my AC knowledge ;)

> Thank you erlendaasland for working on GH-22484.

… and thank you for taking the time to review this!

New changeset 1ba82bbc50a52f40ad05f3c4aaf2e159e0ce126d by Erlend Egeberg Aasland in branch 'master':
bpo-40956: Convert _sqlite3.Connection to Argument Clinic (GH-23341)
https://github.com/python/cpython/commit/1ba82bbc50a52f40ad05f3c4aaf2e159e0ce126d

New changeset 2179349d8cf45b1202775547df384b1fde31630a by Dong-hee Na in branch 'master':
bpo-40956: Fix sqlite3 AC code (GH-23837)
https://github.com/python/cpython/commit/2179349d8cf45b1202775547df384b1fde31630a

New changeset 3ccef1ca474592e191a00e131dfbaf777db271e9 by Erlend Egeberg Aasland in branch 'master':
bpo-40956: Convert _sqlite3.Connection to Argument Clinic, part 2 (GH-23838)
https://github.com/python/cpython/commit/3ccef1ca474592e191a00e131dfbaf777db271e9

New changeset 84d79cfda947f6bc28a5aa11db8055aa40a6b03a by Erlend Egeberg Aasland in branch 'master':
bpo-40956: Convert _sqlite3.Row to Argument Clinic (GH-23964)
https://github.com/python/cpython/commit/84d79cfda947f6bc28a5aa11db8055aa40a6b03a

New changeset c7f8d3caf0b10c19cd46b1b334a98628eac15672 by Erlend Egeberg Aasland in branch 'master':
bpo-40956: Convert _sqlite3.Cursor to Argument Clinic (GH-24007)
https://github.com/python/cpython/commit/c7f8d3caf0b10c19cd46b1b334a98628eac15672

GH-24007 introduced a regression:

>>> import sqlite3
>>> help(sqlite3)
Traceback (most recent call last):
  File "Lib/inspect.py", line 2049, in wrap_value
    value = eval(s, module_dict)
  File "<string>", line 1, in <module>
NameError: name 'cursor' is not defined

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "Lib/inspect.py", line 2052, in wrap_value
    value = eval(s, sys_module_dict)
  File "<string>", line 1, in <module>
NameError: name 'cursor' is not defined

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "Lib/_sitebuiltins.py", line 103, in __call__
    return pydoc.help(*args, **kwds)
  File "Lib/pydoc.py", line 2000, in __call__
    self.help(request)
  File "Lib/pydoc.py", line 2059, in help
    else: doc(request, 'Help on %s:', output=self._output)
  File "Lib/pydoc.py", line 1779, in doc
    pager(render_doc(thing, title, forceload))
  File "Lib/pydoc.py", line 1772, in render_doc
    return title % desc + '\n\n' + renderer.document(object, name)
  File "Lib/pydoc.py", line 472, in document
    if inspect.ismodule(object): return self.docmodule(*args)
  File "Lib/pydoc.py", line 1267, in docmodule
    contents.append(self.document(value, key, name))
  File "Lib/pydoc.py", line 473, in document
    if inspect.isclass(object): return self.docclass(*args)
  File "Lib/pydoc.py", line 1433, in docclass
    attrs = spill("Methods %s:\n" % tag, attrs,
  File "Lib/pydoc.py", line 1382, in spill
    push(self.document(value,
  File "Lib/pydoc.py", line 474, in document
    if inspect.isroutine(object): return self.docroutine(*args)
  File "Lib/pydoc.py", line 1492, in docroutine
    signature = inspect.signature(object)
  File "Lib/inspect.py", line 3129, in signature
    return Signature.from_callable(obj, follow_wrapped=follow_wrapped,
  File "Lib/inspect.py", line 2877, in from_callable
    return _signature_from_callable(obj, sigcls=cls,
  File "Lib/inspect.py", line 2351, in _signature_from_callable
    return _signature_from_builtin(sigcls, obj,
  File "Lib/inspect.py", line 2164, in _signature_from_builtin
    return _signature_fromstr(cls, func, s, skip_bound_arg)
  File "Lib/inspect.py", line 2102, in _signature_fromstr
    p(name, default)
  File "Lib/inspect.py", line 2084, in p
    default_node = RewriteSymbolics().visit(default_node)
  File "Lib/ast.py", line 410, in visit
    return visitor(node)
  File "Lib/inspect.py", line 2071, in visit_Attribute
    return wrap_value(value)
  File "Lib/inspect.py", line 2054, in wrap_value
    raise RuntimeError()
RuntimeError

Looks like this is the culprit: https://github.com/python/cpython/blob/2396614b8958ad202378fd71a598eb4106ac5896/Modules/_sqlite/cursor.c#L825

I don't see how to easily solve this in AC: we want the arraysize of the cursor instance, but I don't see how I'm going to get this via the AC namespace/context. Using a "bogus" default value works:

size as maxrows: int(c_default='self->arraysize') = 1

Is there a better way?

Use NULL.

Thanks, but I prefer the following:

size as maxrows: int(c_default='self->arraysize', py_default='<unrepresentable>') = 1

New changeset a330365ca5ae836075f306334ab648bf23471481 by Erlend Egeberg Aasland in branch 'master':
bpo-40956: Fix sqlite3.Cursor.fetchmany() default value (GH-24214)
https://github.com/python/cpython/commit/a330365ca5ae836075f306334ab648bf23471481

New changeset ea46579067fd2d4e164d6605719ffec690c4d621 by Erlend Egeberg Aasland in branch 'master':
bpo-40956: Fix segfault when Connection.backup is called without target (GH-24503)
https://github.com/python/cpython/commit/ea46579067fd2d4e164d6605719ffec690c4d621

New changeset 185ecdc1463c527743eeb16a5deef75c1985de79 by Erlend Egeberg Aasland in branch 'main':
bpo-40956: Convert sqlite3.connect and sqlite3.Connection.__init__ to AC (GH-24421)
https://github.com/python/cpython/commit/185ecdc1463c527743eeb16a5deef75c1985de79

Erlend, is anything left in this issue?

> Erlend, is anything left in this issue?

Nothing left; thank you for your guidance and reviews, Dong-hee, Berker, Serhiy, and Pablo.

stage: patch review -> resolved