Since the code for rule NR2 is already here, to support CJK unified
ideographs and Tangut ideographs, it can also be used for other names
derived by rule NR2.
Add TYPE_FROZENDICT to the marshal module.
Add C API functions:
* PyAnyDict_Check()
* PyAnyDict_CheckExact()
* PyFrozenDict_Check()
* PyFrozenDict_CheckExact()
* PyFrozenDict_New()
Add PyFrozenDict_Type C type.
Co-authored-by: Hugo van Kemenade <[email protected]>
Co-authored-by: Adam Johnson <[email protected]>
Co-authored-by: Benedikt Johannes <[email protected]>
Add `_Py_type_getattro_stackref`, a variant of type attribute lookup
that returns `_PyStackRef` instead of `PyObject*`. This allows returning
deferred references in the free-threaded build, reducing reference count
contention when accessing type attributes.
This significantly improves scaling of namedtuple instantiation across
multiple threads.
* Add blurb
* Rename PyObject_GetAttrStackRef to _PyObject_GetAttrStackRef
* Apply suggestion from @vstinner
Co-authored-by: Victor Stinner <[email protected]>
* Apply suggestion from @vstinner
Co-authored-by: Victor Stinner <[email protected]>
* format
* Update Include/internal/pycore_function.h
Co-authored-by: Victor Stinner <[email protected]>
---------
Co-authored-by: Victor Stinner <[email protected]>
Add `gi_state`, `cr_state`, and `ag_state` attributes to generators,
coroutines, and async generators respectively. These attributes return the
current state as a string (e.g., `GEN_RUNNING`, `CORO_SUSPENDED`).
The `inspect.getgeneratorstate()`, `inspect.getcoroutinestate()`, and
`inspect.getasyncgenstate()` functions now return these attributes directly.
This is in preparation for making `gi_frame` thread-safe, which may involve
stop-the-world synchronization. The new state attributes avoid potential
performance cliffs in `inspect.getgeneratorstate()` and similar functions by
not requiring frame access.
Also removes unused `FRAME_COMPLETED` state and renumbers the frame state enum
to start at 0 instead of -1.
Instead of long and uninteresting output for all checked libraries, only print found issues by default.
Add a new -v/--verbose option to list all symbols (useful for checking that the script finds the symbols).
If we are specializing to `LOAD_GLOBAL_MODULE` or `LOAD_ATTR_MODULE`, try
to enable deferred reference counting for the value, if the object is owned by
a different thread. This applies to the free-threaded build only and should
improve scaling of multi-threaded programs.
Add the unicodedata.iter_graphemes() function to iterate over grapheme
clusters according to rules defined in Unicode Standard Annex #29.
Add unicodedata.grapheme_cluster_break(), unicodedata.indic_conjunct_break()
and unicodedata.extended_pictographic() functions to get the properties
of the character which are related to the above algorithm.
Co-authored-by: Guillaume "Vermeille" Sanchez <[email protected]>
The behaviour of Cut in nested parentheses, Repeat, Opt, and similar
is somewhat chaotic. Apparently even the academic papers on PEG aren't
as clear as they could be.
And it doesn't really matter. Python only uses top-level cuts.
When that changes, we can clarify as much as necessary (and even
change the implementation to make sense for what we'll need).
Document that this is deliberately unspecified, and add a test to
make sure any decision is deliberate, tested and documented.
- Gather all documented names into a set in a single pass.
This makes the check much faster.
- Do not match substrings (e.g. documenting `PyErr_WarnEx`
doesn't mean that `PyErr_Warn` is documented)
- Consider `PY`-prefixed names (a lot of old macros use this)
In optimized and `-Og` builds, arguments and local variables are frequently
unavailable in gdb. This makes `py-bt` fail to print anything useful. Use the
`PyThreadState*` pointers `_Py_tss_gilstate` and `Py_tss_tstate` to find the
interpreter frame if we can't get the frame from the
`_PyEval_EvalFrameDefault` call.
Co-authored-by: Victor Stinner <[email protected]>
The PyObject header reference count fields must be initialized using
atomic operations because they may be concurrently read by another
thread (e.g., from `_Py_TryIncref`).
The use of memmove and _Py_memory_repeat were not thread-safe in the
free threading build in some cases. In theory, memmove and
_Py_memory_repeat can copy byte-by-byte instead of pointer-by-pointer,
so concurrent readers could see uninitialized data or tearing.
Additionally, we should be using "release" (or stronger) ordering to be
compliant with the C11 memory model when copying objects within a list.
This makes generator frame state transitions atomic in the free
threading build, which avoids segfaults when trying to execute
a generator from multiple threads concurrently.
There are still a few operations that aren't thread-safe and may crash
if performed concurrently on the same generator/coroutine:
* Accessing gi_yieldfrom/cr_await/ag_await
* Accessing gi_frame/cr_frame/ag_frame
* Async generator operations
Now that we specialize range iteration in the interpreter for the common
case where the iterator has only one reference, there's not a
significant performance cost to making the iteration thread-safe.
