Description

sobolevn

opened on Oct 6, 2022

Issue body actions

Feature or enhancement

Right now PyUnicode_Count from

cpython/Objects/unicodeobject.c

Lines 8968 to 9040 in cbdeda8

	Py_ssize_t
	PyUnicode_Count(PyObject *str,
	PyObject *substr,
	Py_ssize_t start,
	Py_ssize_t end)
	{
	Py_ssize_t result;
	int kind1, kind2;
	const void *buf1 = NULL, *buf2 = NULL;
	Py_ssize_t len1, len2;
	if (ensure_unicode(str) < 0 || ensure_unicode(substr) < 0)
	return -1;
	kind1 = PyUnicode_KIND(str);
	kind2 = PyUnicode_KIND(substr);
	if (kind1 < kind2)
	return 0;
	len1 = PyUnicode_GET_LENGTH(str);
	len2 = PyUnicode_GET_LENGTH(substr);
	ADJUST_INDICES(start, end, len1);
	if (end - start < len2)
	return 0;
	buf1 = PyUnicode_DATA(str);
	buf2 = PyUnicode_DATA(substr);
	if (kind2 != kind1) {
	buf2 = unicode_askind(kind2, buf2, len2, kind1);
	if (!buf2)
	goto onError;
	}
	switch (kind1) {
	case PyUnicode_1BYTE_KIND:
	if (PyUnicode_IS_ASCII(str) && PyUnicode_IS_ASCII(substr))
	result = asciilib_count(
	((const Py_UCS1*)buf1) + start, end - start,
	buf2, len2, PY_SSIZE_T_MAX
	);
	else
	result = ucs1lib_count(
	((const Py_UCS1*)buf1) + start, end - start,
	buf2, len2, PY_SSIZE_T_MAX
	);
	break;
	case PyUnicode_2BYTE_KIND:
	result = ucs2lib_count(
	((const Py_UCS2*)buf1) + start, end - start,
	buf2, len2, PY_SSIZE_T_MAX
	);
	break;
	case PyUnicode_4BYTE_KIND:
	result = ucs4lib_count(
	((const Py_UCS4*)buf1) + start, end - start,
	buf2, len2, PY_SSIZE_T_MAX
	);
	break;
	default:
	Py_UNREACHABLE();
	}
	assert((kind2 != kind1) == (buf2 != PyUnicode_DATA(substr)));
	if (kind2 != kind1)
	PyMem_Free((void *)buf2);
	return result;
	onError:
	assert((kind2 != kind1) == (buf2 != PyUnicode_DATA(substr)));
	if (kind2 != kind1)
	PyMem_Free((void *)buf2);
	return -1;
	}

and unicode_count from

cpython/Objects/unicodeobject.c

Lines 10854 to 10916 in cbdeda8

	static PyObject *
	unicode_count(PyObject *self, PyObject *args)
	{
	PyObject *substring = NULL; /* initialize to fix a compiler warning */
	Py_ssize_t start = 0;
	Py_ssize_t end = PY_SSIZE_T_MAX;
	PyObject *result;
	int kind1, kind2;
	const void *buf1, *buf2;
	Py_ssize_t len1, len2, iresult;
	if (!parse_args_finds_unicode("count", args, &substring, &start, &end))
	return NULL;
	kind1 = PyUnicode_KIND(self);
	kind2 = PyUnicode_KIND(substring);
	if (kind1 < kind2)
	return PyLong_FromLong(0);
	len1 = PyUnicode_GET_LENGTH(self);
	len2 = PyUnicode_GET_LENGTH(substring);
	ADJUST_INDICES(start, end, len1);
	if (end - start < len2)
	return PyLong_FromLong(0);
	buf1 = PyUnicode_DATA(self);
	buf2 = PyUnicode_DATA(substring);
	if (kind2 != kind1) {
	buf2 = unicode_askind(kind2, buf2, len2, kind1);
	if (!buf2)
	return NULL;
	}
	switch (kind1) {
	case PyUnicode_1BYTE_KIND:
	iresult = ucs1lib_count(
	((const Py_UCS1*)buf1) + start, end - start,
	buf2, len2, PY_SSIZE_T_MAX
	);
	break;
	case PyUnicode_2BYTE_KIND:
	iresult = ucs2lib_count(
	((const Py_UCS2*)buf1) + start, end - start,
	buf2, len2, PY_SSIZE_T_MAX
	);
	break;
	case PyUnicode_4BYTE_KIND:
	iresult = ucs4lib_count(
	((const Py_UCS4*)buf1) + start, end - start,
	buf2, len2, PY_SSIZE_T_MAX
	);
	break;
	default:
	Py_UNREACHABLE();
	}
	result = PyLong_FromSsize_t(iresult);
	assert((kind2 == kind1) == (buf2 == PyUnicode_DATA(substring)));
	if (kind2 != kind1)
	PyMem_Free((void *)buf2);
	return result;
	}

share a lot of code.

They can be unified, because the do the same thing.

Pitch

Citing @encukou:

Apparently unicode_count missed an optimization in 2011, otherwise they're equivalent (except arg parsing & converting the return value). Merging them could add the optimization to unicode_count.
If you want to work on that, note that there's also anylib_count that duplicates the main switch.

Previous discussion

Link: #96929

PR in the works.

Activity

1. sobolevn

   added

   type-featureA feature request or enhancementA feature request or enhancement

   interpreter-core(Objects, Python, Grammar, and Parser dirs)(Objects, Python, Grammar, and Parser dirs)

   on Oct 6, 2022
2. sobolevn

   self-assigned this

   on Oct 6, 2022
3. sobolevn

   added a commit that references this issue on Oct 7, 2022

   pythongh-97982: Reuse `PyUnicode_Count` in `unicode_count`

   b239ae5
4. vstinner

   added a commit that references this issue on Oct 11, 2022

   gh-97982: Remove asciilib_count() (#98164)

   df3a6d9
5. vstinner commented on Oct 11, 2022

   

   vstinner

   on Oct 11, 2022

   Member

   More actions

   I remove asciilib_count() function in commit df3a6d9.
6. vstinner

   added a commit that references this issue on Oct 12, 2022

   gh-97982: Factorize PyUnicode_Count() and unicode_count() code (#98025)

   ccab67b
7. vstinner commented on Oct 12, 2022

   

   vstinner

   on Oct 12, 2022

   Member

   More actions

   Fixed by ccab67b

   Reacted by Sergey Silaev
8. vstinner

   closed this as completedon Oct 12, 2022
9. carljm

   added 2 commits that reference this issue on Oct 14, 2022

   Merge branch 'main' into typewatch

   6e53514

   Merge branch 'main' into cached_property

   0bb0c69

Issue actions