Issue #23935: Argument Clinic's understanding of format units
accepting bytes, bytearrays, and buffers is now consistent with
both the documentation and the implementation.
diff --git a/Misc/NEWS b/Misc/NEWS
index 0074417..3ba17a4 100644
--- a/Misc/NEWS
+++ b/Misc/NEWS
@@ -187,6 +187,10 @@
Tools/Demos
-----------
+- Issue #23935: Argument Clinic's understanding of format units
+ accepting bytes, bytearrays, and buffers is now consistent with
+ both the documentation and the implementation.
+
- Issue #23944: Argument Clinic now wraps long impl prototypes at column 78.
- Issue #20586: Argument Clinic now ensures that functions without docstrings
diff --git a/Modules/_dbmmodule.c b/Modules/_dbmmodule.c
index bcdea1d..b815e97 100644
--- a/Modules/_dbmmodule.c
+++ b/Modules/_dbmmodule.c
@@ -272,7 +272,7 @@
self: dbmobject
- key: str(length=True)
+ key: str(types={'str', 'robuffer'}, length=True)
default: object = None
/
@@ -282,7 +282,7 @@
static PyObject *
dbm_dbm_get_impl(dbmobject *dp, const char *key, Py_ssize_clean_t key_length,
PyObject *default_value)
-/*[clinic end generated code: output=4f5c0e523eaf1251 input=aecf5efd2f2b1a3b]*/
+/*[clinic end generated code: output=4f5c0e523eaf1251 input=f81478bc211895ef]*/
{
datum dbm_key, val;
diff --git a/Modules/arraymodule.c b/Modules/arraymodule.c
index 49f837c..4704490 100644
--- a/Modules/arraymodule.c
+++ b/Modules/arraymodule.c
@@ -1600,7 +1600,7 @@
/*[clinic input]
array.array.fromstring
- buffer: Py_buffer(types='str bytes bytearray buffer')
+ buffer: Py_buffer(types={'str', 'buffer'})
/
Appends items from the string, interpreting it as an array of machine values, as if it had been read from a file using the fromfile() method).
@@ -1610,7 +1610,7 @@
static PyObject *
array_array_fromstring_impl(arrayobject *self, Py_buffer *buffer)
-/*[clinic end generated code: output=31c4baa779df84ce input=1302d94c97696b84]*/
+/*[clinic end generated code: output=31c4baa779df84ce input=fdde1a56cbe2b05b]*/
{
if (PyErr_WarnEx(PyExc_DeprecationWarning,
"fromstring() is deprecated. Use frombytes() instead.", 2) != 0)
@@ -1929,7 +1929,7 @@
array._array_reconstructor
arraytype: object(type="PyTypeObject *")
- typecode: int(types='str')
+ typecode: int(types={'str'})
mformat_code: int(type="enum machine_format_code")
items: object
/
@@ -1942,7 +1942,7 @@
int typecode,
enum machine_format_code mformat_code,
PyObject *items)
-/*[clinic end generated code: output=6ecbf0e8e4d92ab9 input=f72492708c0a1d50]*/
+/*[clinic end generated code: output=6ecbf0e8e4d92ab9 input=a9ae223306d7b262]*/
{
PyObject *converted_items;
PyObject *result;
diff --git a/Tools/clinic/clinic.py b/Tools/clinic/clinic.py
index 8a2e601..ba7cb88 100755
--- a/Tools/clinic/clinic.py
+++ b/Tools/clinic/clinic.py
@@ -2493,12 +2493,12 @@
class char_converter(CConverter):
type = 'char'
- default_type = str
+ default_type = (bytes, bytearray)
format_unit = 'c'
c_ignored_default = "'\0'"
def converter_init(self):
- if isinstance(self.default, str) and (len(self.default) != 1):
+ if isinstance(self.default, self.default_type) and (len(self.default) != 1):
fail("char_converter: illegal default value " + repr(self.default))
@@ -2531,18 +2531,18 @@
if not bitwise:
fail("Unsigned shorts must be bitwise (for now).")
-@add_legacy_c_converter('C', types='str')
+@add_legacy_c_converter('C', types={'str'})
class int_converter(CConverter):
type = 'int'
default_type = int
format_unit = 'i'
c_ignored_default = "0"
- def converter_init(self, *, types='int', type=None):
- if types == 'str':
+ def converter_init(self, *, types={'int'}, type=None):
+ if types == {'str'}:
self.format_unit = 'C'
- elif types != 'int':
- fail("int_converter: illegal 'types' argument")
+ elif types != {'int'}:
+ fail("int_converter: illegal 'types' argument " + repr(types))
if type != None:
self.type = type
@@ -2633,63 +2633,64 @@
self.type = type
-@add_legacy_c_converter('s#', length=True)
-@add_legacy_c_converter('y', types="bytes")
-@add_legacy_c_converter('y#', types="bytes", length=True)
+#
+# We define three string conventions for buffer types in the 'types' argument:
+# 'buffer' : any object supporting the buffer interface
+# 'rwbuffer': any object supporting the buffer interface, but must be writeable
+# 'robuffer': any object supporting the buffer interface, but must not be writeable
+#
+
+@add_legacy_c_converter('s#', types={"str", "robuffer"}, length=True)
+@add_legacy_c_converter('y', types={"robuffer"})
+@add_legacy_c_converter('y#', types={"robuffer"}, length=True)
@add_legacy_c_converter('z', nullable=True)
-@add_legacy_c_converter('z#', nullable=True, length=True)
+@add_legacy_c_converter('z#', types={"str", "robuffer"}, nullable=True, length=True)
+# add_legacy_c_converter not supported for es, es#, et, et#
+# because of their extra encoding argument
class str_converter(CConverter):
type = 'const char *'
default_type = (str, Null, NoneType)
format_unit = 's'
- def converter_init(self, *, encoding=None, types="str",
+ def converter_init(self, *, encoding=None, types={"str"},
length=False, nullable=False, zeroes=False):
- types = set(types.strip().split())
- bytes_type = {"bytes"}
- str_type = {"str"}
- all_3_type = {"bytearray"} | bytes_type | str_type
- is_bytes = types == bytes_type
- is_str = types == str_type
- is_all_3 = types == all_3_type
-
self.length = bool(length)
+
+ is_b_or_ba = types == {"bytes", "bytearray"}
+ is_str = types == {"str"}
+ is_robuffer = types == {"robuffer"}
+ is_str_or_robuffer = types == {"str", "robuffer"}
+
format_unit = None
if encoding:
self.encoding = encoding
- if is_str and not (length or zeroes or nullable):
+ if is_str and not length and not zeroes and not nullable:
format_unit = 'es'
- elif is_all_3 and not (length or zeroes or nullable):
- format_unit = 'et'
- elif is_str and length and zeroes and not nullable:
+ elif is_str and length and zeroes and nullable:
format_unit = 'es#'
- elif is_all_3 and length and not (nullable or zeroes):
+ elif is_b_or_ba and not length and not zeroes and not nullable:
+ format_unit = 'et'
+ elif is_b_or_ba and length and zeroes and nullable:
format_unit = 'et#'
- if format_unit.endswith('#'):
- fail("Sorry: code using format unit ", repr(format_unit), "probably doesn't work properly yet.\nGive Larry your test case and he'll it.")
- # TODO set pointer to NULL
- # TODO add cleanup for buffer
- pass
-
else:
if zeroes:
fail("str_converter: illegal combination of arguments (zeroes is only legal with an encoding)")
- if is_bytes and not (nullable or length):
- format_unit = 'y'
- elif is_bytes and length and not nullable:
- format_unit = 'y#'
- elif is_str and not (nullable or length):
+ if is_str and not length and not nullable:
format_unit = 's'
- elif is_str and length and not nullable:
- format_unit = 's#'
- elif is_str and nullable and not length:
+ elif is_str and not length and nullable:
format_unit = 'z'
- elif is_str and nullable and length:
+ elif is_robuffer and not length and not nullable:
+ format_unit = 'y'
+ elif is_robuffer and length and not nullable:
+ format_unit = 'y#'
+ elif is_str_or_robuffer and length and not nullable:
+ format_unit = 's#'
+ elif is_str_or_robuffer and length and nullable:
format_unit = 'z#'
if not format_unit:
@@ -2700,10 +2701,12 @@
class PyBytesObject_converter(CConverter):
type = 'PyBytesObject *'
format_unit = 'S'
+ # types = {'bytes'}
class PyByteArrayObject_converter(CConverter):
type = 'PyByteArrayObject *'
format_unit = 'Y'
+ # types = {'bytearray'}
class unicode_converter(CConverter):
type = 'PyObject *'
@@ -2725,43 +2728,29 @@
self.length = True
self.format_unit = format_unit
-#
-# We define three string conventions for buffer types in the 'types' argument:
-# 'buffer' : any object supporting the buffer interface
-# 'rwbuffer': any object supporting the buffer interface, but must be writeable
-# 'robuffer': any object supporting the buffer interface, but must not be writeable
-#
-@add_legacy_c_converter('s*', types='str bytes bytearray buffer')
-@add_legacy_c_converter('z*', types='str bytes bytearray buffer', nullable=True)
-@add_legacy_c_converter('w*', types='bytearray rwbuffer')
+@add_legacy_c_converter('s*', types={'str', 'buffer'})
+@add_legacy_c_converter('z*', types={'str', 'buffer'}, nullable=True)
+@add_legacy_c_converter('w*', types={'rwbuffer'})
class Py_buffer_converter(CConverter):
type = 'Py_buffer'
format_unit = 'y*'
impl_by_reference = True
c_ignored_default = "{NULL, NULL}"
- def converter_init(self, *, types='bytes bytearray buffer', nullable=False):
+ def converter_init(self, *, types={'buffer'}, nullable=False):
if self.default not in (unspecified, None):
fail("The only legal default value for Py_buffer is None.")
self.c_default = self.c_ignored_default
- types = set(types.strip().split())
- bytes_type = {'bytes'}
- bytearray_type = {'bytearray'}
- buffer_type = {'buffer'}
- rwbuffer_type = {'rwbuffer'}
- robuffer_type = {'robuffer'}
- str_type = {'str'}
- bytes_bytearray_buffer_type = bytes_type | bytearray_type | buffer_type
format_unit = None
- if types == (str_type | bytes_bytearray_buffer_type):
+ if types == {'str', 'buffer'}:
format_unit = 's*' if not nullable else 'z*'
else:
if nullable:
fail('Py_buffer_converter: illegal combination of arguments (nullable=True)')
- elif types == (bytes_bytearray_buffer_type):
+ elif types == {'buffer'}:
format_unit = 'y*'
- elif types == (bytearray_type | rwbuffer_type):
+ elif types == {'rwbuffer'}:
format_unit = 'w*'
if not format_unit:
fail("Py_buffer_converter: illegal combination of arguments")