Repository navigation
memoryview & ctypes: incorrect itemsize for empty array #76963
Description
| BPO | 32782 |
|---|---|
| Nosy | @terryjreedy, @amauryfa, @abalkin, @skrah, @meadori, @eric-wieser |
| PRs |
Note: these values reflect the state of the issue at the time it was migrated and might not reflect the current state.
Show more details
GitHub fields:
assignee = None
closed_at = None
created_at = <Date 2018-02-06.18:53:21.025>
labels = ['interpreter-core', 'type-bug', '3.8', '3.9', '3.10', 'ctypes']
title = 'memoryview & ctypes: incorrect itemsize for empty array'
updated_at = <Date 2020-05-23.15:47:49.464>
user = 'https://github.com/eric-wieser'bugs.python.org fields:
activity = <Date 2020-05-23.15:47:49.464>
actor = 'cheryl.sabella'
assignee = 'none'
closed = False
closed_date = None
closer = None
components = ['Interpreter Core', 'ctypes']
creation = <Date 2018-02-06.18:53:21.025>
creator = 'Eric.Wieser'
dependencies = []
files = []
hgrepos = []
issue_num = 32782
keywords = ['patch']
message_count = 9.0
messages = ['311740', '321290', '324705', '340230', '340231', '340232', '340235', '340236', '348004']
nosy_count = 8.0
nosy_names = ['terry.reedy', 'teoliphant', 'amaury.forgeotdarc', 'belopolsky', 'skrah', 'meador.inge', 'Eric.Wieser', 'Eric Wieser']
pr_nums = ['5576']
priority = 'normal'
resolution = None
stage = 'patch review'
status = 'open'
superseder = None
type = 'behavior'
url = 'https://bugs.python.org/issue32782'
versions = ['Python 3.8', 'Python 3.9', 'Python 3.10']
Activity
eric-wieser commentedon Feb 6, 2018
Take the following simple structure:
And construct some arrays with it:
In most cases, this works as expected, returning the size of one item:
>>> get_array_view(10).itemsize 4 >>> get_array_view(1).itemsize 4But when N=0, it returns the wrong result
>>> get_array_view(0).itemsize 0Which contradicts its
.format, which still describes a 4-byte struct>>> get_array_view(0).format 'T{>I:one:}'This causes a downstream problem in numpy:
>>> np.array(get_array_view(0)) RuntimeWarning: Item size computed from the PEP 3118 buffer format string does not match the actual item size.eric-wieser commentedon Jul 8, 2018
Pinging, as recommended by https://devguide.python.org/pullrequest/#reviewing. Ideally this and https://bugs.python.org/issue32780 would make the same patch release.
eric-wieser commentedon Sep 6, 2018
Pinging again, for lack of a clearer path forward
terryjreedy commentedon Apr 14, 2019
This issue is about the itemsize attribute of instances of the built-in memoryview class. Ctypes in only involved in providing format information. Hence the nosy additions.
On Win10 with 3.8, ctypes has no uint attributes. Using 'c_int32' instead, I see the same behavior (itemsize 0 for empty structure).
About itemsize, https://www.python.org/dev/peps/pep-3118/ says
"This is a storage for the itemsize (in bytes) of each element of the shared memory. It is technically un-necessary as it can be obtained using PyBuffer_SizeFromFormat, however an exporter may know this information without parsing the format string and it is necessary to know the itemsize for proper interpretation of striding. Therefore, storing it is more convenient and faster."
The first line could be seen as implying that itemsize is undefined if there are no items (and as justifying numbytes/numitems otherwise). The 0 return could be seen as equivalent to a None return from a python-coded function. If so, it is not a bug, and there might be code that would break if it is changed.
On the other hand, the next lines imply that itemsize is *usually*, though not necessarily, a cache for PyBuffer_SizeFromFormat. This could be seen as implying that in the absence of other information, the itemsize should be calculated from the format, making 0 a bug.
[-]ctypes: memoryview gives incorrect PEP3118 itemsize for array of length zero[/-][+]memoryview gives incorrect PEP3118 itemsize for empty array[/+]terryjreedy commentedon Apr 14, 2019
This is actually about memoryview.itemsize within ctypes.
[-]memoryview gives incorrect PEP3118 itemsize for empty array[/-][+]memoryview & ctypes: incorrect PEP3118 itemsize for empty array[/+]terryjreedy commentedon Apr 14, 2019
[-]memoryview & ctypes: incorrect PEP3118 itemsize for empty array[/-][+]memoryview & ctypes: incorrect itemsize for empty array[/+]4 remaining items
gh-76963: PEP3118 itemsize of an empty ctypes array should not be 0 (G…
pythongh-76963: PEP3118 itemsize of an empty ctypes array should not …
pythongh-76963: PEP3118 itemsize of an empty ctypes array should not …
[3.10] gh-76963: PEP3118 itemsize of an empty ctypes array should not…
[3.11] gh-76963: PEP3118 itemsize of an empty ctypes array should not…
mattip commentedon Dec 23, 2022
Thanks for the fix that closed this. @eric-wieser do you remember where this impacted NumPy? I don't see the relevant issue or PR.
eric-wieser commentedon Dec 23, 2022
I think the workaround in numpy needs to stay till after #5561
eric-wieser commentedon Dec 23, 2022
The reference in numpy is here:
https://github.com/numpy/numpy/blob/85b26f4afc6e98694eab7a6d6abefbd3fed31fde/numpy/core/_dtype_ctypes.py#L21
Merged main into regmachine branch (#50)