FazBrowse GitHub Viewer | Trending |
URL:
| Home
Tools: [Download Repo ZIP]   [Original HTTPS Page]

gh-154387: fix `structseq_repr` by labeling unnamed fields in repr as `<unnamed@N>` by XuehaiPan · Pull Request #154687 · python/cpython · GitHub

  1. python
  2. cpython

Repository navigation

gh-154387: fix structseq_repr by labeling unnamed fields in repr as <unnamed@N> - #154687

Open
XuehaiPan wants to merge 16 commits into
python:mainfrom
XuehaiPan:fix-structseq-unnamed
ConversationCommits (16)ChecksFiles changed
Open

gh-154387: fix structseq_repr by labeling unnamed fields in repr as <unnamed@N>#154687
XuehaiPan wants to merge 16 commits into
python:mainfrom
XuehaiPan:fix-structseq-unnamed

Conversation

XuehaiPan commented Jul 25, 2026 •
edited
Loading

Copy link
Copy Markdown
Contributor

Walk the members alongside the visible slots and label unnamed slots positionally as <unnamed@N>, which is correct regardless of where the unnamed fields sit among the visible ones.

Before this PR, the member slots are accessed incorrectly in repr when there are unnamed fields. It can cause a spurious SystemError when the index reaches the number of named fields (n_fields - n_unnamed_fields), or display the unnamed fields with <name>=<value> using a later field's name.

For example, a structseq type with WithUnnamed(first, <unnamed>, third, <unnamed>) + [hidden .fifth], before this PR:

>>> seq = WithUnnamed(range(5))
>>> repr(seq)
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
    repr(seq)
    ~~~~^^^^^
SystemError: In structseq_repr(), member 3 name is NULL for type WithUnnamed
# -> WithUnnamed(first=0, third=1, fifth=2, <reach NULL-terminated tp_members' end for member 3>)

After this PR:

>>> seq = WithUnnamed(range(5))
>>> seq
WithUnnamed(first=0, <unnamed@1>=1, third=2, <unnamed@3>=3)
>>> seq.fifth
4

Before this PR, for os.stat_result, the unnamed slots printed as the following field names st_atime/st_mtime/st_ctime.

>>> st = os.stat_result(range(os.stat_result.n_fields))
>>> st
os.stat_result(st_mode=0, st_ino=1, st_dev=2, st_nlink=3, st_uid=4, st_gid=5, st_size=6,
               st_atime=7, st_mtime=8, st_ctime=9)  # borrow the following fields' names but use the unnamed fields' values
>>> st[7]
7
>>> st.st_atime
10

>>> st = os.stat_result(tuple(range(10)) + (7.1, 8.2, 9.3))
>>> st
os.stat_result(st_mode=0, st_ino=1, st_dev=2, st_nlink=3, st_uid=4, st_gid=5, st_size=6,
               st_atime=7, st_mtime=8, st_ctime=9)
>>> st[7]        # index = 7 is int
7
>>> st.st_atime  # index = 10 is float
7.1

After this PR, a new custom statresult_repr is added analogously to statresult_new. The custom repr shows the fields 10-12 (always initialized in statresult_new) instead of generic <unnamed@N> (introduced in this PR).

>>> st = os.stat_result(range(os.stat_result.n_fields))
>>> st
os.stat_result(st_mode=0, st_ino=1, st_dev=2, st_nlink=3, st_uid=4, st_gid=5, st_size=6,
               st_atime=10, st_mtime=11, st_ctime=12)  # make the field names and values match
>>> st[7]
7
>>> st.st_atime
10

>>> st = os.stat_result(tuple(range(10)) + (7.1, 8.2, 9.3))
>>> st
os.stat_result(st_mode=0, st_ino=1, st_dev=2, st_nlink=3, st_uid=4, st_gid=5, st_size=6,
               st_atime=7.1, st_mtime=8.2, st_ctime=9.3)
>>> st[7]        # index = 7 is int
7
>>> st.st_atime  # index = 10 is float
7.1

>>> os.stat(os.curdir)
os.stat_result(st_mode=16877, st_ino=429294614, st_dev=16777232, st_nlink=58, st_uid=501, st_gid=20, st_size=1856,
               st_atime=1784996316.4868717, st_mtime=1784996316.472091, st_ctime=1784996316.472091)

Without the custom statresult_repr, the generic <unnamed@N> will be used, which is not that legible.

>>> st = os.stat_result(range(os.stat_result.n_fields))
>>> st
os.stat_result(st_mode=0, st_ino=1, st_dev=2, st_nlink=3, st_uid=4, st_gid=5, st_size=6,
               <unnamed@7>=7, <unnamed@8>=8, <unnamed@9>=9)

>>> os.stat(os.curdir)
os.stat_result(st_mode=16877, st_ino=429294614, st_dev=16777232, st_nlink=58, st_uid=501, st_gid=20, st_size=1856,
               <unnamed@7>=1784996316, <unnamed@8>=1784996316, <unnamed@9>=1784996316)

Fixes #154387

fix(structseq): label unnamed fields in repr as <unnamed@N>
34b2b44 
structseq_repr() labeled each visible slot by indexing tp_members by slot
position, but tp_members is packed with unnamed fields skipped, so every slot
at or after the first unnamed field borrowed the name of a later hidden field.
For os.stat_result the integer time slots printed as st_atime/st_mtime/st_ctime
(which are actually separate hidden float fields), so the repr disagreed with
both st[i] and the .st_atime attribute.

Walk the members alongside the visible slots and label unnamed slots
positionally as "<unnamed@N>", which is correct regardless of where the
unnamed fields sit among the visible ones.
bedevere-app Bot mentioned this pull request Jul 25, 2026
Open
XuehaiPan changed the title gh-154387: label unnamed fields in repr as <unnamed@N> gh-154387: fix structseq_repr by labeling unnamed fields in repr as <unnamed@N> Jul 25, 2026
XuehaiPan added 5 commits July 25, 2026 18:25
fix: fix missing import due to git rebase
23411b4
test(structseq): cover a hidden named field in the interspersed test
9b3f8fe 
Lower n_in_sequence to 4 in the _testcapi interspersed helper so the trailing
named field ("fifth") is hidden, and expand
test_repr_with_interspersed_unnamed_fields to assert the full surface: repr
(hidden field excluded), field counts, __match_args__, tuple()/len(), and
attribute access to visible and hidden fields (the latter both provided and
defaulted to None).
fix: remove test_ prefix for functions in _testcapi
227e089
feat(os): custom repr for os.stat_result showing time fields by name
50a2e22 
os.stat_result has three unnamed integer time slots (7-9) whose values are
mirrored by the named float fields st_atime/st_mtime/st_ctime (statresult_new
fills the latter from the former). The generic struct sequence repr shows
those unnamed slots as <unnamed@7..9>, which is uninformative.

Add statresult_repr showing the named fields (st_mode..st_size and the float
st_atime/st_mtime/st_ctime) by name, matching the attributes. Install it
through the type dict via PyObject_SetAttrString so repr() and st.__repr__()
resolve to the same function; a raw tp_repr assignment would leave the
__repr__ wrapper readied by PyStructSequence_NewType pointing at the generic
repr. Other struct sequences keep the generic <unnamed@N>.
Merge remote-tracking branch 'upstream/main' into fix-structseq-unnamed
59e2238
XuehaiPan marked this pull request as ready for review July 25, 2026 17:00
XuehaiPan commented Jul 29, 2026
View reviewed changes
Comment thread Lib/test/test_class.py Outdated
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters. Learn more about bidirectional Unicode characters
Show hidden characters
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

No reviews

Assignees

No one assigned

Labels

awaiting review

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

structseq_repr mislabels unnamed PyStructSequence fields (e.g. os.stat_result slots 7–9)

1 participant

Back | FazBrowse Home | New Git URL