IDLE:not able to complete the hashlib module

[RamchandraApte]

BPO	13933
Nosy	@terryjreedy, @ned-deily, @serwy, @merwok
Files	issue13933.patch

^{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 = <Date 2012-02-04.17:40:32.894>
created_at = <Date 2012-02-03.16:00:05.729>
labels = ['expert-IDLE', 'type-bug', 'library']
title = 'IDLE:not able to complete the hashlib module'
updated_at = <Date 2012-02-05.19:34:10.636>
user = 'https://bugs.python.org/RamchandraApte'

bugs.python.org fields:

activity = <Date 2012-02-05.19:34:10.636>
actor = 'terry.reedy'
assignee = 'none'
closed = True
closed_date = <Date 2012-02-04.17:40:32.894>
closer = 'ned.deily'
components = ['IDLE', 'Library (Lib)']
creation = <Date 2012-02-03.16:00:05.729>
creator = 'Ramchandra Apte'
dependencies = []
files = ['24419']
hgrepos = []
issue_num = 13933
keywords = ['patch']
message_count = 10.0
messages = ['152531', '152588', '152589', '152633', '152635', '152636', '152642', '152673', '152700', '152701']
nosy_count = 6.0
nosy_names = ['terry.reedy', 'ned.deily', 'roger.serwy', 'eric.araujo', 'python-dev', 'Ramchandra Apte']
pr_nums = []
priority = 'normal'
resolution = 'fixed'
stage = 'resolved'
status = 'closed'
superseder = None
type = 'behavior'
url = 'https://bugs.python.org/issue13933'
versions = ['Python 2.7', 'Python 3.2', 'Python 3.3']

[RamchandraApte]

No completion appears when I import hashlib in IDLE and type in "hashlib." and press Tab to complete.
With any other module it works.

[terryjreedy]

Ramchandra: what OS and Python version? Please always include in reports.

I verified on win7,3.2.2, with hashlib versus itertools, idlelib, heapq, trace. (On my new machine, popup is so fast I do not need <tab>.) In fact, 'hashlib.' and <tab> results in a tab to next tab position. ^<space> to <show completions> also does not work, but it does without the '.'. Some files (only builtins?) work without being imported. Everything seemed to work the same in an edit window.

hashlib.__all__ exists as a tuple. Modules without .__all__ also work.
Both cases are handled in idlelib.AutoComplete.fetch_completions().
I see nothing strange about hashlib.

Looking at idlelib.AutoComplete.py, I see other problems to fix.

"This extension can complete either attribute names of file names."
/of/or/, but ^<sp> in name brings up globals list with cursor at closest name to what one has entered. Are these handled elsewhere?

# This string includes all chars that may be in a file name (without a path
# separator)
FILENAME_CHARS = string.ascii_letters + string.digits + os.curdir + "._~#$:-"
# This string includes all chars that may be in an identifier
ID_CHARS = string.ascii_letters + string.digits + "_"

FILENAME_CHARS looks Windows specific. Both definitions are obsolete. IDLE should not separately define such things if at all possible.

But 'hash' is not in the file and I see little specific to the text context. Without being able to insert print calls to see what is and is not executed, I have no idea how to proceed.

[RamchandraApte]

I am using Ubuntu 11.04 with Python 3.2.2 from the package reprositeries (binaries)

[serwy]

Attached is a patch against 3.3a0 to fix the problem. The cause of the bug is that a tuple can not be sorted and calling .sort was raising a silenced exception. The patch explicitly converts __all__ to a list.

[python-dev]

New changeset 741d4aaf7947 by Ned Deily in branch '2.7':
Issue bpo-13933: IDLE auto-complete did not work with some imported
http://hg.python.org/cpython/rev/741d4aaf7947

New changeset 3e9a7fdf0498 by Ned Deily in branch '3.2':
Issue bpo-13933: IDLE auto-complete did not work with some imported
http://hg.python.org/cpython/rev/3e9a7fdf0498

New changeset 443772a82dcd by Ned Deily in branch 'default':
Issue bpo-13933: merge
http://hg.python.org/cpython/rev/443772a82dcd

[ned-deily]

Roger's patch solves the problem. Thanks! Applied to 2.7 (for 2.7.3), 3.2 (for 3.2.3), and default (for 3.3).

[terryjreedy]

Thank you both. I saw the answer (hashlib.__all__ is, unusally, a tuple) but did not see how it was the answer.

[merwok]

BTW I wonder if setting __all__ to something but a list is valid. I think I ran into a similar bug once and fixing __all__ fixed it. The IDLE robustness fix is probably a good thing, but maybe we should also check the stdlib’s __all__s and the docs.

[python-dev]

New changeset 2b93ee675ec4 by Terry Jan Reedy in branch '2.7':
bpo-13933 refine patch using 'new' builtin
http://hg.python.org/cpython/rev/2b93ee675ec4

New changeset d8f988b0c959 by Terry Jan Reedy in branch '3.2':
bpo-13933 refine patch using 'new' builtin
http://hg.python.org/cpython/rev/d8f988b0c959

New changeset 60beb14636b7 by Terry Jan Reedy in branch 'default':
Merge with 3.2
http://hg.python.org/cpython/rev/60beb14636b7

[terryjreedy]

6.11 Import statement says " __all__; if defined, it must be a sequence of strings which are names defined or imported by that module."

I am tempted to open an issue to change 'sequence' to 'list', or even 'sorted sequence', but that is difficult to enforce.