Closed mdickinson closed 9 years ago
Python's current grammar allows a trailing comma after the argument list in:
def f(a, b,):
pass
but not in
def f(*, a, b,):
pass
I propose allowing trailing commas in both situations.
See python-dev discussion starting at
http://mail.python.org/pipermail/python-dev/2010-July/101636.html
Here's a patch. I've checked with PEP-306, but besides changing Grammar, test_grammar.py and the parser module (which there's a separate issue open for), I don't think any other changes are required.
I take it the AST generation just throws the extra comma away? You're sure this doesn't upset any of the node counts in that stage of the compiler?
No, I'm not sure. :) I'll double check.
So I'm looking at ast_for_arguments and handle_keywordonly_args in ast.c. As far as I can tell, that's the only relevant bit; is there anywhere else I should be checking?
There was one place that needed to be changed in ast.c: namely, the check to make sure that there are keyword-only arguments following a bare star.
Here's a new patch, that fixes that issue, updates the grammar in the ast.c comment to match that in Grammar/Grammar, and also updates the production list given in the docs for function definitions.
Retargetting, as this falls under the moratorium, and also because 3.2b1 has been released.
In bpo-10682, several committers indicated that they would prefer not to change this. So I'm closing this as rejected. Per convention, it would probably require a PEP to modify Python in this aspect (as there is no clear consensus).
In bpo-10682, several committers indicated that they would prefer not to change this.
Issue bpo-10682 has been open for less than 24 hours before it was rejected. In contrast, this issue was open after an almost week-long discussion on python-dev where the proposal was well received.
I think bpo-10682 should have been closed as a duplicate of this issue and this issue should be marked as "after moratorium".
I stand by my evaluation: there is clearly no consensus about this change, so it certainly requires more discussion, potentially leading to proponents being asked to write a PEP.
An open issue more accurately reflects the lack of consensus than a closed one, though. We just won't commit it until there *is* consensus that it is a better option than the status quo.
From 10682: the grammar is also inconsistent as to when trailing commas are allowed in function calls, not just definitions.
From 10682: The patch proposed in this (bpo-9232) issue does not fix call syntax but def sytax only. I think it should fix call sytax as well (see code examples in bpo-10682).
python-dev discussion continuation: http://mail.python.org/pipermail/python-dev/2010-December/106770.html
Moratorium's long over. Will this patch rise from the dead?
Will this patch rise from the dead?
It's really down to getting consensus that it's a good idea. That might require another python-dev discussion.
See also bpo-22942 about existing problems with the language documentation.
I would like to see trailing commas supported consistently, especially in function calls. (I think the patch here only does function definitions?) I like to use them when writing arguments on multiple lines, and it is surprising that adding packed *positional arguments can trigger a syntax error.
Maybe this is stretching the scope a bit too far, but it would also be nice to allow more keyword arguments after the **keyword unpacking:
print(1, 2, end=".\n", *(3, 4)) # Supported, but confusing
print(1, 2, *(3, 4), end=".\n") # Better; also suported
print(1, 2, **dict(sep="-"), end=".\n") # Unsupported, but would be nice
print(end=".\n", 1, 2) # Unsupported, for good reason
Maybe some of this is covered by bpo-2292 (generalizing * unpacking), but I haven’t been following that, so I’m not sure.
It looks like if it was not for Raymond's mild dissent, [1], we would have a consensus last time this was raised on python-dev, [2-7].
[1] -? Raymond Hettinger https://mail.python.org/pipermail/python-dev/2010-December/106782.html [2] +0 Guido van Rossum https://mail.python.org/pipermail/python-dev/2010-December/106780.html [3] +0.5 Alexander Belopolsky https://mail.python.org/pipermail/python-dev/2010-December/106782.html [4] +1 Antoine Pitrou https://mail.python.org/pipermail/python-dev/2010-December/106783.html [5] +1 Glenn Linderman https://mail.python.org/pipermail/python-dev/2010-December/106784.html [6] +1 Cameron Simpson https://mail.python.org/pipermail/python-dev/2010-December/106788.html [7] +1 Terry Reedy https://mail.python.org/pipermail/python-dev/2010-December/106789.html
Actual post by Raymond: \https://mail.python.org/pipermail/python-dev/2010-December/106790.html\
Just noticed there are some arguments for trailing commas in the FAQ: \https://docs.python.org/dev/faq/design.html#why-does-python-allow-commas-at-the-end-of-lists-and-tuples\
Have also been confronted to this bug (imo) and this happen from time to time to me :
I often like to ends my (big) functions defs and calls (those that span over multiple lines thus..) with that extra comma,
so that when/if I add another argument (on a new line) later then there will be only a + line in my VCS).
There seems to have a consensus to apply the patch (unless it's not finished?)..
regards,
Reposting from from my newest duplicate of this issue (bpo-24677), which is now closed:
I think that a trailing comma in function definition should be allowed also after *.
Current situation with definitions: def f(args, ): pass # SyntaxError def f(, ): pass # SyntaxError def f(, a, ): pass # SyntaxError def f(, a=2, ): pass # SyntaxError
def f(a, ): pass # Ok
def f(, ): pass # SyntaxError – this should probably stay error
Corresponding calls: f(args, ) # Ok f(args, a, ) # Ok f(*args, a=2, ) # Ok
f(a, ) # Ok
f(, ) # SyntaxError – this is why the corresponding def behavior should stay
My use case: def f(*, long = 1, list = 2, of = 3, kwonly = 4, parameters = 5, ): ...
FWIW I would like to see this, but I think it does need a PEP given the contention so far. For that, we need a BDFL delegate AIUI.
Some remarks:
• A trailing comma after a non-empty argument list is allowed in every call form, including class statement and optional call in decorator syntax. In the grammar, this correponds to arglist
.
• In function definition, trailing comma is allowed only if there is no star before:
def f(a, b, c,): # allowed
def f(a=1, b=2, c=3,): # allowed
def f(*args,): # disallowed
def f(*kwargs,): # disallowed
def f(, a, b, c,): # disallowed
The last example is what bothers me. The presence of the star should not affect whether trailing comma is allowed or not. If f(a, b, c,)
is allowed as a call, it should be allowed in a definition, and if def f(a, b, c,)
is allowed, f(*, a, b, c,)
should be allowed as well.
In the grammar this corresponds to typedargslist
for functions and varargslist
for lambdas.
• A traling comma is allowed in tuples, lists, dicts, sets, the corresponding comprehensions, augmented assignments, and subscripts. It is also allowed in from module import names
in the names part, but only if there are surrounding parentheses. Also a trailing semicolon is allowed for multiple statements in one line.
• A traling comma is not allowed in with statement, import modules
, assert statement (there is just optional second argument), global and nonlocal statements. In all these cases surrounding parentheses are not allowed.
I'm +1 on adding this. I don't believe it requires a PEP. A trailing comma in definitions is already supported in some places, so I don't buy the argument that it catches errors. During the moratorium we were perhaps too strict.
New changeset 419ceb531bab by Robert Collins in branch 'default': Issue bpo-9232: Support trailing commas in function declarations. https://hg.python.org/cpython/rev/419ceb531bab
The patch had some conflicts in the reference docs, I think I resolved it correctly: if someone wanted to cross check my work that would be great. However I was feeling (perhaps wrongly :)) confident so I have committed it as-is.
Do we want to allow a trailing comma after *args or **kwargs in a function definition? Unlike in a call, **kwargs is always the last thing in the list and nothing can be added after that. Just asking.
With PEP-448, we can now have
fronkulate(**kwargs, **kwargs2)
To be explicit, yes, I want to allow trailing comma even after *args or **kwds. And that's what the patch does.
New changeset 6db349fac3ec by Terry Jan Reedy in branch 'default': Issue bpo-9232: Escape rst markup char in NEWS entry to avoid Sphinx warning. https://hg.python.org/cpython/rev/6db349fac3ec
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: ```python assignee = None closed_at =
created_at =
labels = ['interpreter-core', 'type-feature']
title = 'Allow trailing comma in any function argument list.'
updated_at =
user = 'https://github.com/mdickinson'
```
bugs.python.org fields:
```python
activity =
actor = 'python-dev'
assignee = 'none'
closed = True
closed_date =
closer = 'rbcollins'
components = ['Interpreter Core']
creation =
creator = 'mark.dickinson'
dependencies = []
files = ['17969', '18129']
hgrepos = []
issue_num = 9232
keywords = ['after moratorium']
message_count = 30.0
messages = ['110089', '110122', '110143', '110161', '111194', '123829', '123851', '123896', '123906', '123909', '123910', '123915', '123917', '224470', '224488', '239081', '239084', '239085', '239797', '246697', '247032', '247150', '248399', '248411', '248425', '248426', '248427', '248430', '248449', '252252']
nosy_count = 16.0
nosy_names = ['gvanrossum', 'loewis', 'rhettinger', 'mark.dickinson', 'ncoghlan', 'belopolsky', 'larry', 'eric.smith', 'rbcollins', 'zuo', 'Trundle', 'python-dev', 'gstarck', 'martin.panter', 'Drekin', 'pconnell']
pr_nums = []
priority = 'normal'
resolution = 'fixed'
stage = 'resolved'
status = 'closed'
superseder = None
type = 'enhancement'
url = 'https://bugs.python.org/issue9232'
versions = ['Python 3.6']
```