#black-formatter
flat krakenBOT
bright glacier
rightttt
plain atlas
since it currently needs to be triggered manually
bright glacier
readthedocs still borks when we update the stable tag
plain atlas
there's an endpoint, someone just needs to write a script for it
bright glacier
since it currently needs to be triggered manually
no, it does get triggered, it just picks up the wrong commit ...
bright glacier
there's an endpoint, someone just needs to write a script for it
or we could switch to a branch and I bet this would disappear
plain atlas
ah, so someone could still write a script for it π
bright glacier
I'm not in a happy mood fwiw
(unrelated to this, don't worry - just want to caution if I am a little short)
and triggered: https://readthedocs.org/projects/black/builds/15319540/
plain atlas
I'm not in a happy mood fwiw
me 24/7
bright glacier
D'oh that's unfortunate :(
plain atlas
huh
are comments on issues not relayed here
darn
Pipenv has updated a bunch of unrelated packages unfortuantely when updating the lock
lol
why yes, of course it would
flat krakenBOT
plain atlas
huh that is somewhat annoying since its a comment in the pr
bright glacier
import random
n = random.randint(1, 10)
match n%2:
case 0:
print("even")
case 1:
print("odd")
this is valid code?
plain atlas
!e ```py
import random
n = random.randint(1, 10)
match n%2:
case 0:
print("even")
case 1:
print("odd")
late dewBOT
@plain atlas :white_check_mark: Your eval job has completed with return code 0.
even
I wonder if they're going to post a comment or not :p
plain atlas
srsly tho I respect when ppl close like that
bright glacier
Yeah, just unusual to close without a comment
bright glacier
Eh, we don't get too much arguing in the tracker
plain atlas
they commented
plain atlas
Eh, we don't get too much arguing in the tracker
lucky tracker π
dense jungle
except for the tabs people π
plain atlas
I've seen too many github issues with argues
ofc most of the github issues I look at are ones from searches for issues or ones ppl mention to me that I should take a look so my set is very skewed
silent apex
except for the tabs people π
This issue is hilarious ngl https://github.com/psf/black/issues/2584
it's the best one I've seen
plain atlas
or funny issues like that which ppl mention around me smh ^
silent apex
They're like, "hey, you say you're listening to the community, so just give us tabs"
black: we also say:
Black is a PEP 8 compliant opinionated formatter with its own style.
(I mean, I know it says "Tabs should be used solely to remain consistent with code that is already indented with tabs," but still, black's point is to be consistent)
plain atlas
No, not rejected. Give me one good reason why you won't add tab support. Don't say "no eww tabs are gross and everyone who uses black automatically has to hate tabs"
The response was this lol
Our general principle is to avoid options that affect formatting, so that Black-formatted style is always consistent. We have given in and added a few formatting options, but we're unlikely to add more.
Jelle never mentioned tabs or anything lmfao
bright glacier
It's honestly these sort of things that make me consider quitting the project (or OSS in general) - not so the much arguments (I knew what I was signing up for, for the most part) but more the constant decisions on whether to uphold the project's ethos or give into user demands is rather tiring -- hence why I let others do it for me (thanks @dense jungle!)
plain atlas
uphold the project's ethos
I have a solution for you, never have them in the first place /s
we don't need ethics where we're going!
dense jungle
I love making decisions, it's great
bright glacier
I'm both conflict-averse and indecisive.. it's not a good duo trust me.
plain atlas
hey same
plain atlas
I love making decisions, it's great
i love supporting everything I can get my hands on whether its a good idea or not, unless it goes against my vision
dense jungle
I'm pretty conflict averse too, if I have a mild conflict at work I tend to feel bad all day
silent apex
Saaame. But I'm usually fine since when I get an idea in my head for a project of mine (eg. yeeting truthiness as a concept), I will not let it go.
bright glacier
> uphold the project's ethos
I have a solution for you, never have them in the ...
it's one and zeros anyway, who needs ethics in electronics? /s
plain atlas
when I get an idea in my head for a project of mine....I will not let it go
Yeah uh
yeah mhm that's defintely what I do
no feature creep or anything yep mhm
when I get an idea for something while working on something I'll add it. I try to write all of my stuff future proofed for stuff I'm thinking of adding even if I never intend to add it.
I added yaml configuration support because I was like, why not???
and we are offtopic
Thankfully, there's one thing I never need to think about when writing my code, and that's the formatting. Black takes care of that for me, leaving me to focus on what matters.
bright glacier
and now diff-shades has native zip file support (for read operations only though)
plain atlas
niiice, I'm working on zipfile stuff right now and want to end it all #ot2-the-original-pubsta
silent apex
^^^ I absolutely hate thinking about consistent formatting in my code. I hate having to go back and divide up stuff into multiple lines or whatever, and black makes it so easy, so thank you
y'all deserve some more love π€
plain atlas
Well, spacing and stuff need to be done yourself, nothing good can come out of something choosing line grouping automatically
silent apex
well I mean like when I have function arguments spanning 200 characters
toxic stormBOT
dense jungle
the longer it goes without anybody reproing it, the more I'm convinced the OP was confused
plain atlas
(i commented on it, my thoughts are there)
plain atlas
the longer it goes without anybody reproing it, the more I'm convinced the OP wa...
lol
dense jungle
(i commented on it, my thoughts are there)
thanks for that, it's good to have another data poiint
plain atlas
I feel like this happened in a ci workflow
or something
perhaps pip install black --no-deps was used (idfk why)
dense jungle
I'm sure there's many ways to mess up python package installation
plain atlas
yeah
flat krakenBOT
flat krakenBOT
mint barn
how can i make black ignore lists and tuples
errant barn
You can't, because it's not an option we're willing to give users! We like it consistent ππΏ
plain atlas
how can i make black ignore lists and tuples
what's it doing?
mint barn
keeps messing with my large lists and tuples
plain atlas
there's one option that may be affect your diff
mint barn
makes them look like ass
plain atlas
iirc its --skip-magic-trailing-comma
mint barn
iirc its `--skip-magic-trailing-comma`
what does that do
errant barn
if you want to collapse small lists that is
collapses small lists to one line again
and function declarations, calls, tuples etc.
plain atlas
what does that do
its what it doesn't do π
tldr if something has a trailing comma black will expand it no matter the length
this flag turns off that check
also hi felix i haven't talked to you in a while
mint barn
this looks good, why can't black make my lists look like this and instead it makes one large shit that adds a ton of lines for nothing
plain atlas
because that's not an option that black provides
mint barn
π
errant barn
this looks good, why can't black make my lists look like this and instead it mak...
this looks good for you, but not in general
plain atlas
also IMO that only looks good because of the lengths of every param being the same
errant barn
this looks good, why can't black make my lists look like this and instead it mak...
if you're writing a matrix of values, you could use inner lists
mint barn
ig.. but still looks a lot better than what black does to it imo
errant barn
but if you're not writing a matrix of values, I'd argue it is formatted poorly
errant barn
i.e. a 2-d list
plain atlas
also IMO that only looks good because of the lengths of every param being the sa...
which would be extra processing and a really weird rule....
if all_lengths_in_list_are_same then format this one way, else format a different way
mint barn
oh
mint barn
this looks good, why can't black make my lists look like this and instead it mak...
no this is just a list of roles from my server
plain atlas
this looks good, why can't black make my lists look like this and instead it mak...
TBH you can just # fmt: off and # fmt: on around it
mint barn
i use it to check if at least one of the user's role matches one of the id from that tuple
plain atlas
but that should probably be in a datafile anyways π
errant barn
then I'd say you're doing it wrong π hear me out: if you want to remove something from the list, do you fix up the lines to be the same length every time?
mint barn
then I'd say you're doing it wrong π hear me out: if you want to remove somethi...
ok i see your point
plain atlas
why do i have so many this emojis lmao
mint barn
and in fact yeah that's what i do π
errant barn
:D oh no
mint barn
TBH you can just `# fmt: off` and ` # fmt: on` around it
imma do that
errant barn
oh no
errant barn
that's the secret configuration option :D
the "I want to use black but then again I don't"
mint barn
it worked :D
errant barn
fair enough, but please don't π
plain atlas
the "I want to use black but then again I don't"
at that point I just say, use autopep8
lament crow
or yapf
plain atlas
say, since 21.11b0 is messed up, could someone yank it from pypi?
it shouldn't cause any problems, since ppl can still use it
but it tells clients not to, if they can
flat krakenBOT
flat krakenBOT
plain atlas
lmfao wowwwww there's been a bug with @toxic storm for so long here
tired shard
GitHub link?
plain atlas
yeah
glacial drum
lmfao wowwwww there's been a bug with <@!528937022996611082> for so long here
What's the bug?
it seems fine to me
tired shard
GitHub hyperlink is to pydis org
plain atlas
What's the bug?
should it link pythondiscord if you click the github at the top of the embed? its really minor
glacial drum
Ahh, more of an oversight really. The feature only ever used to work for pydis github mentions
plain atlas
yeah
glacial drum
when we extended it to other repos, we didn't update that link
glacial drum
Should be fairly easy to change that url to the org mentioned
bright glacier
I can reproduce the regex issue @dense jungle
seems like if you build your own version of regex via setup.py install it breaks something
dense jungle
I can reproduce the regex issue <@!783088578363523104>
interesting, I tried that too
bright glacier
actually even building your own wheel breaks it somehow ...
dense jungle
not for me on macos
plain atlas
seems like if you build your own version of regex via setup.py install it breaks...
some weird internal direct link to calver somewhere?
bright glacier
how is this relevant?
plain atlas
theory: it builds correctly when the version matches the current date
for some weird internal reason
bright glacier
it's setuptools, I doubt it'd be that broken
plain atlas
!pip regex
late dewBOT
late dewBOT
setup.py line 8
SRC_BASE = 'regex_%i' % MAJOR```it uses different directories for Python 2 and 3
bright glacier
the very old string C-style interpolation :)
dense jungle
but both define Pattern I checked
wait @plain atlas you seriously hadn't seen % interpolation before?
plain atlas
yes lol
dense jungle
wow I'm oold
plain atlas
lol
bright glacier
What's weird is that Pattern exists even in the locally built version of regex 2021.4.4 but it's just that it's only in regex.regex.Pattern
wait a second
does the PyPI wheel contain different code than the source distribution ..
dense jungle
I wouldn't put it past them
plain atlas
hey totally random thing
found this very useful commit
its the change from 2021.4.4 to 2021.7.1
dense jungle
good find, but I think we're only using Pattern not Match
bright glacier
β― diff regex-2021.4.4/regex_3/regex.py regex/regex.py -u --color
--- regex-2021.4.4/regex_3/regex.py 2021-11-19 12:21:46.303850129 -0500
+++ regex/regex.py 2021-04-04 17:00:14.000000000 -0400
@@ -723,6 +723,9 @@
Match = type(_pat.match(''))
del _pat
+# Make Pattern public for typing annotations.
+__all__.append("Pattern")
+
# We'll define an alias for the 'compile' function so that the repr of a
# pattern object is eval-able.
Regex = compile
well then that's this commit https://github.com/mrabarnett/mrab-regex/commit/1e6986b92f978087eb34ca79732d3b0c45be8652
bright glacier
regex-2021.4.4/ is the tar.gz and regex/ is the wheel
dense jungle
well that explains it
bright glacier
Yup.
plain atlas
wait i don't get it
oh got it now nvm
the wheel includes those two lines but the source doesn't.....
dense jungle
could have sworn I tried with --no-binary yesterday but I guess I didn't
bright glacier
So regex's 2021.4.4 release contains multiple artifacts, right? The wheels contain the __all__ code needed to make Pattern public but the .tar.gz doesn't. Which btw is a big no-no in the world of packaging
dense jungle
they might just be doing it manually
bright glacier
I bet it was manually uploaded and someone messed up
plain atlas
and if its still a diff with the most recent code
plain atlas
I bet it was manually uploaded and someone messed up
there's a travis.yml workflow tho
or was, at that commit
dense jungle
so is the 2021.7 version good on both the sdist and the wheel? the bug report claimed that we need 2021.11
bright glacier
I'm checking right now
dense jungle
thanks!
bright glacier
β― diff regex-2021.7.1/regex_3/regex.py regex/regex.py -u --color
the workflow looks good? at this point? https://github.com/mrabarnett/mrab-regex/blob/hg/.github/workflows/main.yml
bright glacier
seems like it
dense jungle
the workflow *looks* good? at this point? https://github.com/mrabarnett/mrab-reg...
where does it build the sdist?
plain atlas
nope nope nope nope it is so bad
lol i think i am overreacting slightly but there is a fatal bug in that workflow
it runs tests but builds and pushes to pypi no matter if they pass or fail
plain atlas
where does it build the sdist?
not sure
late dewBOT
.github/workflows/main.yml lines 26 to 29
- name: Run test
run: |
python setup.py install
python -m unittest -v regex.test_regex```it's an old package. people don't tend to change things until they break
and setup.py install still works. kind of
bright glacier
Yeah, although I'd've expected them to do some cleanup since it seems like they just moved to GitHub and (GitHub Actions)
I assumed it was a new workflow
plain atlas
mrabarnett/mrab-regex#435
toxic stormBOT
plain atlas
mrabarnett/mrab-regex#445 okay I think I'll stop shaking in a few minutes whew
toxic stormBOT
plain atlas
lol that might have been the problem
late dewBOT
.github/workflows/main.yml lines 79 to 83
- name: Build source distribution & wheels
run: |
python setup.py sdist --formats=gztar
python -m pip install -U cibuildwheel
python -m cibuildwheel --output-dir wheelhouse```good point
bright glacier
I thought cibuildwheel handles the Musl emulation for ya
plain atlas
i'll be back in a little bit, I'm still shaking from the tests....
dense jungle
also @bright glacier which wheel did you check? I wonder whether all wheels actually have the right code
plain atlas
idk why that stressed me out so much....
bright glacier
regex-2021.7.1-cp38-cp38-manylinux2014_x86_64.whl and regex-2021.4.4-cp38-cp38-manylinux2014_x86_64.whl
bright glacier
@dense jungle all wheels + the sdist from 2021.7.1 seem to be correct
I downloaded every single distribution for regex 2021.7.1 and checked the sdist manually and the wheels with this script:
import zipfile
from pathlib import Path
import rich
console = rich.get_console()
THIS_DIR = Path(__file__).parent
wheels = THIS_DIR.rglob("*.whl")
for whl in wheels:
with zipfile.ZipFile(whl) as zfile:
with zfile.open("regex/regex.py") as f:
src = f.read().decode("utf-8")
if '__all__.append("Pattern")' in src:
console.print(f"{whl.name}: [bold green]all good")
else:
console.print(f"{whl.name}: [bold red]oh no ...")
the question is, is it worth bumping the min dep up again?
bright glacier
I do wonder how they get the sdist tho.. are they on a platform without a wheel or are they using --no-binary?
bright glacier
ohhh they're on ARM
dense jungle
<@!783088578363523104> all wheels + the sdist from 2021.7.1 seem to be correct
Thanks for checking this! Is it easy to check the other 2021.4.4 wheels too?
plain atlas
idk why that stressed me out so much....
tldr
dense jungle
I might spend some time tonight trying to just get rid of the regex dep
would make life a lot easier
I don't really want to make another release just bumping the regex version
bright glacier
Thanks for checking this! Is it easy to check the other 2021.4.4 wheels too?
Yeah, all of the wheels for 2021.4.4 are fine, just the sdst is borked
dense jungle
phew
bright glacier
Unfornately regex 2021.4.4 doesn't have a wheel for apple arm or linux musl so the sdist is more popular than ideal. OTOH it's a rather old regex at this point.
Man did that issue take forever to reproduce π
Should we at least yank 21.11b0 as it's virtually identical to the less broken 21.11b1?
bright glacier
You're so fast π
bright glacier
flat krakenBOT
gaunt pecan
how do I make black not format a folder
neon loom
how do I make black not format a folder
There is an ignore regex cli / toml config setting
gaunt pecan
There is an ignore regex cli / toml config setting
How do I use it?
abstract halo
any simple bank application projects?
errant barn
any simple bank application projects?
Hi, but this isn't the channel for that mate ππΏ maybe one in lobby or off-topic?
abstract halo
ahh okay, my bad, sorry :)
errant barn
all good mate, you can use Black in your application ;^)
heavy heron
That's an interesting problem, can I not use Black with Pipenv due to how Black specify versions?
plain atlas
you can, yes
heavy heron
Or at least not give it * as the version lock
toxic stormBOT
plain atlas
lol that's a funny number given what its about
plain atlas
Or at least not give it `*` as the version lock
i.e. using it with pipenv forces allow_prereleases flag to be set to true
lol oof
This will happen soon, you're right. We'll be releasing one last (planned) beta, and then an RC of a stable release.
Right. Surely before Christmas, likely not in November though.
three years later
errant barn
Never set deadlines.
red sedge
It is before Christmas though
plain atlas
It is before Christmas though
but the first stable release is supposed to be 22.0, right after christmas π
cold citrus
Hello
red sedge
Hehe
plain atlas
<:this:470903994118832130>
First of all I want to make it clear that I'm not announcing a stable release or a date or really any concrete plans. We've been burned too many times by suggesting dates and the whatnot. What I do want to do is bring more transparency and let y'all know what the project has been up to since the last major update (which yikes is from 10 months ago!).
that first part is very wise
cold citrus
TΓΌrk yok mu ya
heavy heron
Hey @cold citrus, could you please speak english?
errant barn
:D
plain atlas
(4)
heavy heron
!rule 4
late dewBOT
errant barn
Just because "respect staff and their commands?" π
heavy heron
I blame CTRL + K for being super ineffective
plain atlas
also akarys: y'all need to use black.
red sedge
Commands seems a bit strong
cold citrus
Hey @Rabia, could you please speak english?evet akar
errant barn
Let the arbitrary power trips begin
plain atlas
Commands seems a bit strong
i agree, actually
actually its worded as > Respect staff members and listen to their instructions.
cold citrus
Yes akarys
heavy heron
excellent
red sedge
Yes akarys
Do you have a question about the black formatter?
cold citrus
No graingert
bright glacier
> First of all I want to make it clear that I'm not announcing a stable release ...
I wasn't going to get roasted by the hundreds of people watching that thread for making another promise π
plain atlas
lol
bright glacier
How do I use it?
--extend-exclude some_folder/
heavy heron
@dataclasses.dataclass
class AddressPacket:
"""Class representing a single interaction from the client to the server."""
action: str # "get", "set" or "delete"
key: str
value: Optional[str]
@dataclasses.dataclass
class ReturnPacket:
"""
Class representing a value or ack from the server to the client.
"""
status: str # "ok", "error" or "not_found"
value: Optional[str]```Black is happy to leave two different docstring styles. Is it because you consider the docstring as a logical piece of information and removing newlines could have side effects?or
[tool.black]
extend-exclude = '''
/some_folder/
'''
Black is happy to leave two different docstring styles. Is it because you consider the docstring as a logical piece of information and removing newlines could have side effects?
not sure if that's part of flake8
pep8*
bright glacier
```py
@dataclasses.dataclass
class AddressPacket:
"""Class representing a si...
Yup, people were already mad enough with us converting """ """ to """""" so yeah, I'm hesitant to touch 'em more.
heavy heron
it is
bright glacier
TBF sphinx treats those two examples differently
plain atlas
Yup, people were already mad enough with us converting `""" """` to `""""""` so ...
other way around iirc
plain atlas
does Black also sort imports?
no, that's isort
heavy heron
I guess I can check it right now
plain atlas
!pip isort
late dewBOT
plain atlas
black doesn't touch sorting of imports
bright glacier
with --profile black obviously π
heavy heron
I know isort does that
plain atlas
tldr if it did that would be half of the issue tracker
...maybe two-thirds
interestingly i don't think isort is pep8 compat
@heavy heron its actually in pep 257
and surprisingly
def complex(real=0.0, imag=0.0):
"""Form a complex number.
Keyword arguments:
real -- the real part (default 0.0)
imag -- the imaginary part (default 0.0)
"""
if imag == 0.0 and real == 0.0:
return complex_zero
...
plain atlas
? unread
delicate marsh
plain atlas
ah
flat krakenBOT
red sedge
What are the limits of the black match parser?
Is it just because match is a soft keyword?
dense jungle
What are the limits of the black match parser?
I think it got run around a library of test cases and could parse >90% of them, but haven't looked into the exceptions
red sedge
I think it got run around a library of test cases and could parse >90% of them, ...
I mean what sorts of things can it not do in principal?
dense jungle
I mean what sorts of things can it not do in principal?
Nothing I'm aware of
red sedge
Because it's not a PEG parser it can't ever get them all - but which types can it not do?
At least I thought it couldn't
plain atlas
Is it just because match is a soft keyword?
soft keyword?
dense jungle
soft keyword?
you can still write re.match even though match is a language keyword
it's only a keyword in certain positions
red sedge
Yeah match = re.match(r"...") and if match: all work
dense jungle
this is also true for case; it was true for async and await for a few versions too
red sedge
But async and await aren't as soft as match?
red sedge
async def was a hard keyword that promotes await into a hard keyword
plain atlas
Yeah `match = re.match(r"...")` and `if match:` all work
even in the same block of code?
plain atlas
I thought it was 3.5.3
dense jungle
definitely not a change for a point release π
plain atlas
why was 3.5.3 so major anyhow?
red sedge
Sorry 3.7 made them hard
dense jungle
why was 3.5.3 so major anyhow?
typing and asyncio were provisional around 3.5, so they still made major changes in point releases. the syntax wouldn't have changed though, just the libraries
red sedge
Python 3.6.15 (default, Sep 10 2021, 00:26:58)
[GCC 9.3.0] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> await = 2
>>>
KeyboardInterrupt
>>>
$ python3.7
Python 3.7.12 (default, Sep 10 2021, 00:20:04)
[GCC 9.3.0] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> await = 2
File "<stdin>", line 1
await = 2
^
SyntaxError: invalid syntax
>>>
ah, because I know there was a while when many libraries supported, 3.5.3 and up but not 3.5.2
plain atlas
Is it just because match is a soft keyword?
so to recurse on this, black can format 3.6 code and it doesn't seem to have problems with await...
red sedge
ok but I'm sure someone said match could not be done without a PEG parser
same for parenthesized with
plain atlas
what does peg stand for?
dense jungle
I dunno, it's a parsing technique
red sedge
The current Python grammar is an LL(1)-based grammar. A grammar can be said to be LL(1) if it can be parsed by an LL(1) parser, which in turn is defined as a top-down parser that parses the input from left to right, performing leftmost derivation of the sentence, with just one token of lookahead.
plain atlas
interesting
red sedge
But a PEG parser has infinite look ahead
Which means you can parse multiple alternatives of match and use the subsequent context to find out if it was a match block or just the Name match
plain atlas
given that there's the ast module, why can't black just use that to remake the code? (apologies if this is oversimplified, there's probably a clear reason why)
sour jay
Didn't Guido have to do a lot of work to get these soft keywords to work in the LL(1) parser?
uncut patio
@plain atlaswell, ast doesn't include comments
sour jay
Doesn't that module somewhat require running the code?
plain atlas
no, not that I know of, but I know black parses the ast anyways
that's what the safe mode does, which is on by default
cold citrus
Hello
cold citrus
hello arl
dense jungle
given that there's the ast module, why can't black just use that to remake the c...
you got some good answers today but in more detail: the AST is a representation of all the parts of the code that affect how it's run, but it throws away a lot of details that black may want to preserve. off the top of my head those include: comments, whitespace (we preserve newlines between blocks of code, for example), the trailing comma (for the magic trailing comma feature), raw vs. normal strings, string quote style (e.g. triple quotes, which we preserve), hex vs. normal ints, parentheses
we do also parse the AST, but that's only as an additional safety check to make sure our formatting changes don't change the AST (except for a few limited exceptions)
plain atlas
never heard of them
red sedge
Seems like this should be a list of sponsors now
bright glacier
Haha, yea we get paid nothing for this work for the most part :p
ornate patio
not sure if this is the right channel to ask questions about black, but here goes
I have a block of commented code similar to this:
# fmt: off
# x = 3
# if x == 5:
# print("hello")
# print("goodbye")
# fmt: on
and when I run black I get this:
# fmt: off
# x = 3
# if x == 5:
# print("hello")
# print("goodbye")
# fmt: on
is that expected? I would think the fmt tags should leave it indented as is?
plain atlas
huh the fmt indicators don't apply to the comments
plain atlas
is that expected? I would think the `fmt` tags should leave it indented as is?
I would check the bug tracker and make an issue TBH
if there isn't one already
wait no
what's your black version
black --version
ornate patio
black, version 21.9b0
bright glacier
Pretty sure this is a known bug
ornate patio
i get the same behaviour with black, version 19.10b0
ornate patio
Pretty sure this is a known bug
ah ok, let me look at open issues then
bright glacier
huh the fmt indicators don't apply to the comments
yeah the implementation probably doesn't handle this well
plain atlas
not sure if this is the right channel to ask questions about black, but here goe...
a stopgap might be moving the # to the beginning of the print statements btw
and commenting out the whitespace
ornate patio
yeah that works
plain atlas
(it also plays better with ide comment implementations)
ornate patio
so the actual use case is some internal tool that's using comments as tags
but yeah the indent inside the comment instead of outside is a good intermediate solution
bright glacier
This is total guesswork (from what I barely remember about the fmt: on/off implementation) but the AST leaf containing # fmt: on wouldn't be ignored, everything else contained is though. This is a problem as comments and whitespace is stored in the AST as prefix attributes on AST nodes and leaves, so since this is attached to the non-ignored # fmt: on leaf, the indentation normalization logic edits it.
plain atlas
so the actual use case is some internal tool that's using comments as tags
how does that work, lol
questionable design choice
ornate patio
*how does that work, lol*
idk, I'm new lol
bright glacier
And fmt: on is its own leaf instead of being part of the prefix because that's literally how the fmt: on/off implementation works.
plain atlas
idk, I'm new lol
what's it do?
ornate patio
what's it do?
I'm pretty sure it's for removing blocks of code depending on where the file is deployed
plain atlas
that's.... interesting and I have many questions
ornate patio
yeah honestly not a fan
plain atlas
yet I want to hear none of the answers
bright glacier
psf/black#1245 -- this one?
toxic stormBOT
plain atlas
I'm pretty sure it's for removing blocks of code depending on where the file is ...
has anyone heard of if statements in your workplace /hj
bright glacier
I almost fixed this issue matter of fact but there was one edge case I couldn't work around
plain atlas
could use an enum for deployment location, and then if the lines of code.
my consulting rate is $18/h I'll send you a bill /s
bright glacier
!rule 9
red sedge
Are you doing anything special for Friday?
dense jungle
Could a workaround be to start your line with # print instead of # print?
bright glacier
Are you doing anything special for Friday?
Hmm, is this for me?
ornate patio
Could a workaround be to start your line with `# print` instead of ` # pr...
yeah that's probably what I'll do
red sedge
Hmm, is this for me?
The project in general
dense jungle
ah black friday π
bright glacier
d'oh
dense jungle
@plain atlas might reduce their consulting rate
plain atlas
lmfao
plain atlas
Could a workaround be to start your line with `# print` instead of ` # pr...
ahead of ya :D #black-formatter message
plain atlas
we'll ship PRs with discounted review quality
wait someone's gonna let me review?? oooo neat
jk all of my reviews are of Superior Quality
bright glacier
!eval
a = 11
b = 12
match a, b:
case _, 5:
print("a5")
case 11, 12:
print("default")
case _:
print("badass")
@bright glacier :white_check_mark: Your eval job has completed with return code 0.
default
Alright this is valid code
plain atlas
I believe you were missing a space after eval and the first ``` I am behind
bright glacier
Yeah I was but I've thought it'd ignore that
plain atlas
nope, dpy doesn't care
bright glacier
!eval
```python
a = 11
b = 12
match a, b:
case _, 5:
print("a5")
...
Huh, I thought this was fixed by isidentical
ah, isidentical fixed this similar case:
value = (1, 2)
match value:
case x, y:
pass
what is isidentical? a lib?
bright glacier
a core dev
bright glacier
anyway I fixed this issue with this sketchy patch:
diff --git a/src/blib2to3/Grammar.txt b/src/blib2to3/Grammar.txt
index c2a6254..92bd22c 100644
--- a/src/blib2to3/Grammar.txt
+++ b/src/blib2to3/Grammar.txt
@@ -238,7 +238,7 @@ yield_arg: 'from' test | testlist_star_expr
# to reformat them.
match_stmt: "match" subject_expr ':' NEWLINE INDENT case_block+ DEDENT
-subject_expr: namedexpr_test
+subject_expr: namedexpr_test [(',' namedexpr_test)*]
# cases
case_block: "case" patterns [guard] ':' suite
will test later -- this is probably dumb but I don't have time right now
plain atlas
why would we need a method that is basically is
bright glacier
?
plain atlas
isidentical
bright glacier
that was a bad joke imo
errant barn
you must be joking :D
plain atlas
thought you were first meaning isidentical(a, b) which is the same as a is b
plain atlas
that was a bad joke imo
I wasn't joking
bright glacier
@muted haven it's this person
errant barn
bright glacier
!eval
```python
a = 11
b = 12
match a, b:
case _, 5:
print("a5")
...
anyway, isidentical, black is having trouble parsing this
plain atlas
yeee I got that now lol
rain flower
Hi, just wondering if you all have any easy open issues? Would like to get familiar with the codebase π
mint barn
@obsidian sapphire@junior crystal@mint barn This channel isn't for running bot commands, that's what #bot-commands is for. This channel is for the discussion of an autoformatting tool called "black"
junior crystal
Oh woops sorry about that!
obsidian sapphire
Srry
muted haven
anyway, isidentical, black is having trouble parsing this
Will take a look at it tonight, thanks for CC-ing π
leaden dove
How can I ask black to format with single quotes
errant barn
-S skips the normalisation, but I don't think there's an option for switching the quote style
errant barn
How can I ask black to format with single quotes
I write with single quotes too since they are easier to type, but then Black can just change them to doubles
glad pebble
How can i use Black (not racist) in android command line ?
plain atlas
I write with single quotes too since they are easier to type, but then Black can...
i write with whatever I feel like lmao
and then black just makes them standard
ironically using a formatter has made my code format without a formatter very very shit
errant barn
I feel ya π linters on the other hand.. they bullied me into writing code that I'd be surprised if it were modified by black, disregarding the quotes
tacit skiff
challenge: make a algorithm program with \n
red sedge
How can I ask black to format with single quotes
there's this https://github.com/pre-commit/pre-commit-hooks/blob/38b88246ccc552bffaaf54259d064beeee434539/README.md#double-quote-string-fixer which you can run after black -S to change your high quality double quoted strings to use the incorrect/inferior/wrong quote style
plain atlas
@bright glacier an edge case above would have broken your next-pr-number script ^ π
flat krakenBOT
bright glacier
<@!429445826251587616> an edge case above would have broken your next-pr-number ...
Could you elaborate please?
plain atlas
i mean, you can't solve it
but the two above were within seconds of each other
race condition where someone makes a pr after the command is run, so the number now already exists.
bright glacier
Well yeah I thought that was obvious
I could add a (silly) disclaimer but I don't have room for it really
plain atlas
true
dense jungle
Maybe we should do a new release with the match fix and bump regex to make the ARM people happy
I won't have time for it today though
bright glacier
oh gosh I just realized diff-shades is eating 5x the size of the analysis file in RAM while loading it
when it's cached as a pickle there's no extra copies, but without caching it's real bad
I think I just saw a peak of 1GB of RAM usage (with a json file, not a zip as like with the previous comment) π
This file is ~100MB, genuinely what the heck
stuck vapor
Sorry for the somewhat scant descriptions in the above, bit short on time, need to start by day job in a few minutes π
muted haven
Maybe we should do a new release with the match fix and bump regex to make the A...
If you could wait the weekend, I want to try making some enhancements on the error reporting. It doesn't seem to work properly (not sure if I can fix it, but it would be nice if it gets fixed).
E.g:
def something():
match something:
case t:
pass
match arg:
case t:
print(t)
case z := u:
print(a)
shows the line 2, but the actual error happen on the line 9 (and the match block it is bound to is in line 6).
errant barn
How would you feel about restricting the conversation in psf/black#517 to maintainers so that we can avoid the noise and just provide updates to the people that are following the issue?
toxic stormBOT
errant barn
I'm getting irrationally frustrated π
flat krakenBOT
dense jungle
I'm getting irrationally frustrated π
done
errant barn
done
ππΏ surprisingly enough I don't have the privilege to comment either, but I trust you do and that's enough π
flat krakenBOT
plain atlas
...than black, the OG of Python formatters
Has black really been around that long?
(location of snippet not super relevant)
dense jungle
> ...than black, the OG of Python formatters
Has black really been around that l...
nah yapf and autopep8 are older
plain atlas
lol a fork of black gave that in their readme
bright glacier
Hi, just wondering if you all have any easy open issues? Would like to get famil...
!remind 4hours address this
late dewBOT
plain atlas
hey that migration to coveralls, I found an easy official way to do it for gha without causing everything to fail if coveralls is down
late dewBOT
!remind 4hours address this
It has arrived!
bright glacier
@dense jungle , what's the difference between your regex reduction PR and the regex pattern bugfix PR? It seems like the former contains the latter.
bright glacier
it feels good to have so much tooling for black now
like I'm reviewing Jelle's PR and checking impact and performance is relatively easy :)
dense jungle
<@!783088578363523104> , what's the difference between your regex reduction PR a...
yeah it does. I separated the bugfix out because it seemed less controversial
bright glacier
I'm going to approve your regex reduction PR -- what controversy are you expecting?
dense jungle
I mean the tokenize thing is kind of hacky and might break in some way I didn't think of π
bright glacier
yeah hopefully not
it didn't break anything on the 2 020 758 lines of code diff-shades ran over so that's encouraging but I suspect the potential bugs will surface in invalid code the most
certainly worth some more thought I suppose
dense jungle
it makes this work: ```(black) jelle@mbpt-root black % black --fast -c "xΒΏ = 3"
xΒΏ = 3
though I am not sure that will have any practical effect
bright glacier
yeah especially as our match support is very liberal already
bright glacier
It seems like the noregex PR is a bit faster than main somehow. It's probably noise but at least it doesn't make things slower AFAICT.
jolly oriole
So with keyword arguments whenever I format my code,
Black adds spaces around the = when the argument is annotated but when there's no annotation black doesn't add a space around it.
Is this a feature or a flaw?
dense jungle
So with keyword arguments whenever I format my code,
Black adds spaces around th...
That's intentional. PEP 8 also explicitly calls for this.
jolly oriole
Ohh, just wanted to confirm thanks.
red sedge
are parenthesized with handled at all in the formatter?
eg, I'd expect this :
import contextlib
import collections.abc
@contextlib.contextmanager
def foo() -> collections.abc.Generator[None, None, None]:
yield None
@contextlib.contextmanager
def bar() -> collections.abc.Generator[None, None, None]:
yield None
with (
foo() as eggs,
bar() as ham,
):
pass
to remove the parentheses in the with statement:
with foo() as eggs, bar() as ham:
pass
oh wait magic comma
ah it does this though:
with (foo() as eggs, bar() as ham):
pass
those parenthesis are now redundant
flat krakenBOT
dense jungle
ah it does this though:
```python
with (foo() as eggs, bar() as ham):
pass
...
we don't really do anything well-considered there. Batuhan's PR just added support for parsing them, we didn't think about how to do this formatting
plain atlas
now that I know that isidentical exists I will never forget
this is what I first thought richard was referring to...
def isidentical(a: Any, b: Any) -> bool:
return a is b
isidenticalexists is another level entirely
plain atlas
lol
plain atlas
...
i've legit never seen that order of webhooks
its always pull closed, x new commit(s)
flat krakenBOT
flat krakenBOT
plain atlas
black formats this
tldr its the magic trailing comma that black does, you can turn it off by passing skip-magic-trailing-comma
plain atlas
true
this is a set
yknow, I would check open and closed issues and make and issue if there isn't one already
slow ibex
tuples are not expanding because magic comma isnt magic, it is important part ((x) is not (x,))
plain atlas
!e print(type({'d'}))
late dewBOT
@plain atlas :white_check_mark: Your eval job has completed with return code 0.
<class 'set'>
so that's probably why sets don't expand, its not necessary there
plain atlas
unless there a structure which can load into a set which does cause it, wouldn't be surprised
slow ibex
I will clean the messages
slow ibex
ok π
mint barn
+ python3 -m black main.ipynb
Skipping .ipynb files as Jupyter dependencies are not installed.
You can fix this by running ``pip install black[jupyter]``
$ cat requirements.txt | grep black
black
black[jupyter]
$ pip3 list | grep black
black 21.11b1
```Was wondering why my `black[jupyter]` in my `requirements.txt` doesn't seem to install correctly. Installing with `pip install -r requirements.txt` vs `pip3 ...` makes no difference. Running `pip install black[jupyter]` manually does fix it though, but I'd like to avoid doing that.jupyter is a install_extras
it doesn't actually install a package named black[juypter]
it installs.....
mint barn
Is there a way to specify to install it in the requirements.txt file?
mint barn
lmao it's true
plain atlas
okay aside from me spelling it wrong, its actually jupyter
anyway, it should actually install it even in your requirements.txt file
tribal thistle
Fuzzy matching on pip install when
plain atlas
eys
we ned that
it would save me from myself
@mint barn anyways, it should be working....
what is your pip version?
(pip -V)
plain atlas
#selfshame
bright glacier
jupyter
mint barn
ah okay, it doesnt work if you have black and then black[jupyter]
sorry thanks for the help
flat krakenBOT
flat krakenBOT
dense jungle
the fuzzer found a real bug π
dense jungle
well this was on my sketchy tokenizer change so I was worried it was real
dense jungle
I'm trying to think of ways to replace the recursive regex at https://github.com/psf/black/blob/main/src/black/trans.py#L985 so we can stop having regex as a dependency but so far all I can think of is manually writing a recursive descent parser (probably slow) and using libcst (which probably won't improve dependency hell)
late dewBOT
src/black/trans.py line 985
RE_FEXPR: Final = r"""```If the user cache DIR is not writable by the user wouldn't black fail?
slow ibex
accept only certain nesting depth, other code leave as is
deep nested f-strings are rare
elder tusk
I'm trying to think of ways to replace the recursive regex at https://github.com...
do you really need recursive descent? it looks like you only need the spans, and you know you don't have syntax errors, so you might be able to get away with a loop and a stack
flat krakenBOT
bright glacier
do you really need recursive descent? it looks like you only need the spans, and...
Wait where do we verify the f-string doesn't have syntax errors? I wasn't aware the string transformer machinery validates the syntax.
flat krakenBOT
sullen pendant
Any beginner friendly open source projects web based??
red sedge
Any beginner friendly open source projects web based??
Yes
bright glacier
I spent some time and I now have a non-regex f-string scanner implementation that passes the test suite just fine. It's pretty sketchy but at least it should be possible to create something decent without regex.
If I can clean it up a bit more then I might submit it for potential inclusion
dense jungle
I spent some time and I now have a non-regex f-string scanner implementation tha...
nice! let's do it
you can make another PR and then we can actually get rid of regex π
plain atlas
I spent some time and I now have a non-regex f-string scanner implementation tha...
nice, does it work on the provided test strings by fix error#1642 lol
elder tusk
Wait where do we verify the f-string doesn't have syntax errors? I wasn't aware ...
we don't, but i think we wouldn't be able to parse with an invalid f-string, right? btw, i put some good f-strings in cpython's Lib/test/test_unparse.py that you might want to check against, also stuff like f"{'{'}"
flat krakenBOT
flat krakenBOT
dense jungle
we don't, but i think we wouldn't be able to parse with an invalid f-string, rig...
I think we can, because blib2to3 just doesn't parse the contents of f-strings
it occurred to me that a natural extension to your suggested change though would be to format the expression bits of f-strings
elder tusk
hmm, okay, i think that's still fine, since the PR won't crash on invalid f-strings; it just might not yield anything. and then at the end we'll still get a crash in --safe or whatever
yeah, formatting expression bits would be nice
bright glacier
we don't, but i think we wouldn't be able to parse with an invalid f-string, rig...
ah yea mine doesn't handle this well
β― python scanner.py "{'{'}"
VALID INPUT
0 [] {
1 [0] '
2 [0] {
3 [0, 2] '
4 [0, 2] }
pop due to even
0.1
0.07 ("{'}", 2, 5)
FAIL
``` although fwiw my goal was to first match the regex's behaviour ... and it turns out the regex fails on this one too. The first line (one after the pop due to even log line) is my scanner implementation and the other is the regex one.I wasn't happy with my implementation hence the "if I can clean it up" bit, I really didn't like the hacky asserts of the potential expression. I didn't think about checking strings with other styles of quotes though, nice catch!
Wait a second, does the transformer implementation expect open ranges or half open ranges? the implementation is fine, don't worry I got confused
plain atlas
lmao black
- WEBHOOK = ''''''''''''''''''''''''''''''''''''''''''''''''
+ WEBHOOK = """""" """""" """""" """""" """""" """""" """""" """"""
```*this exists because i zoned out with my finger on `'`, then thought to run it through black*good job on getting exactly a multiple of 6 of quotes
bright glacier
FYI I am going to start community testing of the mypyc wheels soon. Please merge any major PRs soon so I don't have to constantly rebuild to keep the data up to date.
I'll try to review the lazy imports PR soon.
bright glacier
gosh dammit our tox setup is not compatible with my needs to test under cibuildwheel as part of the compiled wheel build ..
flat krakenBOT
neon loom
FYI I am going to start community testing of the mypyc wheels soon. Please merge...
Is there any mypyc github actions helpers? (I haven't googled yet)
bright glacier
AFAIK no
The mypyc bit is not too painful but setting up a workflow to build and test across the various platforms that is painful
plain atlas
The mypyc bit is not too painful but setting up a workflow to build and test acr...
π
I sense an imbalance in the github actions force
plain atlas
should be explained in #βο½how-to-get-help
hidden junco
k thx
plain atlas
no problem!
plain atlas
The mypyc bit is not too painful but setting up a workflow to build and test acr...
so which part are you having trouble with?
also #tools-and-devops might be more fitting
neon loom
The mypyc bit is not too painful but setting up a workflow to build and test acr...
I'm a big fan of recommending to start simple with CI/automation - What's the easiest platform to build for? I'd expect Linux? Lets start there so we can at least guard against PRs that break the compile.
Then once we have that we can add other platforms. I'm down to help.
Maybe step 0 is documenting the manual process (sorry if you have)
plain atlas
Then once we have that we can add other platforms. I'm down to help.
same, i defintely do not spend too much time with CI
neon loom
Then I'm willing to take a stab at a action from scratch to do it
neon loom
same, i defintely do not spend too much time with CI
It's my jam and what I live for and spend most time doing in $dayJob, home + open source projects etc.
neon loom
e.g. I can push to my ansible repo and if i don't need the change everywhere right away, I do it on 1 host to test, push and then it will automatically apply a few hours later
plain atlas
wow
neon loom
Love shit like that
https://github.com/cooperlees/ansible_shed - Wrote my own simple version of ansible tower
plain atlas
rn i just have a lint/test workflow, nothing more complex than that yet
bright glacier
I got it all automated and working but I can't use tox as part of the testing step which is incredibly annoying
neon loom
tox is incredibly annoying - so this is good π
(tox is fine, it's good python has a psuedo standard)
plain atlas
i heard nox is nicer, and I think I'll be writing a config thing like nox uses for a project of my own too
neon loom
nox sorted died as I got into doing a lot of CI so I never started using it
plain atlas
!pip nox
late dewBOT
bright glacier
I prefer nox but honestly I don't even know what to do with pipenv
neon loom
Yeah, I never found pipenv to be a game changer like others did
Pretty much everything it did I needed I feel I could do with pip
plain atlas
honestly i feel like dropping pipenv.lock from the repo would solve a lot of problems
bright glacier
late dewBOT
.github/workflows/build.yaml lines 73 to 77
# So these are the environments we are targeting:
# - CPython 3.6 and higher (no PyPy)
# - 64-bit only
# - ARCH: amd64, x86_64, universal2, and arm64
# - OSes: Linux (no musl), Windows, and MacOS```cibuildwheel makes it a lot less painful fwiw
there's also no support for ARM windows because who the heck is using that?
plain atlas
honestly i feel like dropping pipenv.lock from the repo would solve a *lot* of p...
it doesn't seem like it solves any problems, either. I know that you're supposed to check it in, but its also possible to remove it
bright glacier
There's no good reason to keep 3.6 except for that it doesn't really cost us to keep it
plain atlas
there's also no support for ARM windows because who the heck is using that?
raspberry pi tinkerers who install windows π
bright glacier
Pretty much everything it did I needed I feel I could do with pip
Me too, I'd much prefer a pip-compile + nox setup
except pip-compile doesn't have a proper resolver π₯²
plain atlas
poetry does 
bright glacier
You do realize we are scarred by pipenv, right?
plain atlas
slightly
I do tend to see a lot of pipenv sucks but no moves to get away from it here lol
neon loom
There's no good reason to keep 3.6 except for that it doesn't really cost us to ...
O, nox is not what I am thinking it is at all. I was thinking of nose. My bad.
plain atlas
ah yeah, nosetest
neon loom
I *do* tend to see a lot of pipenv sucks but no moves to get away from it here l...
Lukasz (ambv) bought into it and it was looking promising when he started black to be the future (I sat next to him @ work - we were in same tema)
*team
plain atlas
ah yeah, nosetest
i use pytest or unittest depending on the project
plain atlas
Lukasz (ambv) bought into it and it was looking promising when he started black ...
i actually used pipenv for a project when i was new, but switched to poetry when pydis switched
neon loom
The pipenv support deprecation would need to be communicated and see how much noise we get. I could start an issue and see what uproar we get from people who need it
Could open an issue and see - but i don't think people pay enough attention
plain atlas
wouldn't pipenv deprecation only mess with contributors?
bright glacier
oh yeah because people do actually use Pipfile (not PIpfile.lock)
neon loom
Ya
plain atlas
I do think a minor step in the direction would be to remove pipfile.lock from the repo
silent apex
Why? Isn't a lockfile one of the greatest benefits of pipenv?
plain atlas
since.... it ensures the contributors and developers all have the same dependencies, whereas in a real-world use, the dependencies are all different
plain atlas
Why? Isn't a lockfile one of the greatest benefits of pipenv?
I can't really see any benefits from pipenv, its slow and clunky and always has lockfile issues
bright glacier
so anyway the main things I need to do from a mypyc perspective is 1) actually get the community to test my wheels, 2) integrate a build step into CI, and 3) document, document, and document
plain atlas
from an exploration of issues on the pipenv repo: ( pypa/pipenv#954 pypa/pipenv#598 )
The standard is to commit it for applications, for libraries, less so
Given that black is typically used as a third party tool in workflows and environments, in this case I would consider black a library: its not the only thing being installed to an environment.
This means that the contributor libraries are pinned to a very different verision than what could be installed.
In addition, take the recent versioning problem with regex. This should have been caught and found before release, by testing with the lowest supported versions of dependencies. With the pipfile.lock, all of the dependencies are the same for every contributor, and during the development cycle all dependencies are the same.
When installed from pypi, pip or the installation tool uses its own dependency resolver that matches the provided requirements--which may or may not necessarily match the versions in pipfile.lock
Given all of this information, I am a believer that pipfile.lock should not actually be commited to version control for black.
I also believe that the lowest supported dependencies should be used for testing, but that's a different argument
toxic stormBOT
plain atlas
@neon loom@silent apex@bright glacier ^
bright glacier
why do people always think black is a library? like I get your argument but we don't even have an official API
from a packaging standpoint, sure we may package it like a library but that means less way than a fullblown library
plain atlas
for purposes of the above, it is IMO, but i agree it is not, yet
bright glacier
It'd be nice if pip had an upgrade strategy to install the oldest supported deps but unfortunately that feature is still in the planning phase
plain atlas
I could write one (partially)
I don't think it would work on the deps of the deps from how I'd do it lol
bright glacier
I mean if I had to make a solution up I'd use minideps
plain atlas
is that a package?
bright glacier
mindeps?
bright glacier
is that a package?
no.
plain atlas
aw
bright glacier
what do you mean?
plain atlas
everything on pypi is packaged, right?
bright glacier
yeah?
bright glacier
I don't know
it's a subtool of a larger tool by FFY00 that spits out a list of pinned requirements that are the oldest possible
not sure it uses a resolver and it doesn't support packages with spotty wheel coverage but it'd be a good start
plain atlas
I mean, its not a hard tool to write, depending on if it supports sub-deps or not
bright glacier
are you sure? because conflicts are bound to happen
bright glacier
Yeah. Anyway I am getting incredibly annoyed (because of me being an irresponsible person) and tired so I need to go or I'm going to snap.
plain atlas
(pasted for those who want it)```py
black 21.11b1 The uncompromising code formatter.
βββ click >=7.1.2
β βββ colorama *
βββ mypy-extensions >=0.4.3
βββ pathspec >=0.9.0,<1
βββ platformdirs >=2
βββ regex >=2021.4.4
βββ tomli >=0.2.6,<2.0.0
βββ typing-extensions >=3.10.0.0
βββ typing-extensions !=3.10.0.1
you forgot the jupyter and d extras
plain atlas
ah right i don't have them installed in this environment π¬
bright glacier
there's the colorama, python2, and uvloop extras but they each only add one more dep
plain atlas
the jupyter (oh look i spelled it correctly) and d deps add a lot tho.....
bright glacier
yes, I'm going now.
bright glacier
I will not but oh well
slow ibex
is black compiled with mypyc?
tribal thistle
Not currently. Richard is trialing that, but that will still go through a bit of a public testing period
flat krakenBOT
flat krakenBOT
mint barn
Hi - the stable tag on github still points to 21.11b0 which has been yanked on pypi - should it be updated to point to 21.11b1?
dense jungle
Hi - the `stable` tag on github still points to `21.11b0` which has been yanked ...
thanks, done
muted haven
This is a bit weird,
only happens when used with match or case π€ probably some keyword shenanigans
upps, seems like since both of the matches are assumed to be keywords, we just split them. https://github.com/psf/black/blob/e151686c6fc291a72058a26de8e6279669d756cc/src/black/linegen.py#L140-L141
late dewBOT
src/black/linegen.py lines 140 to 141
if child.type == token.NAME and child.value in keywords: # type: ignore
yield from self.line()```
celest grail
bright glacier
is that a function definition? in that case you don't need the quote to define a parameter named file
celest grail
bright glacier
Well I more meant you don't need to quote file
plain atlas
thanks, done
trigger docs rebuild
dense jungle
*trigger docs rebuild*
oh I need to do that manually don't I? will have to figure out what button to push in RTD
plain atlas
yeah, although.... I'm tempted enough at this point to write a workflow for it
the gist is if stable tag edited, send api request to rebuild
flat krakenBOT
bright glacier
Alright I'm kicking off community testing for the mypyc wheels π To get started please see this issue: https://github.com/ichard26/black-mypyc-wheels/issues/12
silent apex
Alright I'm kicking off community testing for the mypyc wheels π To get started...
!remind 3h dew it
late dewBOT
lament crow
Alright I'm kicking off community testing for the mypyc wheels π To get started...
in case anyone wants to verify these, the wheels come directly from this actions run
https://github.com/ichard26/black-mypyc-wheels/actions/runs/1522879877
slow ibex
pip uninstall black && pip cache purge && pip install black -U --extra-index-url https://ichard26.github.io/black-mypyc-index/simple/ && black --version
```what am i doing wrong?
bright glacier
huh
bright glacier
```py
pip uninstall black && pip cache purge && pip install black -U --extra-ind...
what happens if you do pip install black --index-url https://ichard26.github.io/black-mypyc-index/simple/ --no-deps --force-reinstall?
slow ibex
bright glacier
right - I forgot that 3.10 isn't supported on windows
slow ibex
yes, im on 3.10
bright glacier
I think I can get 3.10 working for windows but the last time I tried the C compiler got very mad at me π¦
IIRC a dependency bump should fix it
I should definitely add better instructions though
slow ibex
black works correctly on 3.10
and im using mypyc for my project, it is also working correctly
bright glacier
The last time I tried I'd get compiler errors on mypyc's C shared library w/ mypy==0.910 on Windows Python 3.10
Alright I'm rebuilding the wheels this time with 3.10 enabled for Windows - hopefully this passes π€
Estimated time remaining is roughly 40 minutes (for both CI and me updating the index).
slow ibex
why it is so long? i think, mypyc can build it in about one minute
bright glacier
really?
even the fastest wheel build takes a good two minutes: https://github.com/ichard26/black-mypyc-wheels/runs/4373276347?check_suite_focus=true#step:6:2224
to be fair I am using an interpreted copy of mypyc as their release schedule is really bad and I need some of the bugfixes available only in the master branch but it still takes a while
late dewBOT
!remind 3h dew it
It has arrived!
bright glacier
dawn you aren't on Windows using CPython 3.10 right?
bright glacier
I was asking dawn the moderator not you denball π
slow ibex
sry π
english is not my native language
bright glacier
Can you try again @slow ibex ? I just pushed the 3.10 wheel for windows
slow ibex
bright glacier
that's good
bright glacier
Don't forget to fill out that issue form once you feel ready so I can keep track but thank you!
silent apex
dawn you aren't on Windows using CPython 3.10 right?
Err, I can use whatever version on Windows
or WSL
I don't mind installing any version
bright glacier
It doesn't matter anynow, I recently added 3.10 support - just please don't be using an ARM cpu or a 32-bit OS :p
silent apex
Nope, just a regular old intel 64-bit thing
bright glacier
I could theoretically support 32-bit but that would double the workload on CI and testing requirements for relatively little gain IMO given 64-bit OSes are wayy more popular.
dense jungle
at Quora we ran 32-bit Python until just a few years ago and it was a major pain because it got really hard to build packages. I don't want to go back to that
it did decrease memory usage though
bright glacier
did you have to build pytorch or tensorflow on 32-bit? I sure hope not
dense jungle
no fortunately not
@bright glacier do you think we can merge https://github.com/psf/black/pull/2644/ and https://github.com/psf/black/pull/2649 ?
bright glacier
I won't have time to check mypyc compatibility but I doubt they'd cause issues. I say go for it if the PRs are good themselves.
I am still a bit nervous about the hacky tokenizer change but I guess we can learn whether it's too hacky the hard way
plain atlas
I could theoretically support 32-bit but that would double the workload on CI an...
except for raspberry pis which are common with python
elder tusk
okay, https://github.com/psf/black/pull/2654 should be good to go! the old regex failed several of the new test cases π
flat krakenBOT
flat krakenBOT
bright glacier
jeez our timing haha
muted haven
How does black's grammar selection work? For example if you return 3 grammars (as [3.9, 3.8, 3.10]) and if it successfully gets parsed with 3.9 grammar, would it try the 3.8 or 3.10?
dense jungle
How does black's grammar selection work? For example if you return 3 grammars (`...
it's ordered, it uses the first one that works
muted haven
Cool! The reason I am asking is that, 3.10 grammar might be probably much slower than the others since we do a lot of pre-calculation on every match/case token (even if the code uses any of these as regular names without pattern matching).
dense jungle
So if we always return the 3.10 grammar first, we'd use it to parse most code
muted haven
So if we put this into the first row, parsing regular code will become slower if they use re.match() etc
bright glacier
The order comes from how black.parsing.get_grammers's return statements are layed out
muted haven
just a FYI about the consideration on enabling this by default.
dense jungle
interesting, do you know how big the slowdown would be?
muted haven
it depends on the program's size
we copy() the whole parser stack on each match occurence
and that really takes a lot of time
plain atlas
How does black's grammar selection work? For example if you return 3 grammars (`...
on another related side, i remember the grammar was talked about not keeping it a lis, and just having the program define one grammar
bright glacier
That's probably not happening imo.
muted haven
I think it should be possible to optimize this a lot (by just copying the last top-level statement, instead of the whole stack) but I still didn't get around looking at it
dense jungle
on another related side, i remember the grammar was talked about not keeping it ...
setting --target-version does this to some extent
bright glacier
We do actually use one grammar file but we have to keep multiple different instances of it with various features enabled / disabled at the runtime level.
muted haven
Cool! The reason I am asking is that, `3.10` grammar might be probably much slow...
Another optimization idea would be specializing the current code for match/case. I initially tried to aim the whole backtracking problem, it got landed black as is. But I think it might also just make sense to implement a hacky solution for match/case where we just check the next few tokens (at best we can locate it in 1 token) manually without the proxy/recorder infrastructure. That would nearly close the speed gap.
But I think it is easier for now to have 3.10 at the last, if you plan to enable it by default at this time.
mental current
Hi
I have a task in my work.. we have a application mobil built with react native and flask as backend... Currently the application mobile is authenticating with Aws cognito service from backend. But my team determinated that this implementation is wrong. They want use AWS lambda like microservice that it can authenticate the user with diferentes providers of authentication such auth0 or cognito. They want have this flexibility when will be necessary. Finally the team too want social login. I have investigated other service such AWS ALB and AWS Api gateway these are others alternatives. I want read us about your opinions. how would you solve this problem? what are your tips?
bright glacier
This seems like the wrong channel to ask in :) #web-development seems like a better fit IMO.
elder tusk
we'd probably get fewer lost strangers if the "open-source projects" section in discord went below "topical chat/help"
https://github.com/psf/black/pull/2572 could use some eyes, if anyone has some to spare π
bright glacier
I CCed ambv since he's the main person who knows the magic trailing comma implementation (and the many many pitfalls involved) well
as evident he's quite busy π¦
errant barn
I find it humorous that fixing that problem mostly involves deleting a bunch of lines π
bright glacier
@dense jungle what do you think of merging psf/black#2472?
toxic stormBOT
bright glacier
I just ran diff-shades over it and here's the results: https://paste.pythondiscord.com/gayetacuze.diff
dense jungle
<@!783088578363523104> what do you think of merging psf/black#2472?
I'm OK with that, the diff looks good
bright glacier
Sounds good!
oh lol this PR is gonna break primer as it causes pyanalyze to be reformatted
wait that's a quora project :p
dense jungle
oh lol this PR is gonna break primer as it causes pyanalyze to be reformatted
working on it as we speak
the project, not the formatting change it needs to make
bright glacier
Oh haha
plain atlas
@bright glacier how did you make this comment??
I tried to make a drop down for a comment with colorized formatting within and i couldn't do it
bright glacier
if you want to name the dropdown you can use the <summary> inline tag
plain atlas
<details>
<summary>
thanks ichard
</summary>
\`\`\`py
import python
\`\`\`
</details>
i legit tried for 15 minutes once to get the markdown parsing within the block
turns out i just needed a new line
bright glacier
and this is why I install the pre-commit hooks :)
plain atlas
yummy commits
dense jungle
why run local tests when you have CI
plain atlas
both are good
bright glacier
it's faster imo
plain atlas
IMHO a commit should always always always be atomic
bright glacier
I don't like context switching between my terminal and the web.
plain atlas
i want a project that runs with every single commit
dense jungle
IMHO a commit should always always *always* be atomic
that's why we squash on merge and clean up all the commits
plain atlas
it may not do everything that it should, but it should run
plain atlas
that's why we squash on merge and clean up all the commits
ah neat
bright glacier
Other than mypyc what should we focus on next? I'm currently trying to deal with the PR backlog pushing 'em forward with diff-shades insight. I suppose I could pick up my goal of moving off pipenv to something friendlier.
plain atlas
I suppose I could pick up my goal of moving off pipenv to something friendlier.
pleasepleasepleaseplease
bright glacier
what's the status on https://github.com/psf/black/pull/2278 by the way?
dense jungle
Yeah we need to make a call on those PRs that change formatting
I think we should move forward with the math people who want x**y to be unspaced
bright glacier
It's roughly 10 ten times as disruptive as that spacing consistency PR I just merged in terms of diff size. I linked the diff-shades report on it -- note it's not perfect as I didn't update my base data after I merged that blank line PR.
dense jungle
It's roughly 10 ten times as disruptive as that spacing consistency PR I just me...
I'm supportive of changing it. Unparenthesized ternaries are confusing to read.
bright glacier
I think we should move forward with the math people who want x**y to be unspaced
Ah yeah this sounds important, the PR looks like it'll require a bunch of work π¦
plain atlas
ternaries?
dense jungle
> I suppose I could pick up my goal of moving off pipenv to something friendlier...
I'm not really sure why we need pipenv in the first place
dense jungle
ternaries?
x if y else z expressions
plain atlas
I'm not really sure why we need pipenv in the first place
btw i wrote an argument against pipenv and stuff yesterday #black-formatter message
dense jungle
this is the only ternary operator in many languages so it's often just called "the ternary operator"
bright glacier
I'm supportive of changing it. Unparenthesized ternaries are confusing to read.
It's awaiting your review FYI - you were requested for review a few months ago π
flat krakenBOT