-
-
Notifications
You must be signed in to change notification settings - Fork 30.9k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
bpo-43312: Functions returning default and preferred sysconfig schemes #24644
Conversation
This PR is stale because it has been open for 30 days with no activity. |
acadc40
to
a0d613f
Compare
Uh, how do I remove the stale label? |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I can't say whether the choices in _get_preferred_schemes
are correct, but they seem reasonable. Unless there's anything we can check them against, I'd say we should go with them.
@@ -517,7 +542,7 @@ def get_path_names(): | |||
return _SCHEME_KEYS | |||
|
|||
|
|||
def get_paths(scheme=_get_default_scheme(), vars=None, expand=True): | |||
def get_paths(scheme=get_default_scheme(), vars=None, expand=True): |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Do we need to call this function every time? Wouldn't be better to call it just once and save it into a private global variable, and use that as the default value for all these methods?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
That’s a separate issue already presented in the previous implementation. I’m not sure this is the best plac eto discuss the possible implications in e.g. multiprocessing settings (I don’t even have enough expertise to comment on it).
As it currently stands, the function is only called twice (get_paths
and get_path
) on import time, so the performance difference is pretty close to none anyway.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Agreed, let's keep the change to the minimal required.
Lib/sysconfig.py
Outdated
@@ -731,7 +756,7 @@ def _main(): | |||
return | |||
print('Platform: "%s"' % get_platform()) | |||
print('Python version: "%s"' % get_python_version()) | |||
print('Current installation scheme: "%s"' % _get_default_scheme()) | |||
print('Current installation scheme: "%s"' % get_default_scheme()) |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Would be a good opportunity to migrate this to f-strings?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Probably, I’ll combine this to other needed changes, if there are any.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I ❤️ it!
Several non-blocking questions attached.
However, I have a very strong suggestion: Please add tests.
Lib/sysconfig.py
Outdated
if scheme not in _INSTALL_SCHEMES: | ||
raise KeyError(key) |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Will the above lookup not raise already? Should this be done before calling _get_preferred_schemes
?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
The main intention is to guard against incorrect implementation, since _get_preferred_schemes()
is expected to be modified by redistributors and implementers. Say an implementer makes a typo:
def _get_preferred_schemes():
return {
"prefix": "posix_prefix",
"home": "posix_home",
"user": "posix_uesr", # Oops!
}
And when the value is later used:
scheme = sysconfig.get_preferred_scheme("user")
... # Much later...
paths = sysconfig.get_paths(scheme)
This check would catch that posix_uesr
is not actually a valid scheme and error early, instead of later when the user actually tries to use the (invalid) scheme.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Can we raise a ValueError with a better message then? "Key '{}' returned scheme '{}' which is not valid on this platform" (or something more correct)
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I couldn’t think of a better message and implemented yours for now, any ideas are very welcomed. Ideally I want to point users to the implementer/redistributor instead of reporting the error to CPython, but that’s probably not really feasible.
This reduces logic duplication with a slight performance tradeoff.
I’ve changed |
The implementation looks good to me. |
I’m wondering whether we should add documentation for the private |
At least a docstring would be great, yes. |
Documentation entry added 👍 Edit: Plus converting %-formatting to f-strings. |
Drive-by refactoring.
Curiosity question: Should distutils use this? |
Ideally yes, but distutils’s implementation around this is structured very differently, and tangled with a lot of things inside |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Would like to see that error message improved, but the rest seems fine to me. I'm just taking on trust that it's sufficient for distros to override what they need, because I don't know that side of it well enough.
Lib/sysconfig.py
Outdated
if scheme not in _INSTALL_SCHEMES: | ||
raise KeyError(key) |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Can we raise a ValueError with a better message then? "Key '{}' returned scheme '{}' which is not valid on this platform" (or something more correct)
Another nudge 🙂 |
I believe all comments have been addressed, and I see no remaining concerns raised and general approval of the change, so I'm going to merge this based on my approval above. |
https://bugs.python.org/issue43312