use typed_kwargs for Dependency
#12057
Open
Conversation
This file contains 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
dcbaker
force-pushed
the
submit/typed-kwargs-dependency
branch
from
0e8c1f7
to
4c57557
Compare
tristan957
approved these changes
Jul 31, 2023
|
Looks like this needs a rebase on master |
dcbaker
force-pushed
the
submit/typed-kwargs-dependency
branch
10 times, most recently
from
7de47e7
to
016f303
Compare
|
Ready for another review? |
|
Yes please |
tristan957
reviewed
Dec 5, 2023
dcbaker
force-pushed
the
submit/typed-kwargs-dependency
branch
from
016f303
to
be4a5cc
Compare
|
I know I didn't get to all of your comments @tristan957, but I think I hit the majority of them |
|
Works for me. This is gonna be a great addition to the codebase! |
dcbaker
force-pushed
the
submit/typed-kwargs-dependency
branch
4 times, most recently
from
d323faa
to
f724dc7
Compare
…used" This reverts commit 93c11f2.
This also allows simplifying the DependencyFallbacksHolder.set_fallback a little, since we now know for sure we'll have a list of 0, 1, or 2 elements. It also fixes an incorrect FeatureNew, which said that the single element array/string value was since 0.53; it's actually since 0.54
I'm not sure what validation to do where. Which methods a particular dependency allows are dependency specific, but there are a set of valid ones, and a set of totally deprecated ones/new ones added that it would be good to emit featurechecks for.
I've made native required, not optional. This is required to ensure that sub-dependencies have the same machine target as the parent.
So we can get proper since warnings
We do some injection of values in some cases that we don't allow from the command line, this adds them and allows us to get rid of the rest of our type: ignore comments
Which gets us super close to having this all type safe
We have several values that are deprecated or added, that are currently not handled due to needing access to the Environment. This can be easily fixed by handling them in the interpreter
Since type checkers now view MisingCompiler as a Compiler, and it provides a method to determine that it is missing (language == 'missing'), we don't need to expose it anywhere, it's just an implementation detail of the dependencies.base module.
dcbaker
force-pushed
the
submit/typed-kwargs-dependency
branch
from
46eb3e7
to
bde51f3
Compare
|
If you rebase this, I can give it one more review and approval. Sorry it fell off my radar. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This is a rather length series (though I've tried to structure it in such a way that we can land small pieces at a time), that converts the
dependency()function to use typed_kwargs, and makes some cleanups along the way.I've made a deviation from the way we handle TypedDict elsewhere, and that is that I've used the
total=Falseargument. What I found when I tried not doing that, is that there is such a tight coupling betweenbuild.Dependency()and the Interpreter layer is that any internal Dependency instantiation became very verbose. A followup that might be able to solve this is to to switch Dependency to use all named arguments, such that reasonable defaults are always available. As such we still have to use a great deal ofkwargs.get(), but it does allow for better type checking.I've also converted the dependency cache to use a distinct class for the key, as I found the old approach to give different results when a keyword argument was undefined vs defined to it's default value, which shouldn't be true.
This also gets the
interpreter/dependencyfallbacks.pymodule very close to being type safe.I've opened this as a draft because I'm sure there are still failing tests, but cursory reviews are still welcome.