Use a common configuration model #374
Conversation
nbraud
force-pushed
the
remolisher/config
branch
from
053f44d to
b74a405
Compare
|
Ah, |
|
Would it not be amasing if tests showed the entire error output? |
nbraud
force-pushed
the
remolisher/config
branch
2 times, most recently
from
c385c6f to
a78e908
Compare
nbraud
force-pushed
the
remolisher/config
branch
from
892bade to
ad62d1b
Compare
|
Ah fuck, the zipapp bootstrapping fails because |
|
@duckinator I could turn all the models into dataclasses, and write our own more-basic input validation layer... but I'm honestly unsure whether we should prioritize Bork itself being available as a zipapp, or maintainability. |
nbraud
force-pushed
the
remolisher/config
branch
4 times, most recently
from
1187dfd to
3d63953
Compare
nbraud
force-pushed
the
remolisher/config
branch
from
3d63953 to
dd796f6
Compare
|
@duckinator that's ready-to-review and CI is green 🎉 All in all I'm pretty happy, we are getting to a point where type-checking is able to find bugs (whereas before, too much of the codebase was untyped and |
|
@nbraud sorry, I somehow got distracted from looking at this until today. I like this overall, but I'd like some kind of test for building ZipApps (even if not of Bork). |
This should capture the entirety of Bork's runtime configuration, with the exception of CLI arguments (for now)
That's an extra zipapp-happy testcase.
nbraud
force-pushed
the
remolisher/config
branch
from
936a40a to
883869f
Compare
Done! |
Model the configuration with Pydantic, and convert the rest of Bork to use that rather than
load_pyprojectand raw dicts.This shouldn't change the actual configuration schema, merely document and validate it.
Rationale
Previously, each use of a config entry had its own idea about that
+37 -60 lines, excluding the model's implementation and removal of
load_pyprojectDrawbacks
Pydantic is a new dependency, with
pydantic_coreweighing ~2 MiB.I don't think it matters all that much: Bork is a development tool, and its latest release was already 5 MiB.