Immutable version of Packager::runIncremental #5596
Conversation
|
We have a policy of testing changes to Sorbet against Stripe's codebase before Stripe employees can see the build results here: → https://go/builds/bui_LShtEgPKFz5fbf |
froydnj
left a comment
There was a problem hiding this comment.
Nice catch, I keep forgetting about package magic.
packager/packager.h
Outdated
| template <typename StateType> | ||
| static std::vector<ast::ParsedFile> runIncremental(StateType &gs, std::vector<ast::ParsedFile> files); |
There was a problem hiding this comment.
Slight preference for being explicit here with two different declarations of runIncremental and leaving the template magic in the .cc file: it's easier to catch mistakes in client code that way (e.g. the current state would happily accept you passing in a unique_ptr<GlobalState>...or a MutableCtx, which would normally auto-convert to GlobalState, but not here. you wouldn't see an error until link time).
(The proposed scheme here is also inconsistent with namer and resolver: the non-mutating versions of those have different names, whereas packager would not. I tend to think having identical names is not a good thing for what we're trying to do, even if C++ is happy to overload on const qualification.)
90d672b to
b7e9e8b
Compare
Honestly behaves like a hybrid of namer ("drop trees we don't want") and
resolver ("use templates to do one thing or another conditionally").
b7e9e8b to
ad505ab
Compare
Commit summary
Add a test that fails before these changes (da637be)
pre-work: Refactor Resolve::runIncremental into two explicit functions (e016114)
pre-work:
...WithoutStateMutation->...BestEffort(343d52a)Immutable version of Packager::runIncremental (ad505ab)
Honestly behaves like a hybrid of namer ("drop trees we don't want") and
resolver ("use templates to do one thing or another conditionally").
Motivation
Honestly behaves like a hybrid of namer ("drop trees we don't want") and
resolver ("use templates to do one thing or another conditionally").
For posterity's sake, here are those two previous PRs: #5498, #5532.
Test plan