reflection: move signature union-splitting logic under the control of inference #22144
Conversation
vtjnash
force-pushed
the
jn/method-union-splitting
branch
from
3489694 to
7529a15
Compare
test/replcompletions.jl
Outdated
| @@ -457,7 +457,6 @@ try | |||
| end | |||
| finally | |||
| @test pop!(LOAD_PATH) == path | |||
| rm(path, recursive=true) | |||
There was a problem hiding this comment.
the parent dirs get created. and what does this have to do with the rest of this pr?
There was a problem hiding this comment.
This test is literally for the case "even though the path do no exist" (sic)
It doesn't, but it was tripping up my attempts at testing locally so it ended up on all my branches today.
There was a problem hiding this comment.
only the first call, before mkpath gets run, and files get created under it. something else was happening in your case, open a separate issue instead of putting it into multiple unrelated pr's please
vtjnash
force-pushed
the
jn/method-union-splitting
branch
from
7529a15 to
8ea4f16
Compare
| xapplicable = _methods_by_ftype(sig_n, sv.params.MAX_METHODS, sv.params.world, min_valid, max_valid) | ||
| xapplicable === false && return Any | ||
| append!(applicable, xapplicable) | ||
| end |
There was a problem hiding this comment.
Is it worth to keep track of how many different methods the entries in applicable actually refer to and bail out if that is more than sv.params.MAX_METHODS, too?
There was a problem hiding this comment.
probably, for now I'm just moving the existing logic rather than trying to improve on it much
|
👍 Makes the code clearer by moving things where they belong. (Also a good idea to break #21933 into smaller, more digestible pieces.) |
It just makes more sense and is more useful for inference to have control over the union-splitting.
edit: I split a function, so https://github.com/JuliaLang/julia/pull/22144/files?w=1 is a bit easier to read