Fix #3252: Change logic in adaptNoArgs #3254
Conversation
adaptNoArgs would do an implicit argument search only if the result type conformed to the expected type. scala#3252 shows that this needs to be revised.
Split into more manageable sub-methods
|
test performance please |
|
performance test scheduled: 1 job(s) in queue, 0 running. |
|
To review this it's best to look at the first commit in isolation, which is reasonably small. The rest is just moving code around. |
|
Performance test finished successfully: Visit http://dotty-bench.epfl.ch/3254/ to see the changes. Benchmarks is based on merging with master (443bdb1) |
| case wtp: ExprType => | ||
| adaptInterpolated(tree.withType(wtp.resultType), pt) | ||
| case wtp: ImplicitMethodType | ||
| if constrainResult(wtp, followAlias(pt)) || !functionExpected => |
There was a problem hiding this comment.
Does the order of the checks matter here? I see that constrainResult generates new constraints if it returns true.
If the order does not matter, then it should be less expensive to first check !functionExpected.
There was a problem hiding this comment.
Good observation. The order does matter, actually. constrainResult strenghtens the constraint if it succeeds so we want to always run it.
adaptNoArgs would do an implicit argument search only if the result type
conformed to the expected type. #3252 shows that this needs to be revised.