I don't have a lot of confidence that the data being collected is what
needs to be collected in each of these cases, but I believe that I've
correctly identified the visit methods that need to be collecting data.
I think we need to have a test that ensures that the keys being used in
this table generator are the same as the keys being used by the
relevance feature computer (as computed by the in-scope pass). I don't
know exactly what that test would look like, but that would ensure that
(a) the keys are the same and (b) the table generator isn't missing
keys being used by the completion engine or collecting data that isn't
being used.
I also fixed a bug that prevented this generator from running and tested
to ensure that it could run over several example packages.
Change-Id: I0c0cff148d288942e39ee876e89cdf9d1468f59d
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/473780
Commit-Queue: Brian Wilkerson <brianwilkerson@google.com>
Reviewed-by: Samuel Rawlins <srawlins@google.com>
Brian Wilkerson
f92ea3d58d