The high weight on semantic context turns out to give a lot of
unreasonable completions on common types. For example
[0].ma<here>
was suggesting withUnsafeMutable... over map.
rdar://problem/27393776
... in the experimental code-completion path. They seem to be generally
useful as a baseline result still as long as they are guaranteed to be
the first result.
rdar://problem/25177968
Experiment with having some locals or nominal members come first, but
only a few so that the literals are still predictably near the top.
Part of rdar://problem/23865118
If there is a type context, hide literal suggesetions that don't match
it, unless they are keywords and we have filtered to their names.
Incidentally fix an output buffering issue when combining filtering with
the -raw flag in complete-test.
Part of rdar://problem/23865118
On a per-request basis. Allows hiding/showing at multiple granularities
* everything
* module
* API by name
* keywords (by uid or all of them)
* literals (by uid or all of them)
With more specific rules overriding less specific ones (so you can hide
everything and then selectively show certain API for example).
rdar://24170060
The code goes into its own sub-tree under 'tools' but tests go under 'test',
so that running 'check-swift' will also run all the SourceKit tests.
SourceKit is disabled on non-darwin platforms.
