fix duplicate results when search for *

[CsCherrYY]

fix #2938

[rgrunber]

I think if we really need to fix it this way, we could but there's one regression that might be introduced. The problem is you can't really see it on the VS Code side because of microsoft/vscode#149144 .

Something like java.util.function.* returns results from the LS, but VS Code would fail to display anything. With this change, the above query returns empty results in the LS also. To me, the query should return something in an ideal world 😛 .

Is there any reason this issue can't be solved by just making a temporary symbol set, and then converting to a symbol list (to satisfy the API)

[CsCherrYY]

@rgrunber you're right. The current fix will return an empty array. Why I didn't use a temporary set is I'd like to reduce an unnecessary search, but it's now tricky to set the condition to "exclude" the search.

I would update this commit in the set/list way.

[rgrunber]

LGTM. I would just confirm that test failure in WorkspaceSymbolHandlerTest was just a random failure. Couldn't reproduce locally. Feel free to squash+merge after.

[rgrunber]

I think this cause some issue in https://ci.eclipse.org/ls/job/jdt-ls-master/1625/testReport/ . at WorkspaceSymbolHandlerTest.testProjectSearch .

[CsCherrYY]

I think this cause some issue in https://ci.eclipse.org/ls/job/jdt-ls-master/1625/testReport/ . at WorkspaceSymbolHandlerTest.testProjectSearch .

I can't reproduce it locally. Is it related to the platform?

[rgrunber]

I wasn't able to reproduce in an IDE, but if I run with

$ mvn clean verify -DtestClass=org.eclipse.jdt.ls.core.internal.handlers.WorkspaceSymbolHandlerTest -DfailIfNoTests=false

Failures: 
  WorkspaceSymbolHandlerTest.testProjectSearch:119 expected:<[java]> but was:<[org.sample]>

Tests run: 17, Failures: 1, Errors: 0, Skipped: 0

Update: : I see the test does a search and gets the first entry .. maybe we've disturbed the ordering. Might be a simple fix to make the test more resilient.

It's an easy fix. There are java.IFoo & org.sample.IFoo . We get the first entry and assume it'll be the java package but we make no guarantees about ordering at this point so either is valid. I'll just fix the test to check for both.