Implement native in filter

[navis]

Currently, InDimFilter is translated to "or + selector filters". Value matcher can use hash set for faster filtering.

[b-slim]

can we add check for null ?

[navis]

values cannot be null and Strings.nullToEmpty(input) also cannot be null.

[gianm]

values may contain null, this should match "" but with the current code it will not. Something should be translating null to empty within values.

[fjy]

you can just pass in empty string

[navis]

wanted to look same with line 69. I'll address that.

[nishantmonu51]

is it a allowed for values to be empty set ?
should we consider empty set as equivalent to filtering against null ?

[navis]

could be regarded false for all values imo, but seeing it had thrown exception in OR filter in runtime in original implementation, it would be better to add precondition in InDimFilter not to accept null or empty.

[fjy]

👍

[nishantmonu51]

👍 after travis, @navis please squash commits.

[navis]

@nishantmonu51 squashed

[nishantmonu51]

Ah, It has merge conflicts. @navis can you rebase with the latest master ?

[gianm]

IMO the old signature was better- it enforces at least one value must be passed in. an IN () with empty list seems strange to me and not very meaningful.

[gianm]

I believe the old impl would actually have thrown an exception in this case. It would have tried to construct an empty OrFilter, which isn't allowed.

[gianm]

ah I see you're checking this in InDimFilter's constructor. In that case why not keep value, values....

[navis]

reverted that

[gianm]

would be nice to see some additional tests too, especially tests involving null handling.

[gianm]

Looks good other than some minor comments and a desire for tests.

[gianm]

👍, looks good once the tests can compile.

BaseFilterTest was adjusted in #2753 such that it could also test filtered aggregators; the tests need to call it a little differently since then.

[navis]

rebased to master and squashed

[gianm]

👍