QueryParser handles explicit group+operator badly

[Jared Camins-Esakov]

Although the following four queries should be equivalent:
george || fred
(george) || fred
george || (fred)
(george) || (fred)

The Pg QueryParser driver generates SQL queries that return different numbers of results for each of those queries, with all desired results appearing only with the first query. The problem seems to be in the code that generates the initial query in TPAC and JSPAC. A patch for TPAC will follow.

[Jared Camins-Esakov (jcamins)]

A patch for this can be found in the working repository in the branch user/jcamins/bug_1044721

[Lebbeous Fogle-Weekley (lebbeous)]

Hi Jared.

Wow, I didn't expect an effective workaround to be that simple, but that does indeed seem to solve the problem in the TPAC. Possibly somebody will want to fix the QueryParser Pg driver so that this grouping workaround isn't necessary in the future, but I don't see what this hurts.

This works just fine in my tests, so I'm pushing to master and backporting through rel_2_2.

[Lebbeous Fogle-Weekley (lebbeous)]

Reopening this on upon reports of facets being broken by it, and possibly other side effects. :-(

Here's a reversion commit to be pushed to master and backported through rel_2_2 once signed-off by another:
http://git.evergreen-ils.org/?p=working/Evergreen.git;a=shortlog;h=refs/heads/collab/senator/revert_1044721

[Dan Scott (denials)]

Confirmed that this reversion resolves the facet searching problems on our system.

I've signed off at user/dbs/revert_1044721 in working.

[Ben Shum (bshum)]

Hmm, was committed to master, rel_2_3 and rel_2_2 already. Marking new statuses

[Ben Shum (bshum)]

Gah, just realized right after I did that, that I've read this backwards. Not fixed, needs re-testing now that new QP is on the scene?