Closed jdelStrother closed 2 years ago
Well that's frustrating indeed, but well done on all the debugging!
I have previously managed to get the TS test suite green with Sphinx 3.4, but the batched search testing doesn't look at metadata - so essentially: this isn't a known limitation nor something I've run into previously.
I wonder if it's worth reporting as a bug? - though I must admit, the inconsistent communication around Sphinx (and the fact it's now closed source) is a bit concerning. I'm not sure if it's worth giving Manticore a try as a drop-in replacement given the development there seems more active and open (though I'm still working on getting TS tests passing with v4.2).
though I must admit, the inconsistent communication around Sphinx (and the fact it's now closed source) is a bit concerning.
yeah, likewise. We'll see what they say - http://sphinxsearch.com/bugs/view.php?id=2751
@jdelStrother just catching up on open-source chores - I'm going to close this issue, given the bug seems to be with Sphinx itself. I'd be curious as to whether you can reproduce this issue with Manticore though!
Hi there - I noticed that our batched queries have started behaving oddly. Something like this:
would give the same total_entries for each search term.
I think I've reduced the problem down to this, which is essentially what ThinkingSphinx is doing in ThinkingSphinx::Connection::MRI :
which prints:
Note the same keyword appearing in both META results. It appears as though given a multi-statement query LIKE
SELECT ...; SHOW META; SELECT ...; SHOW META
, the results returned for both SHOW METAs are that of the last SELECT statement.I'm pretty sure this wasn't always the case. I can reproduce the problem against Sphinx 3.3.1 & 3.4.1, but not against an aged 2.2.11 server.
Seems hard/impossible to fix for batched queries, but wondered if it was a known limitation or if you'd run into it before.