531930 - Searching for "browser browser" returns unrelated search results

Status: RESOLVED INVALID

(support.mozilla.org :: Search, defect)

Description

[krupa raj[:krupa]]

steps to reproduce:
1.Search for "browser browser"
2.Click on one of the search results and check if the search phrase exists

observed behavior:
The search results do not contain the phrase "browser browser".(A sphinx bug as per jsocol)

Comment 1

[James Socol [:jsocol, :james]]

Searching for "hello -world" (without quotes) should not return the same results as "hello world" (without quotes). We're setting the mode to SPH_MATCH_EXTENDED2, which is supposed to handle phrases and other advanced search queries (like -), but this doesn't seem to be working.

Trevor, the Sphinx cluster is running 0.9.9, right?

Comment 2

[Paul Craciunoiu [:paulc]]

-s are escaped for words with hyphens, before being passed on to Sphinx. We could modify this behavior to escape only -s neighboring no whitespace.

Comment 3

[James Socol [:jsocol, :james]]

What else do we escape? Do we need to escape hyphens? It seems like Sphinx is OK at figuring that out by itself.

Comment 4

[Paul Craciunoiu [:paulc]]

We need to escape hyphens within words. I can't find the bug, but passing words such as "marque-pages" (bookmarks in French) without escaping the hyphen causes Sphinx to look for results with "marque" but not "pages" instead of searching for the entire word. Maybe there's a Sphinx setting for ignoring hyphens within words?

On my local, at least, not escaping hyphens (commenting out line 776 of sphinxlib.php) and searching for "marque-pages" yields no results.

Comment 5

[chizu]

(In reply to comment #1)
> Trevor, the Sphinx cluster is running 0.9.9, right?

Yes, it's 0.9.9-rc2.

Comment 6

[James Socol [:jsocol, :james]]

Thanks Trevor.

Looks like we did find the phrases eventually (stemmed, of course) and since we're escaping hyphens and stripping exclamation points, I think the other features of SPH_MATCH_EXTENDED2 are working correctly.