Page MenuHomePhabricator

Lucene search result count should take namespace filtering into consideration
Closed, ResolvedPublic

Description

Author: gangleri

Description:
Hallo!

http://en.wikibooks.org/wiki/Special:Search?search=lrm&fulltext=Search
is displaying 19 results.

It displays also:
*Results 1-20 of 29*
and links as
2 :
http://en.wikibooks.org/w/index.php?title=Special:Search&search=lrm&fulltext=Search&offset=20&limit=20
Next » :
http://en.wikibooks.org/w/index.php?title=Special:Search&search=lrm&fulltext=Search&offset=20&limit=20

Both are empty pages.

Regards Reinhardt [[user:gangleri]]


Version: unspecified
Severity: trivial
URL: http://en.wikibooks.org/wiki/Special:Search?search=lrm&fulltext=Search

Details

Reference
bz3679

Event Timeline

bzimport raised the priority of this task from to Medium.Nov 21 2014, 8:50 PM
bzimport set Reference to bz3679.
bzimport added a subscriber: Unknown Object (MLST).

gangleri wrote:

the links disapeard
It shows *Results 1-20 of 20* but there are only 9 results

Created attachment 1312
Example search returning incorrect numbers/listing.

I don't know if this is connected, but on Wikipedia I've just tried a search,
which returns 'Results 1 to 8 of 8', but only shows a single result. I also
received similar 'incorrect' results for a couple of other searches.

I have attached a screenshot.

(Using firefox 1.5)

Attached:

Search_error_example.jpg (357×1 px, 42 KB)

Created attachment 1313
A second example

Here is a second example. It shows '1 to 20 of 51', but only lists 7 results.
The links for subsequent pages are present, but when clicked they return
"Sorry, there were no exact matches to your query."

Attached:

Search_error_example_2.jpg (713×1 px, 66 KB)

Already known issue. The total count is pre-namespace filtering. Doing namespace filtering in
Lucene is too expensive.

Can we change the text to something like the following:

"Showing items $1 to $2 out of $3 results found across all namespaces. Only
showing results in the '$Namespace' namespace"

If it can't be fixed, then we should acknowledge and work round the problem
rather than ignore it.

  • Bug 2733 has been marked as a duplicate of this bug. ***
  • Bug 5866 has been marked as a duplicate of this bug. ***

rickblock wrote:

If the current page contains no results after namespace filtering, there seems to be no way to get to another page that might
have results. See, for example, http://en.wikipedia.org/w/index.php?title=Special:
Search&search=Francis+Ford+Coppola&fulltext=Search&ns10=1&ns11=1&offset=0&limit=20 . First page has one result,
second page has no results and no way to get to third or other pages (without simply editing the URL).

  • Bug 8311 has been marked as a duplicate of this bug. ***

rainman wrote:

*** Bug 9775 has been marked as a duplicate of this bug. ***

ekb87ds02 wrote:

Can't you just fetch 21 results, only show the 'next' button if the 21st result
exists and not display either that result or the total count (or use language
like 'Approximately')?

21? Or 22? Or 23? Or 23,000? How many will you have to fetch to get 20 viewable
results? There's no way to predict that ahead of time.

rainman wrote:

I think ekb87ds02 was suggesting to fetch 21 viewable results, and then show 20,
and interpret the remaining one as "there are more results". This is fine,
however, with some namespace filter caching we could actually get the exact count.

rainman wrote:

Fixed in Lucene Search 2. The queries are always either filtered by cached filters, or rewritten so that they return the articles from given namespaces. So, the number of hits should always be correct.

[Merging "MediaWiki extensions/Lucene Search" into "Wikimedia/lucene-search2", see bug 46542. You can filter bugmail for: search-component-merge-20130326 ]