Drupal.org searches not recognizing phrases (i.e. search terms placed in quotes)

small help, but there is a separate modules seach once you enter the modules area.

Not sure what documentation can do about this is reads more like you want the apache_solr module to do something more than it is doing. Filing an issue with that module, which is running on drupal.org may be a better avenue.

Moving to infrastructure

Blanket statements about how bad search is are not helpful. Specifics are helpful- what you searched for and what you expected to find.

I am not familiar with "cache_content," is this a file included with a module? I am guessing this may be improved by two projects- integrating API search results #84207: Integrate API full text search into the ApacheSolr setup and documenting all contributed modules, which I mention at http://delocalizedham.com/state-of-the-api-module.

"cache_content" is likely the table {cache_content} (which comes from CCK, not Drupal core), so I'm guessing it doesn't even show up on api.d.o in a search.

If you are doing searches about Drupal internals you should be here: http://api.drupal.org

This search does turn up many relevant results: http://drupal.org/search/apachesolr_search/%22cache_content%22
but indeed seems like Solr wants to split on the punctuation into 2 tokens which get stemmed and matched as a phrase, as expected:

http://wiki.apache.org/solr/DisMaxRequestHandler see 'q'

The guts of the search defining the main "query". This is designed to be support raw input strings provided by users with no special escaping. '+' and '-' characters are treated as "mandatory" and "prohibited" modifiers for the subsequent terms. Text wrapped in balanced quote characters '"' are treated as phrases, any query containing an odd number of quote characters is evaluated as if there were no quote characters at all. Wildcards in this "q" parameter are not supported.

It does seem that queries in quotation marks are not being handled correctly. Or maybe just not the way that I would expect.

For example, if I do a search for "report generation" (including quotes) my expectation is that I would see only pages with that exact term, or at least that results with that term would be displayed before any other pages that only have stemmed variations of the component words. Instead, what happens is that the first displayed hit is for an exact match and the rest are a jumble of exact matches and guesses.

If I go to the trouble of putting a search term in quotation marks, it means that I'm trying to exclude those other results (or at least push up the results with an exact match)

Changed title to be more inclusive

Quotes forces a phrase search, not an exact match.

Is it possible to change this behavior?

User expectations for search have been very heavily conditioned by Google, where quotes indicate the searcher is looking for an exact match.

From the Google Help docs "By putting double quotes around a set of words, you are telling Google to consider the exact words in that exact order without any change." http://www.google.com/support/websearch/bin/answer.py?hl=en&answer=136861

Cranking down the 'qs' value gives closer to an exact match (i.e. there can be fewer or no intervening words). So, d.o could change that by adding it to their solrconfig.xml as a query default, or via a module to add it at query time, but Solr's dismax handler doesn't support exact phrase matching. Putting in "solr single" or "single solr" will match the same docs.

So, basically, no. Feel free to use: http://www.google.com/search?q=site%3Adrupal.org

Neil_in_Chicago- yes, that was specific enough to get us started and hopefully we did figure out the root issue. I read the issue as "Search sucks, oh and here is why," where I would like to have seen simply "phrases don't work."

Drupal.org's search has had plenty of problems. We recently put in much more flexible tools, such as Apache Solr, so we can fix many issues if we know what we are looking for. Bug reports are appreciated so we can continue to improve.

Pwolanin- would cranking down qs affect all searches or just phrase searching? Would any affect on all searches be negative or positive?

Checking into this more - the default qs seems to be 0 (or NULL) already, so I'm not sure there is much to do - we could document this - i.e. that "" finds words in immediate proximity, but not exact matches or ordering.

Increasing qs might help some users, but leaving as-is seems fine to me. I think most people will just put in their keywords and never assume that the know the exact string in advance.

We might look at the tokenizer and see if there is some way to avoid splitting on _ or ->, for example, though right now we are limited to using the solr.CharStreamAwareWhitespaceTokenizerFactory since we are using the filter to map accented to non-accented characters. This is probably more relevant for api.d.o searching.

For api.d.o we can also think about having a filter to match exctly on function names, for example.

I have to agree that exact match is what most searchers will expect and it is surprising that solr can easily support this.

http://drupal.org/search/apachesolr_search/%22most%20searchers%22

Seems to work now.