Return highlighted excerpt of result

[tannerwelsh]

Addresses issue #100.

If configuration is set to

:using => {
  :tsearch => { :highlight => true }
}

then queries will return records with an additional pg_highlight attribute.

Currently uses the defaults set by the Postgres function ts_headline, as specified here: http://www.postgresql.org/docs/current/static/textsearch-controls.html#TEXTSEARCH-HEADLINE

Also supports custom delimiters

:using => {
  :tsearch => { 
    :highlight => { 
      :start_sel => "begin", 
      :stop_sel => "end" 
    }
  }
}

will override the default <b> and </b> query delimiters in the highlighted results.

NOTE: the pg function used (ts_headline()) is not supported on pg versions < 9.0. I know that the gem is intended to support 8.0 and above, but this seemed like a useful enough feature get started. Any way to work around?

[tannerwelsh]

So the build is failing because Rubocop wants class definitions to have <= 100 lines of code, and /lib/pg_search/features/tsearch.rb is 107. :/

I could resolve by either:

1. Extracting into a module
2. Changing the Rubocop config

Do you have a preference @nertzy?

[nertzy]

It would be better to refactor some of the methods into service objects. For instance, tsquery_for_term is way too complicated.

But for now it would be OK with me to add a Rubocop comment that turns off the file size check for just tsearch.rb until we can get a good refactor in. Just add a line to the top of the file that turns off that rule:

# rubocop:disable Metrics/ClassLength

[marshall-lee]

This is awesome! 👍

[tannerwelsh]

Thanks, @nertzy. I'd rather do it right the first time, so I'll refactor them out as suggested.

[tannerwelsh]

Ok, so here's the deal. I cleaned up this PR, added documentation, and added a workaround to appease Rubocop.

Then, in a separate branch, I did some serious refactoring.

I didn't want to bundle all of it into this one PR, so we can do it in two phases:

1. If this PR looks good, then you can merge and the new feature will be available.
2. I'll make another PR with the refactorings so that the code is actually a little cleaner (not perfect, but better).

What do you think, @nertzy? Will that work for you?

[nertzy]

I'm sorry that I've taken so long to address this pull request. Unfortunately, the new subquery structure in pg_search 1.0 breaks this pull request. I spent a few minutes trying to refactor what you've built to act similarly to how .with_pg_search_rank works, but to no avail.

So I'm going to close this for now. I'd gladly accept a pull request based off of this that adapts things to the new subquery, and that makes the highlight optional through something like a .with_pg_search_highlight method chained onto the scope.

[tannerwelsh]

Hey @nertzy no worries, I got pulled in other directions too. Sadly this is not a high priority for me anymore, so don't think I'll be able to work on the feature anytime soon.

Thanks for the helpful suggestions and feedback earlier though! Hope that #280 is able to make it through by building on this one. 😄

[olivierlacan]

@nertzy Was documentation for this feature merged into the master branch README by mistake? I keep trying to use with_pg_search_highlight on a search scope and the method always comes up as undefined.

[olivierlacan]

By the way I'm referring to https://github.com/Casecommons/pg_search/blob/ff1af346322ca54808b3f53081149b8cac55bb76/README.md#highlight-postgresql-90-and-newer-only

The recommended configuration causes the following error:

ArgumentError: Unknown key: :start_sel.

[nertzy]

It's not an error. It's merged to master but not yet in a released gem version, I believe. I'll try to get a new version cut very soon, sorry for the confusion!

[olivierlacan]

Oh, fantastic. Thank you.

[sintro]

There are still error in documentation (the sample content is:

:using => {
  :tsearch => { 
      :start_sel => "begin", 
      :stop_sel => "end" 
  }
}

but not

:using => {
  :tsearch => { 
    :highlight => { 
      :start_sel => "begin", 
      :stop_sel => "end" 
    }
  }
}

)
Also I found that there is error when trying to use with_pg_search_highlight method in combination with associated_against, the error description is something like

table "pg_search_3be97ef13e109a88624534" is absent in FROM-clause 

[heisters]

@sintro I fixed the documentation bug in #312