Show other results from GOV.UK in a scoped search #784

alicebartlett · 2015-04-09T13:25:01Z

Pivotal: https://www.pivotaltracker.com/story/show/

What
This PR exposes the top three search results from the whole of GOV.UK in a search that has been scoped to a single document.

Before:

After:

Summary of commits:

6c3d93f, dc90a5e, f9c24d7 and 318731b (the first three commits) are refactoring done as part of the prep work for adding this feature.
5b7049b modifies search_api.rb to add an the additional call to rummager for three GOV.UK wide search results. Tests for this request are also included in this commit.
804af74 refactors out scoped search functionality to a new class ScopedSearchResultsPresenter and adds tests for this new presenter.
e4c0154 is where the bulk of the changes are. This commit takes the new search results returned by the changes to search_api.rb in 5b7049b, and formats them properly and displays them in the view.
248d2e8 is CSS for the new results pull out.

How to review this PR
I've done quite a lot of hindsight based refactoring so each commit should make sense in isolation. Review however you prefer.

Who should review this PR

Anyone in the core team is especially encouraged to pile in.
I paired with @evilstreak, @jennyd, and @binaryberry on different bits of this at various times.
@rivalee / @markhurrell if you could check you're happy with the design and the micro-copy is what was agreed with the content team, I know it changed quite a bit and I'm not sure I got the right stuff in the end.
Anyone else with an interest in search/rummager/manuals

`count` should be a string not an integer, though there should only ever be 1 result when searching by link anyway. Add a trailing comma for cleaner diffs(tm).

Given this is only being used for 3 things and I just spent quite a while getting lost in the meta-programming while debugging something, I've unpacked this to non-metaprogrammed methods to save the next person who comes to this some time.

We're about to do a lot of fiddling around with results, to keep the code nice and clean, extract `result` to a single template.

binaryberry · 2015-04-09T13:36:04Z

This PR is too advanced for me to review but 👍 on the detailed commit messages!

edds · 2015-04-09T13:53:52Z

I have had a quick look and this looks great to me. Thanks for taking the time to make the commits really easy to follow.

rboulton · 2015-04-09T14:42:11Z

I'm very excited to see this, it looks shiny. ✨.

benilovj · 2015-04-09T15:10:18Z

test/unit/presenters/scoped_search_results_presenter_test.rb

+      end
+
+      # iterate remaining results
+      @scoped_results[3..-1].each_with_index do | result, i |


this bit of the test is quite involved. I wonder whether it'd be more obvious if you checked the actual (hard-coded) values instead of variables.

The hard-coded values are more like:

<[{:attributes=>[], :debug_score=>1, :description=>nil, :display_link=>nil, :es_score=>nil, :examples=>nil, :examples_present?=>false, :external=>false, :format=>nil, :formatted_section_name=>nil, :formatted_subsection_name=>nil, :formatted_subsubsection_name=>nil, :link=>nil, :section=>nil, :suggested_filter_link=>nil, :suggested_filter_present?=>false, :suggested_filter_title=>nil, :title=>"scoped_result_1"}, {:attributes=>[], :debug_score=>1, :description=>nil, :display_link=>nil, :es_score=>nil, :examples=>nil, :examples_present?=>false, :external=>false, :format=>nil, :formatted_section_name=>nil, :formatted_subsection_name=>nil, :formatted_subsubsection_name=>nil, :link=>nil, :section=>nil, :suggested_filter_link=>nil, :suggested_filter_present?=>false, :suggested_filter_title=>nil, :title=>"scoped_result_2"}, {:attributes=>[], :debug_score=>1, :description=>nil, :display_link=>nil, :es_score=>nil, :examples=>nil, :examples_present?=>false, :external=>false, :format=>nil, :formatted_section_name=>nil, :formatted_subsection_name=>nil, :formatted_subsubsection_name=>nil, :link=>nil, :section=>nil, :suggested_filter_link=>nil, :suggested_filter_present?=>false, :suggested_filter_title=>nil, :title=>"scoped_result_3"}, {:is_multiple_results=>true, :results=> [{:attributes=>[], :debug_score=>1, :description=>nil, :display_link=>nil, :es_score=>nil, :examples=>nil, :examples_present?=>false, :external=>false, :format=>nil, :formatted_section_name=>nil, :formatted_subsection_name=>nil, :formatted_subsubsection_name=>nil, :link=>nil, :section=>nil, :suggested_filter_link=>nil, :suggested_filter_present?=>false, :suggested_filter_title=>nil, :title=>"unscoped_result_1"}, {:attributes=>[], :debug_score=>1, :description=>nil, :display_link=>nil, :es_score=>nil, :examples=>nil, :examples_present?=>false, :external=>false, :format=>nil, :formatted_section_name=>nil, :formatted_subsection_name=>nil, :formatted_subsubsection_name=>nil, :link=>nil, :section=>nil, :suggested_filter_link=>nil, :suggested_filter_present?=>false, :suggested_filter_title=>nil, :title=>"unscoped_result_2"}, {:attributes=>[], :debug_score=>1, :description=>nil, :display_link=>nil, :es_score=>nil, :examples=>nil, :examples_present?=>false, :external=>false, :format=>nil, :formatted_section_name=>nil, :formatted_subsection_name=>nil, :formatted_subsubsection_name=>nil, :link=>nil, :section=>nil, :suggested_filter_link=>nil, :suggested_filter_present?=>false, :suggested_filter_title=>nil, :title=>"unscoped_result_3"}]}, {:attributes=>[], :debug_score=>1, :description=>nil, :display_link=>nil, :es_score=>nil, :examples=>nil, :examples_present?=>false, :external=>false, :format=>nil, :formatted_section_name=>nil, :formatted_subsection_name=>nil, :formatted_subsubsection_name=>nil, :link=>nil, :section=>nil, :suggested_filter_link=>nil, :suggested_filter_present?=>false, :suggested_filter_title=>nil, :title=>"scoped_result_4"}]>.

which is why I've written it this way, though I understand your point. Do you think this is clearer:

expected_results_list = [{ "title"=> "scoped_result_1" }, { "title"=> "scoped_result_2" }, { "title"=> "scoped_result_3" }, { "is_multiple_results" => true, "results" => [{ "title"=> "unscoped_result_1" }, { "title"=> "unscoped_result_2" }, { "title"=> "unscoped_result_3" }, ] }, { "title"=> "scoped_result_4" }, ] results. # Scoped results expected_results_list[0..2].each_with_index do | result, i | assert_equal result["title"], results[:results][i][:title] end # Check un-scoped sub-list has flag assert_equal true, results[:results][3][:is_multiple_results] # iterate unscoped sublist of results expected_results_list[3]["results"].each_with_index do | result, i | assert_equal result["title"], results[:results][3][:results][i][:title] end # iterate remaining results expected_results_list[3..-1].each_with_index do | result, i | assert_equal result["title"], results[:results][i+4][:title] end end

rivalee · 2015-04-09T15:45:29Z

Design looks good to me and the copy is correct!

boffbowsh · 2015-04-10T13:03:32Z

app/presenters/scoped_search_results_presenter.rb

+    search_response["scope"]["title"]
+  end
+
+  def presentable_result_list


I wonder if this would be better as an override of the results method?

def results presentable_result_list = super # possibly a better local var name than this, but I'm currently unable to brain if unscoped_results.any? insertion_point = [results.count, 3].min unscoped_results_sublist = { results: unscoped_results, is_multiple_results: true } presentable_result_list.insert(insertion_point, unscoped_results_sublist) end presentable_result_list end

Then you wouldn't need to change results in to_hash above, and intent is clearer.

Aaaaactually, I'm already overwriting results in this class to only dish out ScopedResults which I don't want the super to have to know about, so this won't work. There's probably still a clearer way though, want to come pair on it when you're free?

boffbowsh · 2015-04-10T13:13:35Z

Pedantic style comments, but otherwise seems to do the trick 👍

When users search within a manual (this is doing a "scoped" search) we want to show them some results from the rest of GOV.UK (ie some "un-scoped" results) so that they are aware that here are other results available that might be useful. This commit adds a third rummager request for scoped searches which returns three un-scoped results. This request is built by removing the scoping `filer_manual[]=this/manual` and adding an extra parameter `reject_manual[]=this/manual` which tells rummager we don't want any results that occur in this manual.

Since out code for scoped searches is getting a bit more involved with the inclusion of un-scoped results, let's extract this functionality to a new presenter `scoped_search_results_presenter`. This commit just extracts the existing functionality into a new presenter, no new features are added.

This commit: - adds methods to `scoped_search_results_presenter` to show unscoped results from rummager in the view. - Amends `search_results_presenter` to put the `.to_hash` call inside the `.map` on the result builder. So now whenever `results` is called the returned array is read to be sent to the view. This is because we call `results` from `scoped_search_results_presenter` now too. - Adds a template for unscoped results - adds tests for `scoped_search_results_presenter` `presentable_result_list` is the starting point for the biggest change here. I'm not sure the implementation choices here are very obvious so I'll explain them. The design (results with an embedded list of other results) is difficult to represent semantically in HTML. Ideally the extra results would be in an `<aside>` tag outside of the result list, but the design has them interrupting the list so that sighted users will notice them. A compromise that I'm happy with is to have them nested in the list of results like so: ``` <li>result 1</li> <li>result 2</li> <li>result 3</li> <li>More results from GOV.UK <ol> <li>..</li> <li>..</li> <li>..</li> </ol> </li> <li>result 4</li> etc ``` Given this markup, and the constraint of using a logic-less templating language (mustache) the best way I could thing to achieve this HTML was to pass mustache an object that contains this structure. The responsibility for mashing up the scoped results and unscoped results falls to `presentable_result_list`. This software pattern is a bit uncomfortable because it pushes the design of the interface ("the unscoped results should be nested") onto the presenter, though it should be the view's responsibility. However it is the only way I can think to write it.

alicebartlett · 2015-04-13T09:54:52Z

OK I think I've responded to all the comments now, thanks everyone!

boffbowsh · 2015-04-13T09:57:40Z

Looks good to me 👍

evilstreak · 2015-04-13T10:02:06Z

👍

Show other results from GOV.UK in a scoped search

Alice Bartlett added 4 commits April 9, 2015 12:24

count should be a string / trailing comma

6c3d93f

`count` should be a string not an integer, though there should only ever be 1 result when searching by link anyway. Add a trailing comma for cleaner diffs(tm).

Unpack metaprogramming

dc90a5e

Given this is only being used for 3 things and I just spent quite a while getting lost in the meta-programming while debugging something, I've unpacked this to non-metaprogrammed methods to save the next person who comes to this some time.

fix typo

f9c24d7

Extract result to separate view

318731b

We're about to do a lot of fiddling around with results, to keep the code nice and clean, extract `result` to a single template.

alicebartlett force-pushed the surface-other-results branch 2 times, most recently from 73f3d42 to 0db47e9 Compare April 9, 2015 13:42

benilovj reviewed Apr 9, 2015
View reviewed changes

alicebartlett force-pushed the surface-other-results branch 2 times, most recently from 9fe22cf to 4c2d431 Compare April 9, 2015 16:49

boffbowsh reviewed Apr 10, 2015
View reviewed changes

Alice Bartlett added 2 commits April 10, 2015 15:39

alicebartlett force-pushed the surface-other-results branch from 4c2d431 to 3c05577 Compare April 10, 2015 14:39

Alice Bartlett added 2 commits April 10, 2015 16:00

Styles for unscoped results pullout

a933fc9

alicebartlett force-pushed the surface-other-results branch from 3c05577 to a933fc9 Compare April 10, 2015 15:01

evilstreak added a commit that referenced this pull request Apr 13, 2015

Merge pull request #784 from alphagov/surface-other-results

1387536

Show other results from GOV.UK in a scoped search

evilstreak merged commit 1387536 into master Apr 13, 2015

evilstreak deleted the surface-other-results branch April 13, 2015 10:02

evilstreak mentioned this pull request Apr 21, 2015

Cache compiled templates for partials mustache/mustache#205

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Show other results from GOV.UK in a scoped search #784

Show other results from GOV.UK in a scoped search #784

alicebartlett commented Apr 9, 2015

binaryberry commented Apr 9, 2015

edds commented Apr 9, 2015

rboulton commented Apr 9, 2015

benilovj Apr 9, 2015

alicebartlett Apr 9, 2015

rivalee commented Apr 9, 2015

boffbowsh Apr 10, 2015

alicebartlett Apr 10, 2015

boffbowsh commented Apr 10, 2015

alicebartlett commented Apr 13, 2015

boffbowsh commented Apr 13, 2015

evilstreak commented Apr 13, 2015

Show other results from GOV.UK in a scoped search #784

Show other results from GOV.UK in a scoped search #784

Conversation

alicebartlett commented Apr 9, 2015

binaryberry commented Apr 9, 2015

edds commented Apr 9, 2015

rboulton commented Apr 9, 2015

benilovj Apr 9, 2015

Choose a reason for hiding this comment

alicebartlett Apr 9, 2015

Choose a reason for hiding this comment

rivalee commented Apr 9, 2015

boffbowsh Apr 10, 2015

Choose a reason for hiding this comment

alicebartlett Apr 10, 2015

Choose a reason for hiding this comment

boffbowsh commented Apr 10, 2015

alicebartlett commented Apr 13, 2015

boffbowsh commented Apr 13, 2015

evilstreak commented Apr 13, 2015