Hidden fields in displays are still indexed in SOLR [#1977832]

When doing some analytics on our search index to see if we can reduce the number of fields, I came across a problem where fields set to hidden within our content type display still wind up being indexed by SOLR. This is only valid for fields treated as such in SOLR, like taxonomy and boolean checkboxes.

To reproduce:

* Download and install apachesolr 7.x-1.x
* On any content type, add a boolean field and save.
* Go into manage displays for that content type. Verify your new field above is marked as "<Hidden>" under format
* Go into any node of that type, set a value for the new hidden field and save
* Go into your Apache SOLR index settings and reindex the updated node
* Using the out-of-the-box SOLR admin UI, look at the node you just indexed; you will see the hidden field show up.

Attached is a patch that fixes this; a simple typo in a conditional:

$instances[$bundle][$field_name]['display']['search_index'] != 'hidden'

... search_index is an array so it will always evaluate to true.

Let me know if more detail is needed.

Comment	File	Size	Author
#5	apachesolr-hidden_display_indexed-1977832-5.patch	928 bytes	blazindrop
#5
#3	apachesolr-hidden_display_indexed-1977832-3.patch	927 bytes	blazindrop
#3
#1	apachesolr-hidden_display_indexed-1977832-1.patch	799 bytes	blazindrop
#1

Support from Acquia helps fund testing for Drupal Acquia logo

Comments

Comment #1

blazindrop CreditAttribution: blazindrop commented 23 April 2013 at 16:40

Status:

Active

» Needs review

File	Size
apachesolr-hidden_display_indexed-1977832-1.patch	799 bytes

Comment #2

Nick_vh

he/him

Ghent

CreditAttribution: Nick_vh commented 23 April 2013 at 17:39

+++ b/apachesolr.moduleundefined
@@ -2055,7 +2055,7 @@ function apachesolr_entity_fields($entity_type = 'node') {
+          if (empty($instances[$bundle][$field_name]['display']['search_index']) || $instances[$bundle][$field_name]['display']['search_index']['type'] != 'hidden') {

if this is true, we should check if the ['type'] is not empty :)

Comment #3

blazindrop CreditAttribution: blazindrop commented 23 April 2013 at 20:36

File	Size
apachesolr-hidden_display_indexed-1977832-3.patch	927 bytes

Rolled another patch with Nick's suggestions aftering chatting in IRC. I did add some extra code to avoid having the line wrap. Works good locally.

Comment #4

23 April 2013 at 21:49

Status:

Needs review

» Needs work

The last submitted patch, apachesolr-hidden_display_indexed-1977832-3.patch, failed testing.

Comment #5

blazindrop CreditAttribution: blazindrop commented 24 April 2013 at 14:30

Status:

Needs work

» Needs review

File	Size
apachesolr-hidden_display_indexed-1977832-5.patch	928 bytes

Reattaching recreated patch.

Comment #6

blazindrop CreditAttribution: blazindrop commented 12 July 2013 at 14:01

Any thoughts on my previous comments? I imagine this is a pretty big issue because of the potential sensitivity of information behind hidden fields, not to mention the performance and memory improvements by not indexing hidden fields.