NEMA-2900: Change OR Query into OR Condition Group

[cosmin-flo-tr]

NEMA-2900

Description

The issue: The SQL query may get extended by other modules (contrib or custom) before getting executed. If the OR conjunction is applied globally, everything that's being extended will be OR-ed. This will surely make the query incorrect.

Our case: group contrib module is extending the queries with its own checks for anonymous users. Instead of constraining the query, it does the opposite because of the OR conjunction.
The result was that even if we have 1 field and 1 entity ID, the query would select all the nodes in the system. Given there are 100k+ nodes, the query would timeout. Again, this happens for guests and, most likely, non-admins.

Change log

• Changed:
  Moved the global OR conjunction into an OR-ed condition group. This ensures that the OR will get applied to the checked fields only. Any other checks will be added with AND (the default).

[brummbar]

Please provide test coverage for this part. Create a test submodule that simulates what the groups module does.

[brummbar]

Variable name should be snake_case.

[msnassar]

Hello @brummbar
Due to this query, with group module enabled and configured, unpublished media items that are not being used in any node, are accessible to unauthorized users e.g. Anonymous...

The final query before MR:

SELECT "base_table"."vid" AS "vid", "base_table"."nid" AS "nid"
FROM
{node} "base_table"
LEFT JOIN {node__ucpkn_documents} "node__ucpkn_documents" ON [node__ucpkn_documents].[entity_id] = [base_table].[nid]
LEFT OUTER JOIN {group_relationship_field_data} "gcfd" ON base_table.nid=gcfd.entity_id AND gcfd.plugin_id IN (:plugin_ids_in_use[])
INNER JOIN {node_field_data} "node_field_data" ON base_table.nid=node_field_data.nid
LEFT OUTER JOIN {group_relationship_field_data} "gcfd_2" ON gcfd.gid=gcfd_2.gid AND gcfd_2.plugin_id='group_membership' AND gcfd_2.entity_id=:account_id
WHERE ("node__ucpkn_documents"."ucpkn_documents_target_id" = :db_condition_placeholder_0) **OR** (("gcfd"."entity_id" IS NULL) OR (("node_field_data"."status" = :db_condition_placeholder_1) AND (("gcfd"."type" IN (:db_condition_placeholder_2, :db_condition_placeholder_3, :db_condition_placeholder_4, :db_condition_placeholder_5, :db_condition_placeholder_6, :db_condition_placeholder_7)) AND ("gcfd_2"."entity_id" IS NULL))))

The final query after MR:

SELECT "base_table"."vid" AS "vid", "base_table"."nid" AS "nid"
FROM
{node} "base_table"
LEFT JOIN {node__ucpkn_documents} "node__ucpkn_documents" ON [node__ucpkn_documents].[entity_id] = [base_table].[nid]
LEFT OUTER JOIN {group_relationship_field_data} "gcfd" ON base_table.nid=gcfd.entity_id AND gcfd.plugin_id IN (:plugin_ids_in_use[])
INNER JOIN {node_field_data} "node_field_data" ON base_table.nid=node_field_data.nid
LEFT OUTER JOIN {group_relationship_field_data} "gcfd_2" ON gcfd.gid=gcfd_2.gid AND gcfd_2.plugin_id='group_membership' AND gcfd_2.entity_id=:account_id
WHERE ("node__ucpkn_documents"."ucpkn_documents_target_id" = :db_condition_placeholder_0) **AND** (("gcfd"."entity_id" IS NULL) OR (("node_field_data"."status" = :db_condition_placeholder_1) AND (("gcfd"."type" IN (:db_condition_placeholder_2, :db_condition_placeholder_3, :db_condition_placeholder_4, :db_condition_placeholder_5, :db_condition_placeholder_6, :db_condition_placeholder_7)) AND ("gcfd_2"."entity_id" IS NULL))))

Notice the diff in both queries (I put it between ** **)

I confirm, the MR proposed by @cosmin-flo-tr is fixing the issue.