## Problem The table `notes` has columns `original_position`, `position` and `change_position` with type `text`, same happens with the newest `draft_notes`. In DB level it persist a "Ruby object" (YAML I assume?) string: ```psql gitlabhq_development=# SELECT position FROM "draft_notes" limit 1; position -------------------------------------------------------- --- !ruby/object:Gitlab::Diff::Position + attributes: + :base_sha: 1c59dfa64afbea8c721bb09a06a9d326c952ea19 + :start_sha: 1c59dfa64afbea8c721bb09a06a9d326c952ea19+ :head_sha: af14638a616cd93a0757e35b1fc69c3611065b72 + :old_path: ".gitlab-ci.yml" + :new_path: ".gitlab-ci.yml" + :position_type: text + :old_line: + :new_line: 27 + (1 row) ``` In general that makes it impossible to query with postgres `->` syntax as it's not a `json` or `jsonb` column. It limits optimizations where we'd like to query for specific file paths and should keep biting us in other places. We're not able to answer (with queries) how many comments are there in the latest diff version, or how many draft notes. Everything needs to be made in memory now. ## Proposals Either converting all these records to `jsonb`, or create a `diff_positions` table to store this data. `notes` and `draft_notes` could instead refer to `position_id`, `original_position_id` and `change_position_id`.