mglaman/drupalorg-issue-search

Usage

/drupalorg-issue-search <query> [--project=<project>] [--status=all] [--skip=web_search,api_search,drupalorg_scrape]

Instructions

1. Parse inputs: Extract the search query and optional flags:

   • --project: project machine name
   • --status: issue status filter (default: all)
   • --skip: comma-separated list of channels to skip. Valid values: api_search, drupalorg_scrape, web_search. For example --skip=web_search skips the web search, --skip=api_search,web_search runs only the Drupal.org scrape.

2. Detect project and issue queue type: If --project is not provided, try to infer the project machine name from the current git remote:

   git config --get remote.origin.url
   

   Extract the project name from the URL (pattern: */project-name.git). If detection fails, proceed without a project filter.

   Once the project name is known, check whether it uses GitLab work items:

   drupalorg project:issues <project> --limit=1 --format=json
   

   If the output contains a "gitlab_issues" key or the command prints "Project uses GitLab work items" to stderr, the project has migrated. In that case:

   • Skip the api_search and drupalorg_scrape channels (they search the D.o issue queue, which is empty for this project).
   • For web_search, target site:git.drupalcode.org/project/<project>/-/issues instead.
   • Note to the user: "This project uses GitLab work items. Search results are from GitLab."

3. Run enabled searches in parallel (skip any channel listed in --skip):

   a. API search (channel: api_search) — run the CLI command:

   php drupalorg issue:search <query> --status=<status> --format=json
   

   If a project is known, include it as the first argument:

   php drupalorg issue:search <project> <query> --status=<status> --format=json
   

   b. Drupal.org issue queue scrape (channel: drupalorg_scrape) — if a project is known, fetch the project's issue search page directly using WebFetch:

   URL: https://www.drupal.org/project/issues/<project>?text=<query words joined by +>&status=All
   Prompt: Extract all issue NIDs (numeric IDs from URLs like /node/XXXX or /issues/XXXX), titles, and statuses from this page. Return as a compact list.
   

   Replace spaces in the query with + for the URL parameter. This channel searches issue titles and bodies server-side, so it can find older and closed issues that the API search misses. If no project is known, skip this channel.

   c. Web search (channel: web_search) — search the web:

   • If project is known: <query> site:https://www.drupal.org/project/issues/<project>
   • If project is unknown: <query> site:https://www.drupal.org/project/issues/

4. Extract NIDs: Parse NIDs from all active sources:

   • API search: from the JSON response
   • Drupal.org scrape: from the extracted issue list
   • Web search: from URLs matching patterns /issues/{nid} or /node/{nid} where {nid} is a numeric ID

5. Deduplicate: Collect all unique NIDs from all sources.

6. Enrich results without details: For any NIDs found via web search (but NOT in the API or scrape results which already have titles), fetch details:

   drupalorg issue:show <nid> --format=llm
   

7. Present results: Output a combined summary table with columns:

   • NID
   • Title
   • Status
   • Link (https://www.drupal.org/node/{nid})

   Group results by source if helpful (API results first, then scrape results, then web-only results).