[Copilot]

Every project currently counts the same in the overall language percentage, so a 10-line repository has the same weight as a 10,000-line one.

GitLab's languages API still returns only percentages, not per-language line counts. Those percentages are now weighted by the project's precomputed statistics.repository_size. Absolute byte or line counts are used directly when a payload includes them.

• Weighting
  • Default is repository_size: overall share is sum(percentage * size) / sum(size).
  • --weight equal keeps the previous per-project average.
  • A missing size counts as 1 byte and is reported on projects_missing_size_total.
  • repository_size_bytes_total is the sum of sizes actually used.
• Lookup and cache
  • Size comes from project statistics already on the object, otherwise one statistics request.
  • Group scans request statistics on the existing project fetch.
  • --cache stores the size next to language data and reuses it until last_activity_at changes.
  • Statistics need the Reporter role. Without it, those projects fall back to 1 byte.

gitlab-languages                  # weight by repository size
gitlab-languages --weight equal   # previous unweighted average

• Fixes Consider project size in overall statistics #10