Support proportional language attribution via `repository.languages`

[shm11C3]

Summary

Currently, language statistics are attributed only to each repository’s primaryLanguage.
This makes multi-language repositories underrepresented.

Example:

• A large repository / monorepo contains multiple languages (e.g., Language A 60%, Language B 40%)
• Current behavior: 100% attributed to the primary language
• Expected behavior (opt-in): attribute contributions proportionally across languages

This commonly occurs in monorepos and large repositories where multiple languages coexist.

Cause

Language stats are currently attributed to a single language per repository. As a result, other languages used in the same repository are ignored.

Proposed Solution

Introduce a new option to enable proportional language attribution, while keeping the current behavior as the default.

Option design

Use an explicit mode string rather than a boolean for clarity and extensibility:

language_mode: "primary" | "breakdown"

• "primary" (default): current behavior — attribute all contributions to primaryLanguage
• "breakdown": use repository.languages and split repo contributions proportionally

Implementation approach

One concrete approach (verified in a fork) is to fetch a language breakdown from the GitHub API (e.g., repository.languages with size per language) and use the ratios to split the repository’s contribution counts.

commitContributionsByRepository(maxRepositories: 100) {
    repository {
        languages(first: 10, orderBy: {field: SIZE, direction: DESC}) {
            edges {
                size
                node { name, color }
            }
        }
    }
    contributions { totalCount }
}

Proof of concept (verified in a fork): https://github.com/shm11C3/github-profile-3d-contrib/blob/aacbea44a59f51e5074d506dc2abae8161435f9e/src/aggregate-user-info.ts#L53-L77

Attribution algorithm

Given:

• repoContrib = total contributions attributed to a repository
• langSizes = list of (language, size)

Compute ratio = size / sum(size), then attribute repoContrib * ratio to each language.

Example

• Repo contributions: 10
• Language A 60%, Language B 40%
• => Language A 6, Language B 4

Notes / Considerations

API cost

The node cost increase is minimal — up to +10 nodes per repository (capped by first: 10), well within GitHub's 500,000 node limit per query. No additional HTTP requests are needed compared to the current implementation.

Backward compatibility

• Default behavior is unchanged (language_mode: "primary")
• New behavior only when users explicitly set language_mode: "breakdown"
• The GraphQL query should fetch both primaryLanguage and languages fields, switching the attribution logic based on the mode setting

Fractional contributions

Proportional attribution produces fractional values (e.g., 6.37 contributions). Consider:

• Using Math.round() when rendering to avoid displaying decimals
• Accepting that per-language totals may not sum exactly to the repo total due to rounding
• The type.LangInfo.contributions field type changes from effectively integer to float — verify downstream rendering and sorting are unaffected