action-diff-patch

Prepare CONTEXT, FILES, and DIFF for downstream automation by preferring a local git diff between PR base/head (or provided SHAs), and falling back to fetching a PR .diff over HTTPS when necessary. Returns prioritized file lists and batch metadata as action outputs.

Now supports optional adaptive chunking of changed files for large diffs.

Features

• PR mode: Prefer local git diff between PR base/head; fallback to HTTPS .diff
• Non‑PR mode: Use git diff <base_sha> <head_sha>
• Extracts changed files from the diff
• Optional chunk prioritization via critical_paths_json glob patterns
• Outputs minified context JSON; CHANGED_FILES and DIFF as JSON string literals; also writes file outputs under .ai/

Inputs

• token (optional; required in PR mode): GitHub token for API requests
• repository (optional): owner/repo override (defaults to GITHUB_REPOSITORY)
• pr_number (optional): PR number override; if empty, falls back to SHAs
• context_path (optional): Path to a JSON file with optional critical_paths array
• base_sha (optional): Base commit SHA for non‑PR mode
• head_sha (optional): Head commit SHA for non‑PR mode

Chunking (optional)

• chunking_enabled: "true" to enable adaptive chunking (default: "false")
• max_diff_lines_per_chunk: integer budget for added+deleted per chunk (default: "1500")
• max_files_per_chunk: optional cap on number of files per chunk (default: unset)
• include_globs: CSV of globs to include (e.g., *.swift,*.m)
• exclude_globs: CSV of globs to exclude
• critical_paths_json: JSON array of globs to prioritize earlier (e.g., ["**/Payments/**"])

Outputs

• CONTEXT: Minified JSON string of the provided context file (or {})
• TOTAL_FILES: Total changed files count
• CHANGED_FILES: JSON string literal of the newline‑joined list of changed files
• DIFF: JSON string literal of the full diff content (may be large)
• DIFF_FILE: Absolute path to the full unified diff file (e.g., .ai/diff.all.patch)
• CHANGED_FILES_FILE: Absolute path to the newline‑joined changed files file (e.g., .ai/files.all.txt)

When chunking is enabled, additional outputs are provided:

• CHUNK_COUNT: Number of generated chunks
• CHUNK_IDS_JSON: JSON array of chunk IDs as strings, e.g., ["0","1"]
• CHUNK_MANIFEST_DIR: Directory path containing manifests:
  • files.<id>.txt: newline‑joined files in the chunk
  • diff.<id>.patch: unified diff for those files

Usage

PR mode (HTTPS .diff)

- name: Prepare diff/context
  uses: GulerSevil/actions-diff-patch@<ref>
  with:
    token: ${{ secrets.GITHUB_TOKEN }}
    repository: ${{ github.repository }}
    pr_number: ${{ github.event.pull_request.number }}
    context_path: .github/prompts/ios-risk-context.json
    chunking_enabled: "true"
    max_diff_lines_per_chunk: "1500"
    include_globs: "*.swift,*.m"

Non‑PR mode (git diff SHAs)

- name: Prepare diff/context
  uses: GulerSevil/actions-diff-patch@<ref>
  with:
    token: ${{ secrets.GITHUB_TOKEN }}
    repository: ${{ github.repository }}
    base_sha: ${{ github.sha }}^     # example: previous commit
    head_sha: ${{ github.sha }}
    context_path: .github/prompts/ios-risk-context.json
    chunking_enabled: "true"

Matrix example with AI inference

- name: Prepare diff/context
  id: prep
  uses: GulerSevil/actions-diff-patch@<ref>
  with:
    token: ${{ secrets.GITHUB_TOKEN }}
    repository: ${{ github.repository }}
    pr_number: ${{ github.event.pull_request.number }}
    chunking_enabled: "true"

- name: Run AI for each chunk
  if: steps.prep.outputs.CHUNK_COUNT != '0'
  uses: some/ai-action@v1
  with:
    manifest_dir: ${{ steps.prep.outputs.CHUNK_MANIFEST_DIR }}
    chunk_ids_json: ${{ steps.prep.outputs.CHUNK_IDS_JSON }}

Consuming outputs

- name: Print all outputs
  run: |
    echo "CONTEXT=${{ steps.prep.outputs.CONTEXT }}"
    echo "TOTAL_FILES=${{ steps.prep.outputs.TOTAL_FILES }}"
    echo "CHANGED_FILES=${{ steps.prep.outputs.CHANGED_FILES }}"
    echo "DIFF=${{ steps.prep.outputs.DIFF }}"  # may be large
    echo "DIFF_FILE=${{ steps.prep.outputs.DIFF_FILE }}"
    echo "CHANGED_FILES_FILE=${{ steps.prep.outputs.CHANGED_FILES_FILE }}"
    echo "CHUNK_COUNT=${{ steps.prep.outputs.CHUNK_COUNT }}"
    echo "CHUNK_IDS_JSON=${{ steps.prep.outputs.CHUNK_IDS_JSON }}"
    echo "CHUNK_MANIFEST_DIR=${{ steps.prep.outputs.CHUNK_MANIFEST_DIR }}"

Critical paths

Provide critical_paths_json as a JSON array of globs (e.g., ["**/Payments/**", "**/*.swift"]). These globs are used to prioritize files into earlier chunks when chunking is enabled. This does not change the order of CHANGED_FILES.

["src/core/**", "infrastructure/**"]

Architecture

• src/core/config/inputs.ts: parses all action inputs (including chunking and globs)
• src/core/filters/globs.ts: CSV parsing, include/exclude filtering, critical-path checks
• src/core/chunking/chunking.ts: file size computation, sorting, bin packing, manifest writing
• src/core/diff/diff.ts: diff utilities (file extraction, filtering, prioritization helpers)
• src/core/io/fsio.ts and src/core/io/context.ts: file I/O utilities and context loader
• src/adapters/http.ts and src/adapters/git.ts: network and git adapters
• src/index.ts: orchestration using the modules above

Local development

Requirements: Node 20+

• Install: npm install
• Build: npm run build
• Test: npm run test or npm run test:watch

The TypeScript sources are under src/. Tests live under tests/ and are executed with Vitest.

Notes

• If pr_number is provided, the action attempts to fetch the .diff via HTTPS first; on failure, it falls back to git diff if base_sha and head_sha are provided.
• Ensure the token has permission to access the PR if using private repositories.
• Chunking is disabled by default and does not affect legacy outputs when off.