add only_changed input to run rubocop only against changed files

Switch to bash for process substitution and working with arrays.

Ignore if there are more than 100. Protecting from possibly hitting the
command line length limit, it can be much higher and can be calculated
or also extracted as one more input if needed. The effect from limiting
number of files processed by rubocop is also smaller.

Fix ci workflow to fetch all commits for pr branch plus head commit of
base branch to be able to find changed files.

Co-authored-by: Oliver Günther <mail@oliverguenther.de>

.github/workflows/ci.yml

@@ -11,6 +11,8 @@ jobs:
       BUNDLE_GEMFILE: ${{ github.workspace }}/test/using_bundler/Gemfile
     steps:
       - uses: actions/checkout@v4
+        with:
+          fetch-depth: 2 # we are comparing PR merge head with base
       - uses: ruby/setup-ruby@v1
         with:
           ruby-version: 2.6

README.md

@@ -47,6 +47,11 @@ Default is `added`.
 Optional. Report level for reviewdog [`info`, `warning`, `error`].
 It's same as `-level` flag of reviewdog.
 
+### `only_changed`
+
+Optional. Run Rubocop only on changed (and added) files, for speedup [`true`, `false`].
+Default: `false`.
+
 ### `reporter`
 
 Optional. Reporter of reviewdog command [`github-pr-check`, `github-check`, `github-pr-review`].

action.yml

@@ -19,6 +19,9 @@ inputs:
   level:
     description: 'Report level for reviewdog [info,warning,error]'
     default: 'error'
+  only_changed:
+    description: "Run Rubocop only on changed (and added) files, for speedup [`true`, `false`]"
+    default: 'false'
   reporter:
     description: |
       Reporter of reviewdog command [github-pr-check,github-check,github-pr-review].
@@ -61,6 +64,7 @@ runs:
         INPUT_FILTER_MODE: ${{ inputs.filter_mode }}
         INPUT_GITHUB_TOKEN: ${{ inputs.github_token }}
         INPUT_LEVEL: ${{ inputs.level }}
+        INPUT_ONLY_CHANGED: ${{ inputs.only_changed }}
         INPUT_REPORTER: ${{ inputs.reporter }}
         INPUT_REVIEWDOG_FLAGS: ${{ inputs.reviewdog_flags }}
         INPUT_RUBOCOP_EXTENSIONS: ${{ inputs.rubocop_extensions }}
@@ -70,6 +74,8 @@ runs:
         INPUT_TOOL_NAME: ${{ inputs.tool_name }}
         INPUT_USE_BUNDLER: ${{ inputs.use_bundler }}
         INPUT_WORKDIR: ${{ inputs.workdir }}
+        BASE_REF: ${{ github.event.pull_request.base.sha }}
+        HEAD_REF: ${{ github.sha }}
 branding:
   icon: 'check-circle'
   color: 'red'

script.sh

@@ -1,4 +1,7 @@
-#!/bin/sh -e
+#!/usr/bin/env bash
+
+set -e
+set -o pipefail
 
 cd "${GITHUB_WORKSPACE}/${INPUT_WORKDIR}" || exit
 export REVIEWDOG_GITHUB_API_TOKEN="${INPUT_GITHUB_TOKEN}"
@@ -85,9 +88,36 @@ else
   BUNDLE_EXEC="bundle exec "
 fi
 
+if [ "${INPUT_ONLY_CHANGED}" = "true" ]; then
+  echo '::group:: Getting changed files list'
+
+  # get intersection of changed files (excluding deleted) with target files for
+  # rubocop as an array
+  # shellcheck disable=SC2086
+  readarray -t CHANGED_FILES < <(
+    comm -12 \
+      <(git diff --diff-filter=d --name-only "${BASE_REF}..${HEAD_REF}" | sort || kill $$) \
+      <(${BUNDLE_EXEC}rubocop --list-target-files | sort || kill $$)
+  )
+
+  if (( ${#CHANGED_FILES[@]} == 0 )); then
+    echo "No relevant files for rubocop, skipping"
+    exit 0
+  fi
+
+  printf '%s\n' "${CHANGED_FILES[@]}"
+
+  if (( ${#CHANGED_FILES[@]} > 100 )); then
+    echo "More than 100 changed files (${#CHANGED_FILES[@]}), running rubocop on all files"
+    unset CHANGED_FILES
+  fi
+
+  echo '::endgroup::'
+fi
+
 echo '::group:: Running rubocop with reviewdog 🐶 ...'
 # shellcheck disable=SC2086
-${BUNDLE_EXEC}rubocop ${INPUT_RUBOCOP_FLAGS} --require ${GITHUB_ACTION_PATH}/rdjson_formatter/rdjson_formatter.rb --format RdjsonFormatter \
+${BUNDLE_EXEC}rubocop ${INPUT_RUBOCOP_FLAGS} --require ${GITHUB_ACTION_PATH}/rdjson_formatter/rdjson_formatter.rb --format RdjsonFormatter "${CHANGED_FILES[@]}" \
   | reviewdog -f=rdjson \
       -name="${INPUT_TOOL_NAME}" \
       -reporter="${INPUT_REPORTER}" \