add only_changed input to run rubocop only against changed files (#103)

* 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 <[email protected]>

* automatically fetch the tip of the base branch if it is not available

* don't hide the name of the CI job

* test only_changed

---------

Co-authored-by: Oliver Günther <[email protected]>

.github/workflows/ci.yml

@@ -2,7 +2,6 @@ name: CI
 on: [pull_request]
 jobs:
   test-skip-install-and-use-bundler:
-    name: runner / rubocop
     runs-on: ubuntu-latest
     defaults:
       run:
@@ -22,3 +21,55 @@ jobs:
           skip_install: 'true'
           use_bundler: 'true'
       - run: test "$(bundle exec rubocop --version)" == "1.18.1"
+  test-only_changed:
+    runs-on: ubuntu-latest
+    defaults:
+      run:
+        shell: bash
+    env:
+      INPUT_ONLY_CHANGED: 'true'
+    steps:
+      - uses: actions/checkout@v4
+      - name: setup
+        run: |
+          git config user.email "[email protected]"
+          git config user.name "I am an automated workflow"
+      - name: Check when there are relevant files
+        run: |
+          git checkout ${{ github.sha }}
+          rm -f test/only_changed/reviewdog-was-called
+
+          cp test/only_changed/few_relevant/files/* .
+          git add *
+          git commit -m auto
+
+          export PATH=test/only_changed/few_relevant/mock_bins:test/only_changed/shared_mock_bins:$PATH
+          BASE_REF=$(git rev-parse HEAD~) HEAD_REF=$(git rev-parse HEAD) ./script.sh
+
+          [ -f test/only_changed/reviewdog-was-called ]
+      - name: Check when there are no relevant files
+        run: |
+          git checkout ${{ github.sha }}
+          rm -f test/only_changed/reviewdog-was-called
+
+          cp test/only_changed/nothing_relevant/files/* .
+          git add *
+          git commit -m auto
+
+          export PATH=test/only_changed/nothing_relevant/mock_bins:test/only_changed/shared_mock_bins:$PATH
+          BASE_REF=$(git rev-parse HEAD~) HEAD_REF=$(git rev-parse HEAD) ./script.sh
+
+          [ ! -f test/only_changed/reviewdog-was-called ]
+      - name: Check when there are too many relevant files
+        run: |
+          git checkout ${{ github.sha }}
+          rm -f test/only_changed/reviewdog-was-called
+
+          touch a{00..100}.rb
+          git add *
+          git commit -m auto
+
+          export PATH=test/only_changed/too_many_relevant/mock_bins:test/only_changed/shared_mock_bins:$PATH
+          BASE_REF=$(git rev-parse HEAD~) HEAD_REF=$(git rev-parse HEAD) ./script.sh
+
+          [ -f test/only_changed/reviewdog-was-called ]

README.md

@@ -47,6 +47,14 @@ 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`.
+
+Will fetch the tip of the base branch with depth 1 from remote origin if it is not available.
+If you use different remote name or customize the checkout otherwise, make the tip of the base branch available before this action
+
 ### `reporter`
 
 Optional. Reporter of reviewdog command [`github-pr-check`, `github-check`, `github-pr-review`].

action.yml

@@ -19,6 +19,12 @@ 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`].
+      Will fetch the tip of the base branch with depth 1 from remote origin if it is not available.
+      If you use different remote name or customize the checkout otherwise, make the tip of the base branch available before this action.
+    default: 'false'
   reporter:
     description: |
       Reporter of reviewdog command [github-pr-check,github-check,github-pr-review].
@@ -61,6 +67,7 @@ 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 +77,8 @@ 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 +88,46 @@ else
   BUNDLE_EXEC="bundle exec "
 fi
 
+if [ "${INPUT_ONLY_CHANGED}" = "true" ]; then
+  echo '::group:: Getting changed files list'
+
+  # check if commit is present in repository, otherwise fetch it
+  if ! git cat-file -e "${BASE_REF}"; then
+    git fetch --depth 1 origin "${BASE_REF}"
+  fi
+
+  # get intersection of changed files (excluding deleted) with target files for
+  # rubocop as an array
+  # shellcheck disable=SC2086
+  readarray -t CHANGED_FILES < <(
+    comm -12 \
+      <(git diff --diff-filter=d --name-only "${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 ${GITHUB_ACTION_PATH}/rdjson_formatter/rdjson_formatter.rb \
+  --format RdjsonFormatter \
+  --fail-level error \
+  "${CHANGED_FILES[@]}" \
   | reviewdog -f=rdjson \
       -name="${INPUT_TOOL_NAME}" \
       -reporter="${INPUT_REPORTER}" \

test/only_changed/few_relevant/files/a.rb

@@ -0,0 +1 @@
+puts "Hello, " + "world!"

test/only_changed/few_relevant/files/a.txt

@@ -0,0 +1 @@
+puts "Hello, " + "world!"

test/only_changed/few_relevant/mock_bins/rubocop

@@ -0,0 +1,18 @@
+#!/usr/bin/env ruby
+# frozen_string_literal: true
+
+case ARGV
+when %w[
+  --list-target-files
+]
+  puts Dir['**/*.rb']
+when %W[
+  --require #{ENV['GITHUB_ACTION_PATH']}/rdjson_formatter/rdjson_formatter.rb
+  --format RdjsonFormatter
+  --fail-level error
+  a.rb
+]
+  puts 'Mock message for reviewdog'
+else
+  abort "rubocop mock called with unexpected arguments:\n#{ARGV.join("\n")}"
+end

test/only_changed/nothing_relevant/files/a.txt

@@ -0,0 +1 @@
+puts "Hello, " + "world!"

test/only_changed/nothing_relevant/mock_bins/rubocop

@@ -0,0 +1,11 @@
+#!/usr/bin/env ruby
+# frozen_string_literal: true
+
+case ARGV
+when %w[
+  --list-target-files
+]
+  puts Dir['**/*.rb']
+else
+  abort "rubocop mock called with unexpected arguments:\n#{ARGV.join("\n")}"
+end

test/only_changed/shared_mock_bins/bundle

@@ -0,0 +1,9 @@
+#!/bin/bash
+
+if [ "$1" = "exec" ]; then
+  shift
+  eval "$@"
+else
+  echo "Only 'exec' command is supported"
+  exit 1
+fi

test/only_changed/shared_mock_bins/curl

@@ -0,0 +1,8 @@
+#!/bin/bash
+
+arguments="$*"
+
+if [ "$arguments" != "-sfL https://raw.githubusercontent.com/reviewdog/reviewdog/master/install.sh" ]; then
+  echo "curl mock got unexpected arguments: $arguments"
+  exit 1
+fi

test/only_changed/shared_mock_bins/reviewdog

@@ -0,0 +1,10 @@
+#!/bin/bash
+
+touch test/only_changed/reviewdog-was-called
+
+read -r input
+
+if [ "$input" != "Mock message for reviewdog" ]; then
+  echo "reviewdog mock got unexpected input: $input"
+  exit 1
+fi

test/only_changed/too_many_relevant/mock_bins/rubocop

@@ -0,0 +1,17 @@
+#!/usr/bin/env ruby
+# frozen_string_literal: true
+
+case ARGV
+when %w[
+  --list-target-files
+]
+  puts Dir['**/*.rb']
+when %W[
+  --require #{ENV['GITHUB_ACTION_PATH']}/rdjson_formatter/rdjson_formatter.rb
+  --format RdjsonFormatter
+  --fail-level error
+]
+  puts 'Mock message for reviewdog'
+else
+  abort "rubocop mock called with unexpected arguments:\n#{ARGV.join("\n")}"
+end