labeler: switch to using git diff for labeling (#4052)

* add label calculation script

* add additional ignore files

* add debug mode

* handle spaces in filenames safely

* fix compat, simplify

* set correct default

* fix shebang

* add test for new labeler script

* set env vars, remove jq dep

* include history for merge check

* simplify

* Simplify PR size labeling workflow

- Replace pr-size-labeler with custom labeler script
- Remove redundant labeler_v2 workflow file
- Enhance labeler script to clean up existing size labels before adding new ones

* test label failure

* remove test failure

* Update diff tool to show number of changes per file

- Switched from `--shortstat` to `--numstat` for more detailed file change output

* Refactor ignore arguments in git diff script

- Consolidated exclude arguments into IGNORE_ARGS for clarity

* Enforce stricter PR labeler checks

- Added FAIL_IF_XL environment variable for labeler script execution

.github/workflows/labeler.yml

@@ -10,38 +10,12 @@ jobs:
     if: github.event.pull_request.draft == false
     name: Label the PR size
     steps:
-      - uses: mastercactapus/pr-size-labeler@eb6a25f40d6a3327581aca35839b60d419e8019a # https://github.com/CodelyTV/pr-size-labeler/pull/61, updated with latest from main
+      - uses: actions/checkout@v4
         with:
-          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
-          xs_label: 'size/xs'
-          xs_max_size: '15'
-          s_label: 'size/s'
-          s_max_size: '100'
-          m_label: 'size/m'
-          m_max_size: '250'
-          l_label: 'size/l'
-          l_max_size: '500'
-          xl_label: 'size/xl'
-          fail_if_xl: 'true'
-          ignore_deleted: 'true'
-          message_if_xl: >
-            This PR exceeds the recommended size of 500 lines.
-            Please make sure you are NOT addressing multiple issues with one PR.
-            Note this PR might be rejected due to its size.
-          files_to_ignore: |
-            "*.lock"
-            "graphql2/generated.go"
-            "graphql2/maplimit.go"
-            "graphql2/mapconfig.go"
-            "graphql2/models_gen.go"
-            "pkg/sysapi/*.pb.go"
-            "swo/swodb/*.go"
-            "web/src/*.d.ts"
-            "Makefile.binaries.mk"
-            "gadb/*.go"
-            "migrate/schema.sql"
-            "go.mod"
-            "go.sum"
-            ".yarn/releases/*"
-            "devtools/pgdump-lite/pgd/*"
-            "timezone/*.txt"
+          fetch-depth: 0
+      - name: Run labeler script
+        env:
+          GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+          PR_NUMBER: ${{ github.event.number }}
+          FAIL_IF_XL: '1'
+        run: ./devtools/scripts/github-set-pr-label.sh

.labelignore

@@ -0,0 +1,4 @@
+*.lock
+go.mod
+go.sum
+.yarn/releases/*

devtools/scripts/git-diff-label-calc.sh

@@ -0,0 +1,112 @@
+#!/usr/bin/env bash
+set -e
+
+# Outputs a label based on the size of the diff between the current branch and the main branch.
+
+# env setting defaults
+if [ -z "$MAIN_BRANCH" ]; then
+    MAIN_BRANCH=master
+fi
+if [ -z "$FAIL_IF_XL" ]; then
+    FAIL_IF_XL=0
+fi
+
+# const settings
+IGNORE_PATTERN_FILE=.labelignore
+XS_LABEL=size/xs
+XS_MAX=20
+S_LABEL=size/s
+S_MAX=100
+M_LABEL=size/m
+M_MAX=250
+L_LABEL=size/l
+L_MAX=500
+XL_LABEL=size/xl
+
+if [ -z "$DEBUG" ]; then
+    DEBUG=0
+fi
+if [ "$1" = "--debug" ]; then
+    DEBUG=1
+fi
+
+# debug works like echo but only prints when --debug is passed and to stderr
+debug() {
+    if [ $DEBUG -eq 1 ]; then
+        echo "$1" >&2
+    fi
+}
+
+IGNORE_ARGS=()
+
+# Get a list of files including the "generated by" header, then filter them out by only those with DO NOT EDIT on the first line, in case we have code in the repo that does the generation.
+while IFS= read -r file; do
+    # include only files where DO NOT EDIT appears on the very first line
+    if head -n 1 "$file" | grep -q "DO NOT EDIT"; then
+        debug "IGNORE: generated file: $file"
+        IGNORE_ARGS+=(":(exclude)$file")
+    fi
+done < <(git grep -l -e 'generated by .* DO NOT EDIT')
+
+# Read the .labelignore file and add the resolved files to the IGNORE_ARGS array.
+while IFS= read -r pattern; do
+    # Skip empty lines and comments
+    case "$pattern" in
+    '' | \#*) continue ;;
+    esac
+
+    # Use git ls-files with the pattern
+    while IFS= read -r file; do
+        debug "IGNORE: pattern '$pattern': $file"
+        IGNORE_ARGS+=(":(exclude)$file")
+    done < <(git ls-files "$pattern")
+done < <(cat "$IGNORE_PATTERN_FILE" 2>/dev/null)
+
+# Perform diff for merging to the main branch:
+# --ignore-blank-lines: ignore changes that only involve blank lines
+# --numstat: output only the number of changes (per file)
+# -w: ignore whitespace
+# --diff-filter=d: exclude deleted files
+# --minimal: produce the smallest diff possible
+# --merge-base: find the common ancestor of the current branch and the main branch
+# $IGNORE_ARGS[@]: list of `:(exclude)file` patterns to ignore
+OUTPUT=$(
+    if [ "$DEBUG" = "1" ]; then set -x; fi
+    git diff --ignore-blank-lines --numstat -w --diff-filter=d --minimal --merge-base "origin/$MAIN_BRANCH" -- "${IGNORE_ARGS[@]}"
+)
+
+debug
+debug "DIFF FILE STATS"
+debug "==============="
+debug "$OUTPUT"
+debug
+
+sum=$(
+    echo "$OUTPUT" | awk '{ total += $1 + $2 } END { print total }'
+)
+
+debug "TOTAL: $sum"
+
+LABEL=""
+if [ $sum -le $XS_MAX ]; then
+    LABEL=$XS_LABEL
+elif [ $sum -le $S_MAX ]; then
+    LABEL=$S_LABEL
+elif [ $sum -le $M_MAX ]; then
+    LABEL=$M_LABEL
+elif [ $sum -le $L_MAX ]; then
+    LABEL=$L_LABEL
+else
+    LABEL=$XL_LABEL
+fi
+
+debug "LABEL: $LABEL"
+
+if [ $FAIL_IF_XL -eq 1 ] && [ "$LABEL" = "$XL_LABEL" ]; then
+    echo ""
+    echo "ERROR: This PR exceeds the maximum size of $L_MAX lines."
+    echo "Please make sure you are NOT addressing multiple issues with one PR."
+    exit 1
+fi
+
+echo "$LABEL"

devtools/scripts/github-set-pr-label.sh

@@ -0,0 +1,19 @@
+#!/usr/bin/env bash
+set -e
+
+if [ -z "PR_NUMBER" ]; then
+    echo "PR_NUMBER is not set"
+    exit 1
+fi
+
+LABEL=$(./devtools/scripts/git-diff-label-calc.sh --debug)
+
+# Remove any existing test/* labels
+for label in $(gh pr view "$PR_NUMBER" --json labels --jq '.labels[] | select(.name | startswith("size/")) | .name'); do
+    if [ "$label" == "$LABEL" ]; then
+        continue # Skip the label we want to add
+    fi
+    gh pr edit "$PR_NUMBER" --remove-label "$label"
+done
+
+gh pr edit "$PR_NUMBER" --add-label "$LABEL"