From 855c3e80a1e047efeb4799a2cb25ed144a82bcd8 Mon Sep 17 00:00:00 2001
From: Johannes Aubart <johannes.aubart@sap.com>
Date: Wed, 3 Sep 2025 17:56:59 +0200
Subject: [PATCH] improve release note generation

---
 generate-changelog.sh    | 51 +++++++++++++++++--------
 release-notes-to-json.sh | 80 ++++++++++++++++++++++++++++++++++++++++
 2 files changed, 115 insertions(+), 16 deletions(-)
 create mode 100755 release-notes-to-json.sh

diff --git a/generate-changelog.sh b/generate-changelog.sh
index 5912c55..c3f868c 100755
--- a/generate-changelog.sh
+++ b/generate-changelog.sh
@@ -2,12 +2,15 @@
 
 set -euo pipefail
 
+echo "Generating changelog for version: $VERSION"
+
 # Ensure gh CLI available
 if ! command -v gh &> /dev/null; then
   echo "gh CLI is required but not installed." >&2
   exit 1
 fi
 
+RELEASE_NOTES_TO_JSON_SCRIPT="$(realpath "$(dirname $0)/release-notes-to-json.sh")"
 cd $(dirname "$0")/../../
 
 LATEST_RELEASE_TAG=$(gh release list --json tagName,isLatest --jq '.[] | select(.isLatest)|.tagName')
@@ -15,7 +18,7 @@ if [[ -z "$LATEST_RELEASE_TAG" ]]; then # first release?
   LATEST_RELEASE_TAG=$(git rev-list --max-parents=0 HEAD) # first commit in the branch.
 fi
 
-GIT_LOG_OUTPUT=$(git log "$LATEST_RELEASE_TAG"..HEAD --oneline --pretty=format:"%s" main)
+GIT_LOG_OUTPUT=$(git log "$LATEST_RELEASE_TAG"..HEAD --oneline --pretty=format:"%s")
 PR_COMMITS=$(echo "$GIT_LOG_OUTPUT" | grep -oE "#[0-9]+" || true | tr -d '#' | sort -u)
 
 CHANGELOG_FILE=./CHANGELOG.md
@@ -43,38 +46,54 @@ done
 
 for PR_NUMBER in $PR_COMMITS; do
   PR_JSON=$(gh pr view "$PR_NUMBER" --json number,title,body,url,author)
+  echo -n "Checking PR $PR_NUMBER"
 
-  IS_BOT=$(echo "$PR_JSON" | jq -r '.author.is_bot')
+  IS_BOT=$(jq -r '.author.is_bot' <<< "$PR_JSON")
   if [[ "$IS_BOT" == "true" ]]; then
+    echo " [skipping bot PR"]
     continue
   fi
 
-  PR_TITLE=$(echo "$PR_JSON" | jq -r '.title')
-  PR_URL=$(echo "$PR_JSON" | jq -r '.url')
-  PR_BODY=$(echo "$PR_JSON" | jq -r '.body')
+  PR_TITLE=$(jq -r '.title' <<< "$PR_JSON")
+  PR_URL=$(jq -r '.url' <<< "$PR_JSON")
+  PR_BODY=$(jq -r '.body' <<< "$PR_JSON")
+  echo " - $PR_TITLE"
 
   # Determine type from conventional commit (assumes title like "type(scope): message" or "type: message")
   TYPE=$(echo "$PR_TITLE" | grep -oE '^[a-z]+' || echo "feat")
   CLEAN_TITLE=$(echo "$PR_TITLE" | sed -E 's/^[a-z]+(\([^)]+\))?(!)?:[[:space:]]+//')
 
   # Extract release note block, this contains the release notes and the release notes headers.
-  RELEASE_NOTE_BLOCK=$(echo "$PR_BODY" | sed -n '/\*\*Release note\*\*:/,$p' | sed -n '/^```.*$/,/^```$/p')
+  # The last sed call is required to remove the carriage return characters (Github seems to use \r\n for new lines in PR bodies).
+  RELEASE_NOTE_BLOCK=$(echo "$PR_BODY" | sed -n '/\*\*Release note\*\*:/,$p' | sed -n '/^```.*$/,/^```$/p' | sed 's/\r//g')
   # Extract release notes body
-  RELEASE_NOTE=$(echo "$RELEASE_NOTE_BLOCK" | sed '1d;$d' | sed -e 's/^[[:space:]]*//' -e 's/[[:space:]]*$//')
+  RELEASE_NOTE_JSON=$("$RELEASE_NOTES_TO_JSON_SCRIPT" <<< "$RELEASE_NOTE_BLOCK")
+
+  # skip PRs without release notes
+  if [[ "$RELEASE_NOTE_JSON" == "[]" ]]; then
+    echo "  [ignoring PR without release notes]"
+    continue
+  fi
+
+  # Format release notes
+  # Updating NOTE_ENTRY in the loop does not work because it is executed in a subshell, therefore this workaround via echo.
+  NOTE_ENTRY="$(
+    jq -rc 'sort_by(.audience, .type) | .[]' <<< "$RELEASE_NOTE_JSON" | while IFS= read -r note; do
+      NOTE_TYPE=$(jq -r '.type' <<< "$note" | tr '[:lower:]' '[:upper:]')
+      NOTE_AUDIENCE=$(jq -r '.audience' <<< "$note" | tr '[:lower:]' '[:upper:]')
+      NOTE_BODY=$(jq -r '.body' <<< "$note")
+      echo -en "\n  - **[$NOTE_AUDIENCE][$NOTE_TYPE]** $NOTE_BODY"
+    done
+  )"
 
   # Format entry
   ENTRY="- $CLEAN_TITLE [#${PR_NUMBER}](${PR_URL})"
 
-  if [[ -z "$RELEASE_NOTE"  || "$RELEASE_NOTE" == "NONE" ]]; then
-    ENTRY+="."
-  else
-    # Extract and format the release note headers.
-    HEADERS=$(echo "$PR_BODY" | sed -n '/\*\*Release note\*\*:/,$p' | sed -n '/^```.*$/,/^```$/p'| head -n 1 | sed 's/^```//')
-    FORMATED_HEADERS=$(echo "$HEADERS" | sed 's/^[[:space:]]*//; s/[[:space:]]*$//; s/\s\+/ /g' | sed 's/\(\S\+\)/[\1]/g')
+  # Extract and format the release note headers.
+  HEADERS=$(echo "$PR_BODY" | sed -n '/\*\*Release note\*\*:/,$p' | sed -n '/^```.*$/,/^```$/p'| head -n 1 | sed 's/^```//')
+  FORMATED_HEADERS=$(echo "$HEADERS" | sed 's/^[[:space:]]*//; s/[[:space:]]*$//; s/\s\+/ /g' | sed 's/\(\S\+\)/[\1]/g')
 
-    ENTRY="- ${FORMATED_HEADERS} ${CLEAN_TITLE} [#${PR_NUMBER}](${PR_URL}): ${RELEASE_NOTE}"
-  fi
-  ENTRY+="\n"
+  ENTRY="- ${CLEAN_TITLE} [${PR_NUMBER}](${PR_URL})${NOTE_ENTRY}\n"
 
   # Append to appropriate section
   if [[ -n "${PR_ENTRIES[$TYPE]+x}" ]]; then
diff --git a/release-notes-to-json.sh b/release-notes-to-json.sh
new file mode 100755
index 0000000..3386721
--- /dev/null
+++ b/release-notes-to-json.sh
@@ -0,0 +1,80 @@
+#!/bin/bash
+
+# This script reads release notes from STDIN and converts them to JSON format, which is then printed to STDOUT.
+# Expected input format:
+#
+# ```<type> <audience>
+# <body> (may span multiple lines)
+# ```
+# (the above block may repeat multiple times)
+#
+# Output:
+# [
+#   {
+#     "type": "<type>",
+#     "audience": "<audience>",
+#     "body": "<body>"
+#   },
+#   ...
+# ]
+#
+# Additional whitespace is ignored, but each release note block must start with ``` followed by two lowercase words,
+# and it must end with ``` followed by any amount of whitespace and then a newline or end of input.
+#
+# Release notes where the body is empty or "NONE" are ignored and not part of the output JSON array.
+
+set -euo pipefail
+
+type=""
+audience=""
+body=""
+rnj='[]'
+
+start_regex='^[[:blank:]]*```[[:blank:]]*([a-z]+)[[:blank:]]+([a-z]+)[[:blank:]]*$'
+end_regex='^[[:blank:]]*```[[:blank:]]*$'
+empty_regex='^[[:space:]]*$'
+line=""
+
+while IFS= read -r line || [[ -n "$line" ]]; do
+  # check if we are currently reading a release note body
+  if [[ -n "$type" ]]; then
+    # check for end of release note block
+    if [[ "$line" =~ $end_regex ]]; then
+      # end of release note block
+      # add release note to JSON array, unless the body is empty or "NONE"
+      if [[ ! "$body" =~ $empty_regex ]] && [[ "$body" != "NONE" ]]; then
+        rnj="$(jq '. + [{"type": $type, "audience": $audience, "body": $body}]' --arg type "$type" --arg audience "$audience" --arg body "$body" <<< "$rnj")" 
+      fi
+      # reset variables
+      type=""
+      audience=""
+      body=""
+    else
+      # more body content
+      # append line to body
+      if [[ -n "$body" ]]; then
+        body="$body\n$line"
+      else
+        body="$line"
+      fi
+    fi
+  else
+    # not currently reading a release note body
+    # check for start of release note block
+    if [[ "$line" =~ $start_regex ]]; then
+      # start of release note block
+      type="${BASH_REMATCH[1]}"
+      audience="${BASH_REMATCH[2]}"
+      body=""
+    else
+      # invalid line outside of release note block
+      if [[ -n "$line" ]]; then
+      # ignore empty lines, log a warning otherwise
+        echo "unexpected line in release notes: $line" >&2
+      fi
+    fi
+  fi
+done
+
+# output JSON array
+echo "$rnj"