Find the goods in old branches

find_unique_commits.sh

#!/bin/bash

# Tool to find which commits are unique to a given branch using patch-id comparison.
# This is helpful for finding good stuff & new contributions in old branches when they have become out of sync with main.
# Assumes the trip branch is called main.

set -o nounset
set -o errexit
set -o pipefail

BRANCH=${1:-$(git branch --show-current)}
BASE_BRANCH=${2:-main}

echo "=== Unique Commit Finder ==="
echo "Analyzing branch: $BRANCH"
echo "Base branch: $BASE_BRANCH"
echo "Date: $(date)"
echo ""

# Step 1: Quick branch relationship discovery
echo "🔍 STEP 1: Discovering branch relationships..."
TOTAL_COMMITS=$(git log --oneline "$BRANCH" ^"$BASE_BRANCH" --no-merges | wc -l)
echo "Total non-merge commits in $BRANCH (not in $BASE_BRANCH): $TOTAL_COMMITS"

# Find most likely parent branch using merge-base analysis
echo ""
echo "Finding most similar branches using merge-base analysis:"
BEST_CANDIDATE=""
MIN_UNIQUE=999999

git for-each-ref --format='%(refname:short)' refs/heads/ refs/remotes/ | \
grep -v -E "^($BRANCH|HEAD|origin/HEAD)$" | while read branch; do
    # Skip if branch doesn't exist
    if ! git show-ref --verify --quiet "refs/heads/$branch" && ! git show-ref --verify --quiet "refs/remotes/$branch"; then
        continue
    fi

    # Count unique commits compared to this branch
    unique_count=$(git log "$BRANCH" ^"$branch" --no-merges --oneline 2>/dev/null | wc -l || echo "999")

    if [ "$unique_count" -lt 20 ] && [ "$unique_count" -gt 0 ]; then
        echo "  $branch: $unique_count unique commits"
    fi
done | sort -n | head -10

echo ""

# Step 2: Use patch-id for definitive analysis
echo "🔍 STEP 2: Using patch-id for definitive analysis..."

# Check against api-test specifically (our known case)
if git show-ref --verify --quiet "refs/heads/api-test" || git show-ref --verify --quiet "refs/remotes/origin/api-test"; then
    CANDIDATE="api-test"
    echo "Found api-test branch - using patch-id comparison..."

    # Generate patch IDs for both branches
    echo "Generating patch IDs for $BRANCH..."
    git log "$BRANCH" ^"$BASE_BRANCH" --no-merges --format="%H" | while read hash; do
        patch_id=$(git show "$hash" | git patch-id --stable | cut -d' ' -f1)
        echo "$patch_id $hash"
    done > /tmp/branch_patches_$$

    echo "Generating patch IDs for $CANDIDATE..."
    git log "$CANDIDATE" --format="%H" | while read hash; do
        patch_id=$(git show "$hash" | git patch-id --stable | cut -d' ' -f1)
        echo "$patch_id $hash"
    done > /tmp/candidate_patches_$$

    echo ""
    echo "🎯 FINDING TRULY UNIQUE COMMITS..."

    # Find commits that don't have matching patch IDs
    UNIQUE_COMMITS=""
    while read patch_id hash; do
        if ! grep -q "^$patch_id " /tmp/candidate_patches_$$; then
            UNIQUE_COMMITS="$UNIQUE_COMMITS $hash"
            commit_subject=$(git log --format="%s" -n 1 "$hash")
            echo "UNIQUE: $hash $commit_subject"
        fi
    done < /tmp/branch_patches_$$

    UNIQUE_COUNT=$(echo "$UNIQUE_COMMITS" | wc -w)
    TOTAL_BRANCH_COMMITS=$(wc -l < /tmp/branch_patches_$$)

    echo ""
    echo "=== RESULTS ==="
    echo "Total commits in $BRANCH (not in $BASE_BRANCH): $TOTAL_BRANCH_COMMITS"
    echo "Commits also in $CANDIDATE: $((TOTAL_BRANCH_COMMITS - UNIQUE_COUNT))"
    echo "Truly unique commits: $UNIQUE_COUNT"

    if [ "$UNIQUE_COUNT" -gt 0 ]; then
        echo ""
        echo "=== UNIQUE COMMITS (chronological order) ==="
        for hash in $UNIQUE_COMMITS; do
            git log --format="%h %ad %s" --date=short -n 1 "$hash"
        done | sort -k2

        echo ""
        echo "=== FILES MODIFIED BY UNIQUE COMMITS ==="
        for hash in $UNIQUE_COMMITS; do
            git show --name-only --format="" "$hash"
        done | sort | uniq

        echo ""
        echo "=== CHERRY-PICK COMMANDS ==="
        echo "To create a clean branch with only unique commits:"
        echo ""
        echo "git checkout $BASE_BRANCH"
        echo "git checkout -b $BRANCH-unique"

        # Sort by commit date and generate cherry-pick commands
        for hash in $UNIQUE_COMMITS; do
            commit_date=$(git show -s --format=%ct "$hash")
            commit_subject=$(git log --format="%s" -n 1 "$hash")
            echo "$commit_date $hash $commit_subject"
        done | sort -n | while read date hash subject; do
            echo "git cherry-pick $hash  # $subject"
        done
    fi

    # Cleanup
    rm -f /tmp/branch_patches_$$ /tmp/candidate_patches_$$

else
    echo "⚠️  api-test branch not found. Showing all commits not in $BASE_BRANCH:"
    git log "$BRANCH" ^"$BASE_BRANCH" --no-merges --oneline --reverse
fi

echo ""
echo "=== SUMMARY ==="
echo "This script uses the following approaches to find unique commits in a branch:"
echo "1. Quick merge-base analysis to find similar branches"
echo "2. Patch-id comparison for definitive duplicate detection"
echo "3. Chronological ordering of truly unique commits"