Count the number of models in a Rails app over time.

model-count-history.sh

APP_DIR=.
DATE=0

echo "Date, Tier 1, Tier 2, Tier 3, Tier 4, Tier 5, Total" > model_counts.csv
for COMMIT in $(git log --first-parent --reverse --format=format:%H -- $APP_DIR); do
    NEW_DATE=$(git show -s --format=%cs ${COMMIT})
    if [[ "$DATE" == "$NEW_DATE" ]]; then
        continue
    fi

    git checkout ${COMMIT} -f >/dev/null 2>&1;

    DATE=$NEW_DATE
    LAYER_1=$(find $APP_DIR/app/models/* -maxdepth 0 -type f ! -path '*/concerns/*' 2>/dev/null | wc -l)
    LAYER_2=$(find $APP_DIR/app/models/**/* -maxdepth 0 -type f ! -path '*/concerns/*' 2>/dev/null  | wc -l)
    LAYER_3=$(find $APP_DIR/app/models/**/**/* -maxdepth 0 -type f ! -path '*/concerns/*' 2>/dev/null | wc -l)
    LAYER_4=$(find $APP_DIR/app/models/**/**/**/* -maxdepth 0 -type f ! -path '*/concerns/*' 2>/dev/null | wc -l)
    LAYER_5=$(find $APP_DIR/app/models/**/**/**/**/* -maxdepth 0 -type f ! -path '*/concerns/*' 2>/dev/null | wc -l)
    TOTAL=$(find $APP_DIR/app/models -type f ! -path '*/concerns/*' 2>/dev/null | wc -l)

    echo "$DATE, $LAYER_1, $LAYER_2, $LAYER_3, $LAYER_4, $LAYER_5, $TOTAL"
done >> model_counts.csv

Note to self: Even though this tries to only take 1 row per date, commits aren't sorted in date order, so I'll need to add a step to remove duplicates.

Edit: adding --first-parent seems to mostly resolve the issue -- it'll skip intermediary commits and favor the history of merges (and squash merges). This assumes that the history of a repo follows a relatively sane merge pattern 🙂