Command line tool to detect language distribution in a codebase, ignore binary or vendored files.

linguistic.sh

linguistic() {
IFS_ORIGINAL=$IFS
IFS="?"
FILES=($(find -type f -name "*" -not -path "./.git/*" \
                -exec sh -c 'printf "%s " "$(du -b {})"' \; \
                -exec file -b {} \; ))

SUM=($(sed 's|\./[^ ]*| |g' <<< ${FILES[@]} \
     | sort -k2 \
     | awk '{v=$1;$1="";s[$0]+=v}END{for(i in s)print s[i] i "?"}' \
    ))

NON_EXEC=("${SUM[@]//*relocatable*}")
NON_EXEC=("${NON_EXEC[@]//*executable*}")

IFS=$IFS_ORIGINAL

TOTAL_BYTES=$(awk '{split($0,a,","); sum += a[1]} END {print sum}' <<< "${NON_EXEC[@]}")

printf "Total non-executable bytes: %d\n" "$TOTAL_BYTES"

for (( i=0; i<${#NON_EXEC[@]}; i++ )); do
    ITEM="${NON_EXEC[i]/$'\n'/}"
    BYTES=$(printf "%s\n" "$ITEM" | cut -d' ' -f1 )
    TYPE=$(printf "%s\n" "$ITEM" | cut -d' ' -f2- )
    DOTS="..."
    [ "${#TYPE}" -ge "$((COLUMNS-${#DOTS}))" ] && TYPE="${TYPE::72}..." || :
    if [[ $BYTES > 0 ]]; then
        PERCENT=$( echo "scale=20; $BYTES / $TOTAL_BYTES * 100" | bc -l )
        printf "%5.2f%% \t%s\n" "$PERCENT" "${TYPE}"
    fi
done | sort -nr

EXEC=("$(grep 'executable\|relocatable' <<< ${FILES[@]} \
       | grep -v 'text' \
       | sed 's|.*\(\./[^ ]*\).*|\1|g')")
NUM_EXEC=$(wc -l <<< $EXEC)

if [[ ${#EXEC} > 0 ]]; then
    printf "\nTotal number of binary files %d\n" $NUM_EXEC
    printf "Binary files:\n"
    printf "%s\n" "$EXEC"
else
    printf "No binary files found\n"
fi

}