Optimize jh-checksource

I didn't think it needed optimizing, but on my abslibre checkout, it was
taking 48-point-something seconds.  Now it takse 14-point-something
seconds.

1 files changed, 32 insertions, 17 deletions

diff --git a/jh-checksource.sh b/jh-checksource.sh
index e3eff21..9de91c0 100644
--- a/jh-checksource.sh
+++ b/jh-checksource.sh
@@ -8,15 +8,19 @@ sep='<no-filename-ever-contains-this>'
 
 safe_types_regexp=('^(inode|text|image|video|audio)/')
 safe_types_string=('application/pdf' 'application/postscript' 'application/xml' 'application/ogg' 'message/rfc822')
-safe_files_regexp=('/\.(git|hg|svn)/' '/[^/]*\.git/' '/po/[^/]*.gmo$' '\.flw$' '\.odg$' '\.ppt$')
+safe_dirs_glob=(.{git,hg,svn} '*.git')
+safe_files_regexp=('/po/[^/]*.gmo$' '\.flw$' '\.odg$' '\.ppt$')
 safe_files_string=()
 
 # don't care about files less than 3 bytes.
 min_size=3
 
-normalize_filename() {
-	local cwd="$(readlink -m -- "$PWD")"
-	readlink -m -- "$1"|sed "s|^$cwd/|./|"
+cwd="$(readlink -m -- "$PWD")"
+
+normalize_filenames() {
+	while IFS='' read -r filename; do
+		readlink -m -- "$filename"
+	done | sed "s|^$cwd/|./|"
 }
 
 matches_string() {
@@ -53,10 +57,11 @@ print-machine() {
 main() {
 	local format=human
 	# Parse arguments
+	local arg
 	for arg in "$@"; do
 		case "$arg" in
 			-m) format=machine;;
-			*) safe_files_string+=("$(normalize_filename "$arg")");;
+			*) safe_files_string+=("$(normalize_filenames <<<"$arg")");;
 		esac
 	done
 
@@ -65,23 +70,33 @@ main() {
 	trap "$(printf 'rm -f -- %q' "$unsafe_files")" EXIT
 
 	# Heavy lifting
-	find . -type f -printf '%s %h/%f\n' | # find all files
+	local filter_dirs=()
+	local glob
+	for glob in "${safe_dirs_glob[@]}"; do
+		filter_dirs+=(-type d -name "$glob" -prune -o)
+	done
+	find . "${filter_dirs[@]}" -type f -printf '%s %p\n' | # find all files
 	while read -r size file; do # filter out files smaller than $min_size
 		[[ $size -lt $min_size ]] || printf '%s\n' "$file"
 	done |
-	xargs -d'\n' file --mime-type -r -F "$sep" | # identify the filetypes
+	normalize_filenames |
+	xargs -r -d'\n' file --mime-type -r -F "$sep" | # identify the filetypes
 	sed -r "s@(.*)${sep}\s*(.*)@\2:\1@" |  # reformat the output to be easier to parse
 	while IFS=: read -r type file; do
-		file="$(normalize_filename "$file")"
-
-		if \
-			matches_string "$file" "${safe_files_string[@]}" || \
-			matches_string "$type" "${safe_types_string[@]}" || \
-			matches_regexp "$file" "${safe_files_regexp[@]}" || \
-			matches_regexp "$type" "${safe_types_regexp[@]}" ; then
-			: # do nothing
-		else
-			printf "%s\n" "$file"
+		declare -A cached_types
+		if ! { matches_string "$file" "${safe_files_string[@]}" || \
+		       matches_regexp "$file" "${safe_files_regexp[@]}" ;}; then
+			if [[ -z ${cached_types[$type]} ]]; then
+				if matches_string "$type" "${safe_types_string[@]}" || \
+				   matches_regexp "$type" "${safe_types_regexp[@]}" ; then
+					cached_types[$type]=false
+				else
+					cached_types[$type]=true
+				fi
+			fi
+			if "${cached_types[$type]}"; then
+				printf "%s\n" "$file"
+			fi
 		fi
 	done > "$unsafe_files"