Use find -printf instead of per-file stat fork in cleanup scripts

lidarr_cleanup.sh/sonarr_cleanup.sh/radarr_cleanup.sh each forked a
separate stat call per file during classification. find already has to
stat() every entry to know it's -type f, so -printf '%s %T@ %p' gets
size+mtime for free during the walk itself. Measured: 5.77s for all
175,954 files in the Lidarr music root (walk + stat combined) vs 85.98s
for stat alone on a 20K-file subset of the same library -- roughly 130x
faster per file, and collapses two passes into one. Verified path
parsing preserves spaces/parens/unicode exactly via read's trailing-
field capture before switching.
This commit is contained in:
Gmer4Lfe
2026-07-17 01:52:30 -04:00
parent 072df6b153
commit d2e071dece
3 changed files with 29 additions and 17 deletions
+9 -5
View File
@@ -29,6 +29,9 @@
# The filesystem is walked once per run, not twice — classification records which paths are # The filesystem is walked once per run, not twice — classification records which paths are
# eligible for deletion as it goes, and the delete pass (once the size-threshold check below # eligible for deletion as it goes, and the delete pass (once the size-threshold check below
# passes) just acts on that list instead of re-walking and re-classifying the whole tree. # passes) just acts on that list instead of re-walking and re-classifying the whole tree.
# That single walk also gets size+mtime straight from find -printf instead of a separate stat
# fork per file — find already has to stat() every entry to know it's -type f, so this is
# free by comparison. Measured ~130x faster per file (0.033ms vs 4.3ms).
# #
# ============================================================================================== # ==============================================================================================
# OPERATIONAL MODEL # OPERATIONAL MODEL
@@ -452,8 +455,9 @@ NOW=$(date +%s)
TO_DELETE_FILE="$TMP_DIR/to_delete_paths.txt" TO_DELETE_FILE="$TMP_DIR/to_delete_paths.txt"
> "$TO_DELETE_FILE" > "$TO_DELETE_FILE"
while IFS= read -r filepath; do while read -r FILE_SIZE FILE_MTIME filepath; do
[[ -z "$filepath" ]] && continue [[ -z "$filepath" ]] && continue
FILE_MTIME="${FILE_MTIME%%.*}"
# Tracked — leave alone # Tracked — leave alone
if [[ -n "${TRACKED_MAP[$filepath]:-}" ]]; then if [[ -n "${TRACKED_MAP[$filepath]:-}" ]]; then
@@ -468,9 +472,6 @@ while IFS= read -r filepath; do
continue continue
fi fi
# Single stat call for both fields instead of two separate subprocess invocations.
read -r FILE_SIZE FILE_MTIME < <(stat -c '%s %Y' "$filepath" 2>/dev/null || echo "0 0")
if has_extension "$filepath" "${LIDARR_EXTENSIONS[@]}"; then if has_extension "$filepath" "${LIDARR_EXTENSIONS[@]}"; then
FILE_AGE=$(( NOW - FILE_MTIME )) FILE_AGE=$(( NOW - FILE_MTIME ))
@@ -491,7 +492,10 @@ while IFS= read -r filepath; do
echo "$filepath" >> "$TO_DELETE_FILE" echo "$filepath" >> "$TO_DELETE_FILE"
fi fi
done < <(find "$LIDARR_MUSIC_ROOT" -type f 2>/dev/null) # -printf gets size + mtime directly from find's own stat() during the walk, instead of a
# separate stat fork per file (2026-07-17) — measured ~130x faster per file (0.033ms vs
# 4.3ms), since find already has to stat() every entry anyway to know it's -type f.
done < <(find "$LIDARR_MUSIC_ROOT" -type f -printf '%s %T@ %p\n' 2>/dev/null)
TOTAL_DELETE_BYTES=$(( ORPHAN_BYTES + JUNK_BYTES )) TOTAL_DELETE_BYTES=$(( ORPHAN_BYTES + JUNK_BYTES ))
TOTAL_REMOVED=$(( ORPHAN_COUNT + JUNK_COUNT )) TOTAL_REMOVED=$(( ORPHAN_COUNT + JUNK_COUNT ))
+10 -6
View File
@@ -27,7 +27,10 @@
# individual API calls is now a jq filter over data already in hand. The filesystem is # individual API calls is now a jq filter over data already in hand. The filesystem is
# walked once per run, not twice — classification records which paths are eligible for # walked once per run, not twice — classification records which paths are eligible for
# deletion as it goes, and the delete pass (once the size-threshold check below passes) # deletion as it goes, and the delete pass (once the size-threshold check below passes)
# just acts on that list instead of re-walking and re-classifying the whole tree. # just acts on that list instead of re-walking and re-classifying the whole tree. That single
# walk also gets size+mtime straight from find -printf instead of a separate stat fork per
# file — find already has to stat() every entry to know it's -type f, so this is free by
# comparison. Measured ~130x faster per file (0.033ms vs 4.3ms).
# #
# ============================================================================================== # ==============================================================================================
# OPERATIONAL MODEL # OPERATIONAL MODEL
@@ -424,8 +427,9 @@ NOW=$(date +%s)
TO_DELETE_FILE="$TMP_DIR/to_delete_paths.txt" TO_DELETE_FILE="$TMP_DIR/to_delete_paths.txt"
> "$TO_DELETE_FILE" > "$TO_DELETE_FILE"
while IFS= read -r filepath; do while read -r FILE_SIZE FILE_MTIME filepath; do
[[ -z "$filepath" ]] && continue [[ -z "$filepath" ]] && continue
FILE_MTIME="${FILE_MTIME%%.*}"
if [[ -n "${TRACKED_MAP[$filepath]:-}" ]]; then if [[ -n "${TRACKED_MAP[$filepath]:-}" ]]; then
log "TRACKED: $filepath" log "TRACKED: $filepath"
@@ -438,9 +442,6 @@ while IFS= read -r filepath; do
continue continue
fi fi
# Single stat call for both fields instead of two separate subprocess invocations.
read -r FILE_SIZE FILE_MTIME < <(stat -c '%s %Y' "$filepath" 2>/dev/null || echo "0 0")
if has_extension "$filepath" "${RADARR_EXTENSIONS[@]}"; then if has_extension "$filepath" "${RADARR_EXTENSIONS[@]}"; then
FILE_AGE=$(( NOW - FILE_MTIME )) FILE_AGE=$(( NOW - FILE_MTIME ))
@@ -461,9 +462,12 @@ while IFS= read -r filepath; do
echo "$filepath" >> "$TO_DELETE_FILE" echo "$filepath" >> "$TO_DELETE_FILE"
fi fi
# -printf gets size + mtime directly from find's own stat() during the walk, instead of a
# separate stat fork per file (2026-07-17) — measured ~130x faster per file (0.033ms vs
# 4.3ms), since find already has to stat() every entry anyway to know it's -type f.
done < <( done < <(
for host_path in "${SCAN_ROOTS[@]}"; do for host_path in "${SCAN_ROOTS[@]}"; do
[[ -d "$host_path" ]] && find "$host_path" -type f 2>/dev/null [[ -d "$host_path" ]] && find "$host_path" -type f -printf '%s %T@ %p\n' 2>/dev/null
done | sort -u done | sort -u
) )
+10 -6
View File
@@ -28,7 +28,10 @@
# lidarr_missing_art.sh, but the data's there once one does. The filesystem is walked once # lidarr_missing_art.sh, but the data's there once one does. The filesystem is walked once
# per run, not twice — classification records which paths are eligible for deletion as it # per run, not twice — classification records which paths are eligible for deletion as it
# goes, and the delete pass (once the size-threshold check below passes) just acts on that # goes, and the delete pass (once the size-threshold check below passes) just acts on that
# list instead of re-walking and re-classifying the whole tree. # list instead of re-walking and re-classifying the whole tree. That single walk also gets
# size+mtime straight from find -printf instead of a separate stat fork per file — find
# already has to stat() every entry to know it's -type f, so this is free by comparison.
# Measured ~130x faster per file (0.033ms vs 4.3ms).
# #
# ============================================================================================== # ==============================================================================================
# OPERATIONAL MODEL # OPERATIONAL MODEL
@@ -435,8 +438,9 @@ NOW=$(date +%s)
TO_DELETE_FILE="$TMP_DIR/to_delete_paths.txt" TO_DELETE_FILE="$TMP_DIR/to_delete_paths.txt"
> "$TO_DELETE_FILE" > "$TO_DELETE_FILE"
while IFS= read -r filepath; do while read -r FILE_SIZE FILE_MTIME filepath; do
[[ -z "$filepath" ]] && continue [[ -z "$filepath" ]] && continue
FILE_MTIME="${FILE_MTIME%%.*}"
if [[ -n "${TRACKED_MAP[$filepath]:-}" ]]; then if [[ -n "${TRACKED_MAP[$filepath]:-}" ]]; then
log "TRACKED: $filepath" log "TRACKED: $filepath"
@@ -449,9 +453,6 @@ while IFS= read -r filepath; do
continue continue
fi fi
# Single stat call for both fields instead of two separate subprocess invocations.
read -r FILE_SIZE FILE_MTIME < <(stat -c '%s %Y' "$filepath" 2>/dev/null || echo "0 0")
if has_extension "$filepath" "${SONARR_EXTENSIONS[@]}"; then if has_extension "$filepath" "${SONARR_EXTENSIONS[@]}"; then
FILE_AGE=$(( NOW - FILE_MTIME )) FILE_AGE=$(( NOW - FILE_MTIME ))
@@ -472,9 +473,12 @@ while IFS= read -r filepath; do
echo "$filepath" >> "$TO_DELETE_FILE" echo "$filepath" >> "$TO_DELETE_FILE"
fi fi
# -printf gets size + mtime directly from find's own stat() during the walk, instead of a
# separate stat fork per file (2026-07-17) — measured ~130x faster per file (0.033ms vs
# 4.3ms), since find already has to stat() every entry anyway to know it's -type f.
done < <( done < <(
for host_path in "${SCAN_ROOTS[@]}"; do for host_path in "${SCAN_ROOTS[@]}"; do
[[ -d "$host_path" ]] && find "$host_path" -type f 2>/dev/null [[ -d "$host_path" ]] && find "$host_path" -type f -printf '%s %T@ %p\n' 2>/dev/null
done | sort -u done | sort -u
) )