Empty Folder Pruner vs Command-Line ZIP Cleanup

The careful CLI equivalent

The closest CLI to the JAD logic must avoid deleting directories that hold files. A blanket zip -d archive '*/' over-deletes; you'd need to enumerate empties yourself. The browser tool computes this for you.

# Over-deletes ALL dir records (readers recreate them on extract):
zip -d release.zip '*/'

# JAD equivalent (deletes ONLY empty dir records, keeps files' parents):
# drop release.zip into /archive-tools/empty-folder-pruner -> release-pruned.zip

Python: list empty directory records

To match the JAD logic in Python you derive referenced directories from file names, then rewrite without the unreferenced directory entries — exactly what the tool does internally.

import zipfile, posixpath
zin = zipfile.ZipFile('release.zip')
files = [n for n in zin.namelist() if not n.endswith('/')]
refdirs = set()
for f in files:
    p = posixpath.dirname(f)
    while p:
        refdirs.add(p + '/'); p = posixpath.dirname(p)
keep = [i for i in zin.infolist() if not i.filename.endswith('/') or i.filename in refdirs]
# rewrite 'keep' into release-pruned.zip ...

Inspect before you prune

List directory records first so you know what will go. The browser tool shows the same result as an 'Empty dirs removed' count after processing.

# CLI: show only directory records
unzip -l release.zip | awk '$4 ~ /\/$/ {print $4}'
logs/
tmp/
src/   <- has files, will be KEPT by JAD

# JAD: result panel reports 'Empty dirs removed = 2'

Reproducible build pairing

Neither path gives stable timestamps automatically (the browser tool resets them; zip -d preserves originals). For deterministic ZIPs, normalise after pruning.

# CLI:
zip -d release.zip '*/empty-only-dirs*'
strip-nondeterminism release.zip   # or set SOURCE_DATE_EPOCH at build

# JAD:
prune -> Timestamp Normaliser (/archive-tools/timestamp-normalizer) -> stable dates

Non-ZIP archive

The browser pruner is ZIP-only. For tar.gz or 7z, a CLI is the right tool, or convert to ZIP first then prune.

# tar.gz has no real empty-dir bloat problem the same way, but to clean a 7z:
7z d archive.7z -r '*/' -x'!*.* '   # CLI handles 7z; JAD does not

# Or convert then prune in-browser:
# archive.7z -> Archive Format Converter -> archive.zip -> Empty Folder Pruner

A blanket glob removes every directory record, including ones that hold files (most readers recreate them, but it changes the listing). The JAD tool avoids this by keeping any directory referenced by a file.

The browser tool reads ZIP only and rejects other formats with 'Not a valid ZIP archive.' Use 7z/bsdtar, or convert to ZIP first with Archive Format Converter.

The browser tool holds the whole archive in memory and is capped at 2 GB on paid tiers. For larger ZIPs, a streaming CLI or in-place zip -d is the better fit.

Unlike zip -d, which keeps remaining entries' dates, the JAD tool rebuilds the ZIP and resets every date to the processing time. Use Timestamp Normaliser if dates matter.

fflate's unzipSync cannot read encrypted entries, so the browser tool fails on password-protected ZIPs. A CLI with the password, or decrypt-then-prune, is required.

The browser tool always emits a new <name>-pruned.zip; it never mutates the original. If your workflow needs in-place editing, zip -d does that directly.

For a step that runs on every build, a scripted CLI command is more appropriate than a manual browser action. The browser tool shines for one-off, ad-hoc cleanups.

On a machine where you cannot install Info-ZIP or Python, the browser tool needs nothing but a tab — and never uploads the file, so untrusted input stays local.

The browser tool has no preview-only mode; it prunes and reports the count. To inspect first, list directory records with unzip -l or use File Listing Generator before pruning.