Explain tar code

What this tar sample shows

c, t and x

The first letter says what tar does: c creates an archive, t lists what's in one and x extracts it. On a folder holding package.json, src/index.html, src/app.js, node_modules and server.log, -tzf listed 5 entries: my-site/, my-site/package.json, my-site/src/, and the 2 files in src. The listing wrote no files.

z, v and f

z runs the archive through gzip, v prints each file as tar works, and f names the archive file, so the name must come straight after it. Written tar -cfz x.tar.gz my-site, tar took z as the archive's name: it made a file called z and failed on x.tar.gz: Cannot stat: No such file or directory.

-C and --exclude

--exclude patterns kept node_modules and server.log out of the archive. GNU tar only applies them when they come before the folder: put after my-site, it said --exclude ‘node_modules’ has no effect, exited with code 2 and still packed node_modules. -C ~/restore unpacked into ~/restore, so the files landed in ~/restore/my-site. With a folder that didn't exist, GNU tar stopped with Cannot open: No such file or directory.

one file out

A name after the archive extracts only that member, written exactly as -t lists it: my-site/package.json, not package.json. A name that wasn't in the archive gave Not found in archive. A pattern like 'my-site/src/*.js' found nothing in GNU tar until --wildcards was added. bsdtar extracted app.js without it, and refused --wildcards as an option it doesn't support.

Good to know

• Run with GNU tar 1.35, and with bsdtar 3.8.9 from libarchive. The tar in macOS is Apple's own, older build of bsdtar (github.com/apple-oss-distributions/libarchive). Both packed the same 5 entries from the sample. GNU tar's v printed plain names, and bsdtar's put a before each file it added and x before each file it extracted.
• Where they differ: -xzf on a plain .tar stopped GNU tar with gzip: stdin: not in gzip format, while bsdtar unpacked it. Both unpacked the .tar.gz with -xf and no z, because each works out the compression itself when reading.

Watch out: Extracting overwrites files with the same name, without asking. After package.json in ~/restore was changed, extracting the archive into ~/restore again put the archived version back, with no warning and exit code 0. Add -k to keep files that already exist: GNU tar then refused each one with Cannot open: File exists, and bsdtar skipped them silently. Unpack into an empty folder when you aren't sure what's in an archive.