Project: a backup script

The -p option means create the whole chain of parents and do not treat an existing directory as an error. That is what lets the script call it on every run without guarding it with a test first. Plain mkdir would complain when the directory exists, and it would not create the parents either. The script never removes the directory, because doing so would throw away the archives already inside it.

The -C flag changes directory before tar collects anything, and the trailing dot then means everything in that folder. The result is an archive whose entries are relative to the source, so they unpack cleanly wherever you are. Without -C, the dot would refer to your current directory and you would archive the wrong tree. The compression comes from -z, not from -C.

An exit code is a message to whatever ran the script. A run that found the day archive already present did exactly what it should, so it reports success with 0. A failure would be tar being unable to write, which the script reports with 1, or a missing source folder, which it reports with 2. Using 1 for a skip would make a scheduled job look broken when nothing was wrong.

The test is true when the path is not a directory, which covers both a folder that is missing and one that is a file. Catching it first lets the script print its own message and exit with 2, so the two failure cases are easy to tell apart. Left to tar, the same problem produces a longer and less obvious error. tar can compress a directory perfectly well, so that is not the reason.