What does git rev-parse do?

git rev-parse is an ancillary plumbing command primarily used for manipulation.

One common usage of git rev-parse is to print the SHA1 hashes given a revision specifier. In addition, it has various options to format this output such as --short for printing a shorter unique SHA1.

There are other use cases as well (in scripts and other tools built on top of git) that I've used for:

• --verify to verify that the specified object is a valid git object.
• --git-dir for displaying the abs/relative path of the .git directory.
• Checking if you're currently within a repository using --is-inside-git-dir or within a work-tree using --is-inside-work-tree
• Checking if the repo is a bare using --is-bare-repository
• Printing SHA1 hashes of branches (--branches), tags (--tags) and the refs can also be filtered based on the remote (using --remote)
• --parse-opt to normalize arguments in a script (kind of similar to getopt) and print an output string that can be used with eval

Massage just implies that it is possible to convert the info from one form into another i.e. a transformation command. These are some quick examples I can think of:

• a branch or tag name into the commit's SHA1 it is pointing to so that it can be passed to a plumbing command which only accepts SHA1 values for the commit.
• a revision range A..B for git log or git diff into the equivalent arguments for the underlying plumbing command as B ^A

Just to elaborate on the etymology of the command name rev-parse, Git consistently uses the term rev in plumbing commands as short for "revision" and generally meaning the 40-character SHA1 hash for a commit. The command rev-list for example prints a list of 40-char commit hashes for a branch or whatever.

In this case the name might be expanded to parse-a-commitish-to-a-full-SHA1-hash. While the command has the several ancillary functions mentioned in Tuxdude's answer, its namesake appears to be the use case of transforming a user-friendly reference like a branch name or abbreviated hash into the unambiguous 40-character SHA1 hash most useful for many programming/plumbing purposes.

I know I was thinking it was "reverse-parse" something for quite a while before I figured it out and had the same trouble making sense of the terms "massaging" and "manipulation" :)

Anyway, I find this "parse-to-a-revision" notion a satisfying way to think of it, and a reliable concept for bringing this command to mind when I need that sort of thing. Frequently in scripting Git you take a user-friendly commit reference as user input and generally want to get it resolved to a validated and unambiguous working reference as soon after receiving it as possible. Otherwise input translation and validation tends to proliferate through the script.