Choose how to resolve a path

3 min read

When a script receives a path, first decide what must exist. Resolving an existing input, naming a new output file, and planning a directory tree need different checks.

Start with one working directory #

Assume the physical working directory is /home/jack/project, with no symbolic links in that directory path:

  • notes.txt is an existing ordinary file.
  • latest is a symbolic link whose stored target is notes.txt.
  • report.txt does not exist yet.
  • planned and its child directory drafts do not exist yet.

The examples use those files and directories and show the expected output. Relative paths are interpreted from the working directory, which you can inspect with pwd:

$ pwd -P
/home/jack/project

In a script, check the command's exit status before using its output. The examples omit realpath -q: affected builds can report success for invalid operands in quiet mode. See the precise compatibility warning and workaround. Nonempty output alone does not establish success.

Use plain readlink when you want the path stored in the link:

$ readlink "latest"
notes.txt

That is the immediate target, not an absolute path or a check that the target exists. Plain readlink "notes.txt" would fail because notes.txt is an ordinary file; that failure does not prove the file is absent.

Require an existing input #

Use realpath -e when every component, including the final file, must exist:

$ realpath -e "latest"
/home/jack/project/notes.txt

Unlike plain readlink, this resolves the whole path, following symbolic links and removing . and ... It also accepts an ordinary file. Under the example assumptions, realpath -e "report.txt" would fail because that final file is absent.

Name a new file under existing directories #

Default realpath allows the final component to be absent, but requires its parents to exist:

$ realpath "report.txt"
/home/jack/project/report.txt

Choose this when preparing an output filename in an existing directory. Resolving the name does not create the file or establish that a later write will succeed.

Plan a path with missing parents #

Use -m when parent directories may be absent too:

$ realpath -m "planned/drafts/report.txt"
/home/jack/project/planned/drafts/report.txt

Where the path cannot be walked, it is canonicalised as text. This gives you a planned spelling; it does not create directories or establish access. Default realpath would fail for this example because the parent directories are absent.

Split a name or check it #

If you only need part of a path, basename and dirname operate on text without consulting the file system:

$ basename "planned/drafts/report.txt"
report.txt

$ dirname "planned/drafts/report.txt"
planned/drafts

pathchk checks names instead: its default checks include file-system length limits and whether leading directories can be searched. A successful check does not guarantee that a future write will succeed.

For relative output, other link-handling modes, and full option tables, use the realpath reference.

Related Content

Edit this page