String Manipulation in Bash: Trim and Replace

Bash string manipulation lives in parameter expansion — no external commands, no subprocesses, just ${var...} doing trims, slices, replacements, and case changes in place. For the string work that used to mean piping through sed, this is the in-process answer.

Trimming from the edges #

Four operators strip from the front (#) or the end (%), each in short and long forms:

$ f="archive.tar.gz"
$ echo "${f#*.}"      # shortest match from the front: tar.gz
$ echo "${f##*.}"     # longest:                gz
$ echo "${f%.*}"      # shortest from the end:  archive.tar
$ echo "${f%%.*}"     # longest:                archive

The mnemonic: # sits before % on the keyboard, front before back; doubled means greediest. The canonical uses fall straight out of this table — basename is ${path##*/}, extension stripping is ${file%.*}, and the bulk rename idiom "${f%.jpeg}.jpg" is a single expansion.

Replacing and slicing #

$ url="http://old.example.com/x"
$ echo "${url//old.example.com/aospix.com}"
http://aospix.com/x

Substring indexing is zero-based and counts characters, not bytes — with UTF-8 that distinction occasionally matters.

Case and defaults #

Bash 4 added case operators, and the default forms from variables round out the toolkit:

$ echo "${name^}"     # capitalize first char
$ echo "${name^^}"    # all upper
$ echo "${name,,}"    # all lower
$ file="${1:-default.conf}"   # default when unset
$ dir="${dir:?need a directory}"   # fail loudly instead

When to graduate #

Expansion is ideal for single, simple transforms in scripts. The graduation points are worth naming: patterns involving / need escaping gymnastics, multiple sequential edits get unreadable, and anything involving whole words or character classes is regex work — that is sed and awk territory, or a real language.

The edge-stripping operators pair with globbing (same pattern language); the default forms are strict mode's best friend. The scripting series continues.