access series of numbers higher than 9 with [ ]

Question

I want to change files with sed for a series of numbered filemnames (more than 10). 0 - 9 I can easily access using sed -i 's/old/new/g' myfile[0-9] but this doesn't seem to work for numbers higher than 10. How can I do this instead? like [0-50]?

Answer 1

A character class, like [0-9] or [a-f], matches a single character only, by definition. They don't match "numbers", per se -- even if the digits given are numeric, they're being viewed only as character codepoints, not numeric values. That's the same thing in fnmatch()-style patterns (used here by the shell) as they are in regular expressions.

---

If you want 0-50, with globbing behavior (matching only files that exist) that can be done by composing multiple character classes, like so:

shopt -s nullglob # if no files match, return empty result
files=( myfile[0-9] myfile[1-4][0-9] myfile5[0] )

# if list of files is nonzero, run sed:
(( ${#files[@]} )) && sed -i -e 's/old/new/g' "${files[@]}"

To explain how that works:

• myfile[1-9] matches 1-9 (if they exist)
• myfile[1-4][0-9] matches 10-49 (if they exist)
• myfile5[0] matches 50, if and only if it exists.

Putting the list of files into the array and checking the array's length makes sure you don't run sed without any filenames at all listed, which could happen otherwise because of nullglob. (Why use nullglob at all here? Because you don't want myfile5[0] being passed as a literal filename if no myfile50 exists, which is the default behavior otherwise).

---

There are some additional extensions to the POSIX sh standard available as well:

If you don't care if files exist (and want to put contents on the command line even if they don't), you can use brace expansion:

sed -i -e 's/old/new/g' myfile{0..50}

Alternately, if you simply care about matching one-or-more numeric digits at the end of the filename, you can use extglobs:

shopt -s extglob
sed -i -e 's/old/new/g' myfile+([0-9])