Using query-replace-regex to partially replace

Question

I'm struggling with the basics of query-replace-regex after reading the manual.

I'm trying to replace lots of citations formatted like this: #123213213, foo bar

to #123213213\, foo bar

the numbers vary from instance to instance and can be of any length. They always start with a #. Foo bar can be numbers or text. I always want to replace only:

, --> \,

without changing the numbers or the foo bar.

I've tried emacs' regex builder, and I'm stuck at "\#[0-9]*,\s.*\} which doesn't work because it selects too much before the bit I want. Ideally, I'm looking for an answer that explains how it works. Thanks for your help!

Answer 1

Assuming the trailing arbitrary text can legitimately contain commas, so you can't safely replace every , with \, throughout:

1. You don't need to escape the #.
2. Based on your description, you don't need the \}
3. (minor point) You always expect at least one digit in the leading number, correct? Use + rather than *.

Try this:

(query-replace-regexp "#\\([0-9]+\\), \\(.*\\)" "#\\1\\\\, \\2")

Interactively, you halve the backslashes:

#\([0-9]+\), \(.*\)
RET
#\1\\, \2
RET