🏆 I provide private lessons on Emacs, Linux, and Life in general: https://protesilaos.com/coach/. Lessons continue throughout the year.

Use 'M-x query-replace-regexp' to downcase matches

Earlier today I wanted to convert a bunch of Org meta data keywords to lower case notation, with the help of isearch-forward-regexp and query-replace-regexp. This included everything from special comments, like #+TITLE: to property drawers in the form of :PROPERTIES:. There were 340 such cases in my file and I was reluctant to do so manually. Thankfully, Emacs makes such a task fairly simple once you get past the essentials, because it lets you evaluate arbitrary Elisp forms for the replacement text. This includes the possibility of running some function on a matched regexp group. In this case that function was (downcase OBJECT).

First I needed a regular expression that would capture all targets. isearch-forward-regexp and its query-replace-regexp counterpart expect a single backslash for the escape character, so I ended up with this pattern:

^\(#\|:\)[^ ].*?:

If I were to test this in the buffer, I could use M-x re-builder, which however requires double backslashes (same for when you write Elisp code):

^\\(#\\|:\\)[^ ].*?:
  • Once I was sure of the regular expression I had to use, I went to the top of the buffer and invoked isearch-forward-regexp (bound to C-M-s by default).

  • Entered the pattern ^\(#\|:\)[^ ].*?:, got live feedback of the matching items, and switched to query-replace with C-% (isearch-query-replace). When the Isearch is regexp-aware so is the corresponding query-replace.

  • For the replacement text I instructed the command to evaluate the downcase function. This is done by escaping the comma operator (,) and then supplying the function with a regexp group. Because I wished to match everything, the group should be \0. Which means that the replacement should be expressed thus: \,(downcase \0).

Try this with any string in a buffer, say, Hello:

  • M-x query-replace-regexp
  • Hello
  • \,(upcase \0)

Confirm the operation and you should get HELLO. This works with multiple groups and can read a series of Elisp forms. An example with Hello world:

  • M-x query-replace-regexp
  • \(Hello\) \(world\)
  • \,(downcase \1) \,(capitalize \2)

Which should give you hello World.

The elegant minimalism of the query-replace interface grants you the power to either replace each match one at a time or hit the exclamation mark (!) to answer “yes to all”. This is what I did to downcase all 340 matches. Voila! All of my Org file’s meta data were converted to lower case in one go.

Finally, I discovered query-replace-regexp-eval which saves you from adding the escaped comma operator for the replacement’s Elisp form. However its doc string reads thus:

Interactive use of this function is deprecated in favor of the
‘\,’ feature of ‘query-replace-regexp’.  For non-interactive use, a loop
using ‘search-forward-regexp’ and ‘replace-match’ is preferred.