How the rules match, and the details that trip people up
Each rule replaces every match at once, like Replace All in a word processor, and the rules run from top to bottom. Order matters: with cat → dog followed by dog → bird, every cat ends up as bird. Matching ignores upper and lower case unless Match case is ticked, using Unicode case folding, so Ä matches ä and Σ matches σ. A plain rule matches the text exactly as typed: a dot is a dot, and $ in the replacement is a dollar sign.
Line breaks are the change people ask for most. With the line-break option on (the default), \n in Find or Replace with means a line break and \t a tab, so the preset Line breaks → commas is simply \n replaced by a comma and a space. Turn the option off when you search for a real backslash, for example in a Windows path such as C:\new. Windows line endings (CR LF) are read as plain line breaks, so one rule covers files from any system.
Pattern (regex) uses JavaScript regular expressions with Unicode support, and ^ and $ match at the start and end of every line. In the replacement, $1 to $99 insert numbered groups, $<name> a named group, $& the whole match and $$ a dollar sign. The pattern (\d{4})-(\d{2})-(\d{2}) with $3/$2/$1 turns 2026-11-12 into 12/11/2026, and ^(.+)$ with '$1', wraps each line in quotes. Some patterns, such as (a+)+ on a long run of letters, can take minutes to fail, so patterns run in a background thread and are stopped after 4 seconds with a message.
Whole words only checks that a match does not touch a letter, accent mark, digit or underscore on either side, in any script. The common \b shortcut knows only the English letters A to Z, which is why it splits catégorie or a Hindi word in tools that rely on it. In our test run, replacing colour with color in 1,054,000 characters (51,000 matches) took 7 ms, and three rules, including a pattern and a line-break join, took 22 ms. Changed lines lists up to 100 edited lines with their text before and after, whenever the rules keep the number of lines.