Find and replace

Replacing every instance and knowing how many there were

The value of doing this in a tool rather than by hand is the count. A replacement that reports three when you expected thirty tells you the search was wrong before the result goes anywhere.

1. 1Paste the text. It stays in the page, which is the reason to use this rather than an online editor when the text is a contract or a client document.
2. 2Type what to find and what to replace it with. The result updates as you type, so you can watch the count settle.
3. 3Check the count before copying. Zero usually means a trailing space in the search box or a smart apostrophe in the text that does not match the straight one you typed.
4. 4Turn on match case when replacing a name or an identifier, so Apple and apple are not treated as the same word.
5. 5Turn on whole word when the search term is a fragment of longer words, which is what stops cat becoming dog inside category.

Match case and whole word, and when each one saves you

These two options are the difference between a clean replacement and a mess you have to unpick by reading the whole document. Both are off by default because most replacements are simple.

• Match case off treats Word, word and WORD as the same target, which is what you want when fixing a term that appears in headings and in prose.
• Match case on is essential for code, identifiers and brand names, where the capitalisation carries meaning rather than style.
• Whole word off replaces every occurrence including the ones inside longer words, which is right for a substring you know is unique.
• Whole word on requires a non-word character or the end of the text on both sides, so replacing "art" leaves "start" and "particle" alone.
• Whole word uses Unicode letter and digit boundaries rather than plain ASCII, so it behaves correctly in accented and non-Latin text.
• The two combine: match case with whole word off will still hit the inside of a longer word, which is the combination people trip over most often.

The search is literal, and that is deliberate

Most replacements are ordinary text, and a tool that quietly treats the search box as a pattern turns a full stop into a wildcard that matches everything. Here what you type is what it looks for.

• A full stop matches a full stop, not any character, so replacing "9.99" does not also match "9x99".
• A dollar sign, brackets, a plus, a question mark and a backslash are all matched literally, which is what makes prices, code snippets and file paths safe to search for.
• Multi-word phrases work as typed, including the spaces between them, so a phrase replacement does not need any escaping.
• Emoji and non-Latin scripts can be searched for directly, because the matching runs in Unicode mode rather than over bytes.
• A straight apostrophe will not match a curly one. If a search that should hit returns zero, run the text through the text cleaner first and try again.
• The replacement is literal too, so a dollar sign in the replacement box stays a dollar sign instead of being read as a capture group reference.

What this tool does not try to be

It is a single-pass replacement over one block of text, and several jobs people bring to it want a different tool. Saying which ones keeps you from fighting the page.

• There is no regular expression mode. Patterns, capture groups and backreferences belong in an editor, and offering them here would make the literal guarantee above impossible.
• There is no undo history beyond your own paste. The original is still in the input box, so the recovery is to clear the search rather than to step backwards.
• It replaces in one block of text, not across files. For a codebase, an editor with project search does the job and shows you each hit in context.
• It does not preserve the case of what it replaced, so replacing "colour" with "color" in a heading gives you exactly what you typed.
• It is not a translation or rewriting tool. It changes the strings you name and leaves every other word alone, which is the whole promise.