Build a list of find-and-replace rules with regex, case and whole-word options, and watch each rule's match count as it runs on the output of the one above.
Waiting for text and at least one rule
Rules run in order, each on the result of the one above it. Nothing is uploaded.
The rewritten text appears here
Every rule applied in order, with each one's match count beside it.
Find and Replace runs entirely on your device — nothing is uploaded.
You’re on the Free plan. Compare plans
One find-and-replace is what a text editor is for. A list of them applied in order is the thing an editor makes awkward, and it is the only reason this page exists. Anonymising a log, rewriting a set of import paths, normalising the quirks of a vendor's export — all of them are five or six replacements that have to happen together, every time, in the same order.
Rules cascade, and the page says so above the list rather than letting it be discovered. Each rule runs on the output of the one before it, so replacing a with b and then b with c turns every original a into c. That is almost never what the person writing the second rule intended. The rules are numbered in the order they run and each reports its own match count against the text as it reached that rule, which is how you see a later rule eating an earlier rule's output.
Each rule carries its own options because a list where one rule needs case sensitivity and another does not is the normal case, not the exception. Whole-word matching is a convenience that writes the word boundaries for you; in regex mode it is disabled, because the pattern can say it better and two mechanisms fighting over the same job produce surprises.
A pattern that does not compile is reported beside its own rule and passed over for that run, while every other rule still applies. In regex mode a half-typed pattern is the normal state of affairs while someone is typing it, and failing the whole operation because rule four has an unclosed bracket would make the list impossible to build.
A rule can be switched off without being deleted, which is what you want when working out which of two rules is responsible for an unexpected result. Nothing is uploaded: the text and the rules stay in the browser, which is what makes this usable on a log that still has real data in it.
Because each rule runs on the result of the one above it. Replacing a with b and then b with c turns every original a into c. The rules are numbered in running order and each shows its own count so the cascade is visible.
That rule reports its error and is skipped for the run; every other rule still applies. A half-typed pattern is normal while you are writing one, so the list stays usable.
Yes. In regex mode the replacement supports the usual dollar-number references, so a pattern with groups can reorder or reuse what it matched.
Because the pattern can express it directly with a word-boundary assertion, and having two mechanisms competing to insert boundaries produces results that are hard to reason about. Write the boundary into the pattern instead.
No, it is ignored. A rule with nothing to find does nothing, which is the useful behaviour while you are part way through writing one.
No. Both the text and the rules stay in your browser and nothing reaches a server or analytics, which is what makes this safe for a log containing real data.