Find and Replace Text

Rules in sequence, not a single replacement

Each rule is a find and replace pair with its own options. Rules run top to bottom, and each one receives the text already modified by the previous ones. That matters: a rule turning A into B followed by one turning B into C carries A all the way to C, while the reverse order stops at B. Reordering with the arrows changes the result, and disabling a rule makes it pass the text through without breaking the chain.

The counter next to each rule shows how many occurrences it finds in the text it actually receives, not in the original. When you select the second rule or a later one, the left panel switches to the input of that stage, which is the only honest way to highlight the occurrences of a rule that runs after another.

• Add, remove, reorder and disable rules.
• Occurrence and replacement counts per rule.
• A failing rule does not stop the others.
• Apply to source bakes the result and restarts the list.

One-by-one review and capture groups

The Occurrences tab highlights every match of the selected rule in the text and lets you walk through them. On each occurrence you decide: replace it or keep it. Keeping does not delete the rule nor edit the source text, it only records that this specific occurrence stays out, and the result on the right adjusts immediately. Replace all and Keep all make the decision in bulk once the pattern is clear.

In the replacement field, beyond plain text, backreference codes apply: $& returns the whole matched text, $1 returns the first capture group, $<name> returns a named group and $$ writes a literal dollar sign. The escapes \n and \t insert a line break and a tab, since the field is a single line. With accent-insensitive matching on, these codes return the real fragment from the text, accents included, not the accent-stripped copy used for searching.

• Highlight with previous and next navigation.
• Per-occurrence decision, reversible at any time.
• $&, $1, $<name>, $$ and the \n and \t escapes.
• Undo for the actions that change the source text.

Whole word and accents

The JavaScript word boundary treats cedillas and accented vowels as separators, so a naive whole word search would match inside março and ação. Here the boundary uses an explicit class covering accented Latin, Greek and Cyrillic, so searching mar with whole word ignores março and marte, and searching ação ignores reação. The option applies to both literal and regex mode, because the pattern is wrapped in a non-capturing group that leaves your own group numbering untouched.

Ignore accents solves the opposite problem, for people who type without accents or work with inconsistent source text. Searching cafe finds café, and searching acao finds ação. Matching happens over an accent-stripped copy, but positions are converted back before anything reaches the screen, so the highlight marks the right fragment and the replacement preserves the surrounding text.

A bad expression must not freeze the tab

Expressions with a quantifier inside a quantified group, such as (a+)+ or (.*)*, can push the regex engine into explosive run times. In the browser there is no way to interrupt an execution already under way without leaving the main thread, so the defense comes first: patterns in that shape are refused with a warning before they run. On top of that there is a text size cap, a pattern size cap and a time budget checked between occurrences, which stops collection and warns instead of freezing the page.

The limits in numbers: up to 400,000 characters of text, up to 400 characters per pattern, up to 20,000 occurrences per rule, up to 20 rules and files up to 2 MB in the open button. An expression with a syntax error shows the engine message under the rule and leaves the text intact, so nothing is replaced until the pattern is valid.