lookbehinds

Lookbehinds, or lookbehind assertions, are a feature of regular expressions that allow a pattern to assert what must come immediately before the current position without consuming any characters. They come in two main forms: positive lookbehind, written as (?<=...), and negative lookbehind, written as (?<!...). The pattern inside the lookbehind is used only to check the preceding text; the actual match is taken from the content that follows the assertion.

Lookbehinds are zero-width; they do not advance the match position. Most engines require the lookbehind pattern to be fixed length, so the number of characters checked is constant. Some engines or modern extensions support limited variable-length lookbehinds, but support varies by language and version. Not all languages support lookbehinds at all; for example, older JavaScript environments lacked support, while modern engines generally provide it.

Examples illustrate common uses. To match digits that follow a colon and space, you can use (?<=:\s)\d+ in a string such as "ID: 1234" — the overall match would be "1234". To match a word that is immediately preceded by the word "error", you could use (?<=\berror\s)\w+. To find tokens not preceded by a hash, you might use (?<!#)\w+. These constructs allow precise extraction without including the preceding context in the match.

In practice, lookbehinds are useful for tokenization, data extraction, and validation tasks that depend on preceding text. When porting regexes between languages, check the engine’s documentation for support and fixed-length requirements. If lookbehinds are not available, you can often achieve similar results with alternative patterns that combine capturing groups and lookaheads.