Language reference
Replace (substitute text)
Syntax
Replace(<string>, <find>, <replacement>)
π Replace(phone, "-", "") → digits (Contacts)Description
Replace swaps every occurrence of one piece of text for another. Use it to strip characters (replace with an empty string), reformat values, or fix known typos in bulk.
Technical Description
Replace(s, find, replacement) → STRING. Replaces all non-overlapping occurrences of the literal find within s with replacement (Java String.replace — literal, not regex). NULL string input returns NULL. PURE, DETERMINISTIC.
Examples
Strip dashes from phone numbers:
π Replace(phone, "-", "") → digits (Contacts)PROJECT Replace(phone, "-", "") -> digits (Contacts)Normalise a separator:
π Replace(path, "\\", "/") → path (Files)PROJECT Replace(path, "\\", "/") -> path (Files)Redact a token:
π Replace(note, "SECRET", "[redacted]") → note (Logs)PROJECT Replace(note, "SECRET", "[redacted]") -> note (Logs)Pushdown
SQL: folds to REPLACE(<s>, <find>, <replacement>) on all dialects. It is the one string built-in that does. Most SQL backends' REPLACE searches case-sensitively whatever the column's collation, which is what the engine does; SQL Server's searches under the collation, whose default ignores case, so there the string is handed over under a binary one: REPLACE((<s>) COLLATE Latin1_General_100_BIN2, …). The rest of the string library is evaluated in-engine because SQL's same-named function answers differently — see pushdown.
Limitations
Matching is literal text, not a regular expression. STRING arguments; a NULL string returns NULL.
Alternatives
Trim/LTrim/RTrim to remove only surrounding whitespace; Mid/Left/Right to extract rather than substitute.