String Regex Match

String Regex Match is the power tool of the TakkunRed ComfyUI-String-Function pack. Where the slicing nodes and String Find work on fixed positions, this one matches patterns - and patterns are how you pull a version number out of v2.1.4-beta, grab every digit from a caption, or extract the month from a 2026-08-27 date without caring about the exact layout. If you've ever wished a string node could do a "find and capture," this is it.

It's also the escape hatch for the pack's position-based nodes. Fixed slicing breaks the moment text shifts by a character; regex holds on to a structure no matter where it sits. In a workflow full of parsed filenames, LLM output and generated text, that's the difference between a graph that survives format drift and one that silently breaks.

How it works

It's a thin, careful wrapper over Python's re module with four modes:

• search (default) - find the pattern anywhere in the text.
• match - match from the start of the text.
• fullmatch - the entire text must match the pattern.
• findall - return every match, joined into one string with join_str (default ", "). In findall, the group setting is ignored and matches are collected wholesale.

The group input (default 0) picks which capture group to return - 0 is the whole match, 1+ is a parenthesised group, handy for (\d{4})-(\d{2})-(\d{2}) style parsing where you want just the month. ignore_case and multiline (makes ^/$ match line starts/ends) round out the flags.

Two design details worth calling out. First, an invalid pattern doesn't crash the graph - it returns a string like [regex error: ...], matched=False, match_count=0, so a typo is diagnosable instead of fatal. Second, the three-way output (result, matched, match_count) lets downstream logic branch on whether anything matched at all, which is exactly what a parsing pipeline needs.

Inputs and outputs

Required: text, pattern, mode. Optional: group, ignore_case, multiline, join_str. Outputs:

• result (STRING) - the extracted match (or joined matches in findall).
• matched (BOOLEAN) - did anything match? Branch on this.
• match_count (INT) - how many matches (1 for search/match/fullmatch).

Installing it

Same pack, same zero-dependency story:

cd ComfyUI/custom_nodes
git clone https://github.com/TakkunRed/ComfyUI-String-Function

Or search "ComfyUI-String-Function" in ComfyUI Manager, restart, and grab it under String Function.

Where people get burned

• Regex is a different language, and it's case- and whitespace-sensitive. \d+ finds digits; \d+ (with a trailing space) finds digits followed by a space - subtle differences that change results. Test patterns in a scratch tool before wiring them into a workflow.
• match is not "search somewhere in the middle." It anchors to the start of the text, like Python's re.match. People coming from "match = find" habit get confused; use search for anywhere-in-the-text.
• Capturing groups change the result. With a pattern containing groups, group=0 returns the whole match including delimiters; group=1 returns just the first capture. Set group deliberately or you'll get more text than you bargained for.
• In findall, a pattern with groups returns tuples, and the node joins each tuple's pieces with join_str. A (\d+)-(\d+) pattern in findall mode gives you the two numbers joined per match, then all matches joined - read the outputs before trusting them.

When the layout is stable, String Find + String Mid is simpler. When it isn't, this is the node that keeps your workflow alive across format drift.