search groups (LIST)

Half the regex family in this pack comes in two flavors - "data list" and "LIST" - and this is the LIST one: run a search, collect the capture groups, and hand them back as a single Python list variable rather than fanning them out into individually processed items. Same regex, same matches; the difference is entirely about how you want to consume the result.

Reach for this when you want the collection intact: hold all the groups, index into them with a LIST get item node, count them, or convert to a SET for dedup. Reach for "search groups (data list)" when each group should flow into its own per-item slot downstream. If you only need one value out of a structured string, the RegexGroupDict node is usually the better tool - labels beat indices every time. But when your pattern's groups have a fixed, well-known order, a positional LIST is perfectly readable and this is the node.

How it works

One re.search(pattern, string), then list(match.groups()). The semantics that keep tripping people:

• groups() returns only the parenthesized captures, in order, excluding the full match. Pattern (\d+)x(\d+) on "1920x1080" → ["1920", "1080"].
• Only the first match counts; this is a search, not a scan-everywhere.
• No match → empty LIST, no error. A length check downstream is how you test for "did it match."
• Non-participating optional groups come through as None, which then needs handling before any string-consuming node.

Inputs and output

• string - text to search.
• pattern - regex with (...) groups.

Output is one LIST of capture-group strings, in pattern order. Index it with a list accessor, or hand the whole thing to a SET/LIST node.

Installing it

Ships in Basic data handling by StableLlama:

cd ComfyUI/custom_nodes
git clone https://github.com/StableLlama/ComfyUI-basic_data_handling

Restart, or install via ComfyUI Manager (search "Basic data handling"). The pack declares no dependencies and downloads no models - pure stdlib, so nothing at install time to fight your other nodes.

Gotchas

• Position-based, so it's brittle to pattern edits: changing group order silently reorders your data. If the string format is out of your control and stable, fine; if not, the named-group dict node is the safer read.
• None entries from non-participating groups are real and will surface as confusing errors two nodes down the line. Filter empties early.
• A malformed regex fails with re.error at this node; a valid-but-wrong one returns an empty list. Know which failure mode you're looking at when results vanish.