split (data list)

Plain string split breaks on one literal separator, and that's usually not what your data looks like. Your captions are separated by commas and newlines. Your filenames use -, _, and spaces. Your config dumps have arbitrary whitespace runs. This node is re.split() - split on a pattern - with the results returned as a data list, so each piece becomes its own item ComfyUI can process individually.

That data-list output is the reason this variant exists alongside the "(LIST)" one. Split "red, green , blue" on \s*,\s* and you get red, green, blue as three separate items ready to flow into per-item slots - the difference between "I can batch these" and "I have one list variable I must index into." For turning messy delimited text into a fan-out of clean items, this is the node.

How it works

Straight re.split(pattern, string). Three behaviors to internalize:

• The separator disappears. re.split(r"\s*,\s*", "a, b") → ["a", "b"]. If you want the separators kept, wrap the pattern in a group - re.split keeps captured separators in the output.
• Empty pieces survive. Trailing separators, or a leading match, produce "" entries. "a,b," → ["a", "b", ""].
• It's regex, so the split point can be a whole structure - "(\d+)" splits on every number, which is a genuinely useful trick for pulling text apart into alternating chunks.

The class declares OUTPUT_IS_LIST, so the pieces come out as a data list - each string an individual item rather than one bundled LIST.

Inputs and output

• string - the text to split.
• pattern - the regex separator.

Output is a data list of STRINGs. No match at all just means the whole input comes back as a single-element list - which is correct behavior, though it surprises people expecting an error.

Installing it

Part of the Basic data handling pack by StableLlama:

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

Restart, or grab it via ComfyUI Manager (search "Basic data handling"). Zero dependencies, no models - nothing at install to go wrong.

Gotchas

• Empty strings in the output are normal, not a bug. Filter them with a data-list filter node before they hit something that chokes on "".
• If you're splitting on something you want to keep (like \n when rebuilding lines), wrap it in a group.
• Greedy quantifiers in the separator pattern can merge or skip separators in surprising ways. Test the pattern on a scratch node with your real data before trusting a batch pipeline to it.