1hew-ComfyUI-1hewNodes/web/docs/1hew_TextPathExtract/en.md

Text Path Extract - Read values by path

Node Purpose: Text Path Extract reads a value out of the text widget with a path expression. The text can be JSON, JSON inside a Markdown fence, a Python literal, relaxed JSON (single quotes, trailing commas, True/False/None, unquoted keys) or plain text. The default path [*] splits the whole payload into its items; paths such as [0].content, [-1].bbox[0], data.items[0].zh and [*].content pick the matching content.

Inputs

Name	Type	Default	Range	Description
text	STRING	``	multiline	The payload: JSON / JSON-like / plain text
path	STRING	[*]	-	Path expression; [*] means every item, an empty path means the whole payload
output_format	COMBO	auto	auto / json / text / csv	Affects the text output only; auto means smart rendering (see below)
separator	STRING	\n	supports \n / \t / \r	Joins multiple matches (every format except json)
fallback	STRING	``	-	Returned when the path misses or the input is empty

Outputs

Name	Type	Description
text	STRING	Rendered text; with several matches json emits one array, the other formats join with separator
value	ANY	The raw value, never stringified: lists stay lists, numbers stay numbers; a list when several matches
count	INT	Number of matches; the amount itself when the path ends with .count / .length / [#]; 0 when the path misses or the input is empty. After a run the port label shows the number (e.g. 3 count)

output_format

Using path=[0].bbox (a list of coordinates), output_format only changes text, never value:

Value	Meaning	text for that example
auto	Smart: strings stay raw, lists / dicts become compact JSON, numbers and bools become text	[184,410,845,716]
json	Always JSON; strings keep their double quotes	[184,410,845,716]
text	Python str() style: lists with spaces, strings without quotes	[184, 410, 845, 716]
csv	Lists joined with a comma and a space, no brackets	184, 410, 845, 716

Path Syntax

Form	Meaning
[0] / [-1]	List index, negatives allowed
.content / content	Dict key; the leading dot is optional
[0]["content"] / [0].'content'	Quoted key, for names containing dots or spaces
$.data.items[0].zh	Optional JSONPath-style root $
[*] / [*].content / data.*.content	Wildcard: every item (the default path is [*])
[*].content.count / [0].bbox.length / [0].bbox[#]	Count: the number of matches, or the length of the matched list / dict / string
empty path	The whole payload itself

Full-width 【】, “” and ‘’ are normalised, so a Chinese IME typing 【0】.“content” still works.

Features

• Layered parsing: strict JSON, Python literal, relaxed JSON (trailing commas, True/False/None), a loose parser for unquoted keys, then a key/value scrape for truncated payloads.
• Fence cleaning: Markdown code fences and prose around the JSON block are stripped.
• Double decoding: a payload that is itself a JSON string, such as "{\"a\": 1}", is decoded once more.
• Exact keys: path keys are case-sensitive and are not synonym-matched; use Text Custom Extract when synonyms are wanted.
• Wildcard collection: [*] walks list items or dict values and applies the remaining path to each.
• Counting: the count output is always the number of matched items; a trailing .count / .length / .size / [#] returns that number as the value itself, and the count output then carries it too.
• Front-end value: after a run the count port label is prefixed with the number (e.g. 3 count), matching the count value; reconnecting an input restores the plain count label until the next run.
• Whole-set JSON: with several matches and output_format=json the output is one JSON array ([*].bbox gives [[184,410,845,716],...]) instead of one value per line; the other formats join with separator.

Typical Usage

• See every item: path=[*] prints one element per line.
• Take the first detection text: path=[0].content.
• Take the last box for a downstream node: path=[-1].bbox with output_format=json gives [690,264,992,588].
• Get every content as a list: path=[*].content leaves the list on value; with output_format=json the text output is ["犀牛","犀牛","犀牛"].
• Get every bbox as a list: path=[*].bbox with output_format=json gives [[184,410,845,716],[508,679,1000,953],[690,264,992,588]].
• Get the amount: wire the count output, or use the path [*].content.count.
• Get a comma-separated coordinate string: path=[0].bbox with output_format=csv gives 184, 410, 845, 716.
• Keep the structure: wire the value output instead of the string output.

Notes & Tips

• Blank text returns fallback directly and never raises.
• A missing path also returns fallback; leave it empty to pass an empty value downstream.
• Plain text has no items to walk, so [*] finds nothing; leave path empty to pass plain text through unchanged.
• With auto strings stay raw while lists / dicts become compact JSON; choose json when strings must keep their quotes.
• A real count key is not shadowed: if the object has that key its value wins; write ["count"] to force literal key access.
• The numeric label comes from the web/js/output_label.js extension; without it only the label is missing, the returned values are unaffected.