CLI · MATCHING

Phrase search

Atom order

Each atom produces source candidates through an independent matcher. The phrase matcher connects a following candidate only when it begins after the previous token ends.

By default, a phrase stays within one logical line. It never connects across a newline, so source-line locations remain intact.

Gap

--max-gap limits the number of Unicode scalars between the previous token end and the next token start. The default of 24 includes whitespace and punctuation.

It is neither a UTF-8 byte count nor a UTF-16 code-unit count. Emoji and combining characters follow the same scalar rule.

Phrase span

The result spans from the first atom core start through the last consumed token end. Every atom retains separate core and token spans plus provenance.

Overlapping combinations are sorted deterministically and deduplicated. Analyses that share a surface form merge their origin sets rather than discarding provenance.