Queries & matching

Skeleton — structure plus verified facts; full prose to be written.

Overview

Matchmaker's matcher is built on Nucleo with fzf-style scoring: smart case, consecutive-match boosting, and start-of-word preference.

How matching works

1. Characters match in order but not necessarily consecutively; shorter gaps and consecutive runs score higher.
2. Multiple whitespace-separated tokens must each match (logical AND).
3. Smart case: a lowercase query matches case-insensitively; uppercase letters force case-sensitivity.

Query operators

abc fuzzy match, 'abc literal substring, ^abc prefix, abc$ suffix, ^abc$ exact, !abc exclude, \ escaped space. For the full reference see Queries & misc (this page will host the deeper walkthroughs).

Word boundaries

• Word-movement in the query bar and boundary-aware tokenization stop at whitespace plus configured word_boundaries characters ([query] word_boundaries, e.g. ['.', '/']), in addition to whitespace.

Matcher configuration

• [matcher] configures matching options (e.g. ansi pre-processing of ANSI sequences).
• [sort] configures initial sorting: sort mode, reverse, threshold, and column-aware sorting.
• The matcher runs on a worker thread and rescans as you type (debounce configurable via [tui] sleep_ms / [query]… — to verify exact debounce option).

To write

• Column-scoped queries and how %column filters interact with tokens.
• Ranking details: exact matches, prefix matches, positions of matches.
• Long-input performance notes.