Lesson 21 · The filter, ranked - Build a Fuzzy Finder

The filter, ranked

The filter mode from chapter one still prints in input order. Today you upgrade it to the full pipeline - parse the query, rank every candidate, and print the best matches highlighted - closing the ranking chapter with a real, ordered filter.

The goal

Make filter mode rank its output by score and highlight each printed match.

Start here - the target

TO DO

Scenario: Ranked, highlighted filter output

Giventhe input lines "src/main.go", "README.md", "go.mod", "src/app.go" and the query "go"

Whenthe tool runs in filter mode

Thenit prints three highlighted lines in score order: "src/app.[g][o]" (score 45, length 10), then "src/main.[g][o]" (score 45, length 11, after the shorter tie), then "[g][o].mod" (score 40)

Background

Chapter one’s filter printed matches in input order because that was all it could do. Now the finder has a whole pipeline - parse the query into terms, score every candidate’s best alignment, sort with tie-breaking, recover positions - so filter mode becomes its real self: the ranked, highlighted output that fzf’s -f flag produces. The two go matches at the same score order by length, and the weaker go.mod follows.

This is the tool a script or a pipe would actually use, and it exercises every piece of the back three chapters at once. It is also the natural checkpoint before going interactive: everything an interactive finder does per keystroke is exactly this pipeline, run again on a new query. The next chapter takes this batch pipeline and wraps it in state - a query you can edit, a selection you can move, a frame you can draw.

Make it work

// filter mode = parse -> rank -> sort -> highlight -> print.
results := rank(query, candidates)   // parses terms, scores, sorts
for _, r := range results {
  fmt.Println(highlight(r.Candidate, r.Positions))
}

CheckpointDONE

The non-interactive filter now returns ranked, highlighted matches. Commit and stop here.