Add new tiebreak: 'chunk'

Favors the line with shorter matched chunk. A chunk is a set of
consecutive non-whitespace characters.

Unlike the default `length`, this new scheme works well with tabular input.

  # length prefers item #1, because the whole line is shorter,
  # chunk prefers item #2, because the matched chunk ("foo") is shorter
  fzf --height=6 --header-lines=2 --tiebreak=chunk --reverse --query=fo << "EOF"
  N | Field1 | Field2 | Field3
  - | ------ | ------ | ------
  1 | hello  | foobar | baz
  2 | world  | foo    | bazbaz
  EOF

If the input does not contain any spaces, `chunk` is equivalent to
`length`. But we're not going to set it as the default because it is
computationally more expensive.

Close #2285
Close #2537
- Not the exact solution to --tiebreak=length not taking --nth into account,
  but this should work. And the added benefit is that it works well even
  when --nth is not provided.
- Adding a bonus point to the last character of a word didn't turn out great.
  The order of the result suddenly changes when you type in the last
  character in the word producing a jarring effect.

CHANGELOG.md

@@ -9,13 +9,30 @@ CHANGELOG
         - Word after whitespace characters or beginning of the string
         - Word after common delimiter characters (`/,:;|`)
         - Word after other non-word characters
-      ````sh
+      ```sh
       # foo/bar.sh` is preferred over `foo-bar.sh` on `bar`
-      fzf --query bar --height 4 << EOF
+      fzf --query=bar --height=4 << EOF
       foo-bar.sh
       foo/bar.sh
       EOF
       ```
+- Added a new tiebreak `chunk`
+    - Favors the line with shorter matched chunk. A chunk is a set of
+      consecutive non-whitespace characters.
+    - Unlike the default `length`, this scheme works well with tabular input
+      ```sh
+      # length prefers item #1, because the whole line is shorter,
+      # chunk prefers item #2, because the matched chunk ("foo") is shorter
+      fzf --height=6 --header-lines=2 --tiebreak=chunk --reverse --query=fo << "EOF"
+      N | Field1 | Field2 | Field3
+      - | ------ | ------ | ------
+      1 | hello  | foobar | baz
+      2 | world  | foo    | bazbaz
+      EOF
+      ```
+    - If the input does not contain any spaces, `chunk` is equivalent to
+      `length`. But we're not going to set it as the default because it is
+      computationally more expensive.
 - Bug fixes and improvements
 
 0.31.0

man/man1/fzf-tmux.1

@@ -21,7 +21,7 @@ LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
 OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
 THE SOFTWARE.
 ..
-.TH fzf-tmux 1 "Jul 2022" "fzf 0.31.0" "fzf-tmux - open fzf in tmux split pane"
+.TH fzf-tmux 1 "Aug 2022" "fzf 0.32.0" "fzf-tmux - open fzf in tmux split pane"
 
 .SH NAME
 fzf-tmux - open fzf in tmux split pane

man/man1/fzf.1

@@ -21,7 +21,7 @@ LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
 OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
 THE SOFTWARE.
 ..
-.TH fzf 1 "Jul 2022" "fzf 0.31.0" "fzf - a command-line fuzzy finder"
+.TH fzf 1 "Aug 2022" "fzf 0.32.0" "fzf - a command-line fuzzy finder"
 
 .SH NAME
 fzf - a command-line fuzzy finder
@@ -95,6 +95,8 @@ Comma-separated list of sort criteria to apply when the scores are tied.
 .br
 .BR length "  Prefers line with shorter length"
 .br
+.BR chunk "   Prefers line with shorter matched chunk (delimited by whitespaces)"
+.br
 .BR begin "   Prefers line with matched substring closer to the beginning"
 .br
 .BR end "     Prefers line with matched substring closer to the end"

src/options.go

@@ -35,7 +35,7 @@ const usage = `usage: fzf [options]
     --tac                 Reverse the order of the input
     --disabled            Do not perform search
     --tiebreak=CRI[,..]   Comma-separated list of sort criteria to apply
-                          when the scores are tied [length|begin|end|index]
+                          when the scores are tied [length|chunk|begin|end|index]
                           (default: length)
 
   Interface
@@ -125,6 +125,7 @@ type criterion int
 
 const (
 	byScore criterion = iota
+	byChunk
 	byLength
 	byBegin
 	byEnd
@@ -611,6 +612,7 @@ func parseKeyChords(str string, message string) map[tui.Event]string {
 func parseTiebreak(str string) []criterion {
 	criteria := []criterion{byScore}
 	hasIndex := false
+	hasChunk := false
 	hasLength := false
 	hasBegin := false
 	hasEnd := false
@@ -627,6 +629,9 @@ func parseTiebreak(str string) []criterion {
 		switch str {
 		case "index":
 			check(&hasIndex, "index")
+		case "chunk":
+			check(&hasChunk, "chunk")
+			criteria = append(criteria, byChunk)
 		case "length":
 			check(&hasLength, "length")
 			criteria = append(criteria, byLength)
@@ -640,6 +645,9 @@ func parseTiebreak(str string) []criterion {
 			errorExit("invalid sort criterion: " + str)
 		}
 	}
+	if len(criteria) > 4 {
+		errorExit("at most 3 tiebreaks are allowed: " + str)
+	}
 	return criteria
 }
 

src/result.go

@@ -49,6 +49,21 @@ func buildResult(item *Item, offsets []Offset, score int) Result {
 		case byScore:
 			// Higher is better
 			val = math.MaxUint16 - util.AsUint16(score)
+		case byChunk:
+			b := minBegin
+			e := maxEnd
+			l := item.text.Length()
+			for ; b >= 1; b-- {
+				if unicode.IsSpace(item.text.Get(b - 1)) {
+					break
+				}
+			}
+			for ; e < l; e++ {
+				if unicode.IsSpace(item.text.Get(e)) {
+					break
+				}
+			}
+			val = util.AsUint16(e - b)
 		case byLength:
 			val = item.TrimLength()
 		case byBegin, byEnd:

src/result_test.go

@@ -54,9 +54,9 @@ func TestResultRank(t *testing.T) {
 	// FIXME global
 	sortCriteria = []criterion{byScore, byLength}
 
-	strs := [][]rune{[]rune("foo"), []rune("foobar"), []rune("bar"), []rune("baz")}
+	str := []rune("foo")
 	item1 := buildResult(
-		withIndex(&Item{text: util.RunesToChars(strs[0])}, 1), []Offset{}, 2)
+		withIndex(&Item{text: util.RunesToChars(str)}, 1), []Offset{}, 2)
 	if item1.points[3] != math.MaxUint16-2 || // Bonus
 		item1.points[2] != 3 || // Length
 		item1.points[1] != 0 || // Unused
@@ -65,7 +65,7 @@ func TestResultRank(t *testing.T) {
 		t.Error(item1)
 	}
 	// Only differ in index
-	item2 := buildResult(&Item{text: util.RunesToChars(strs[0])}, []Offset{}, 2)
+	item2 := buildResult(&Item{text: util.RunesToChars(str)}, []Offset{}, 2)
 
 	items := []Result{item1, item2}
 	sort.Sort(ByRelevance(items))
@@ -98,6 +98,23 @@ func TestResultRank(t *testing.T) {
 	}
 }
 
+func TestChunkTiebreak(t *testing.T) {
+	// FIXME global
+	sortCriteria = []criterion{byScore, byChunk}
+
+	score := 100
+	test := func(input string, offset Offset, chunk string) {
+		item := buildResult(withIndex(&Item{text: util.RunesToChars([]rune(input))}, 1), []Offset{offset}, score)
+		if !(item.points[3] == math.MaxUint16-uint16(score) && item.points[2] == uint16(len(chunk))) {
+			t.Error(item.points)
+		}
+	}
+	test("hello foobar goodbye", Offset{8, 9}, "foobar")
+	test("hello foobar goodbye", Offset{7, 18}, "foobar goodbye")
+	test("hello foobar goodbye", Offset{0, 1}, "hello")
+	test("hello foobar goodbye", Offset{5, 7}, "hello foobar") // TBD
+}
+
 func TestColorOffset(t *testing.T) {
 	// ------------ 20 ----  --  ----
 	//   ++++++++        ++++++++++

test/test_go.rb

@@ -754,6 +754,20 @@ def test_tiebreak_length_with_nth
     assert_equal output, `#{FZF} -fh -n2 -d: < #{tempname}`.lines(chomp: true)
   end
 
+  def test_tiebreak_chunk
+    writelines(tempname, [
+                 '1 foobarbaz baz',
+                 '2 foobar baz',
+                 '3 foo barbaz'
+               ])
+
+    assert_equal [
+      '3 foo barbaz',
+      '2 foobar baz',
+      '1 foobarbaz baz'
+    ], `#{FZF} -fo --tiebreak=chunk < #{tempname}`.lines(chomp: true)
+  end
+
   def test_invalid_cache
     tmux.send_keys "(echo d; echo D; echo x) | #{fzf('-q d')}", :Enter
     tmux.until { |lines| assert_equal '  2/3', lines[-2] }