Fix comparison algorithm

Fix the file line ID comparison algorithm to use a Jaccard Index. This
correctly identifies the number of documents in the corpus.

Also add a little status output when files are being ordered by
timestamp.

.gitignore

@@ -2,4 +2,5 @@
 output.*.txt
 .vscode
 files
-files.*/
+files.*
+*.prof

main.go

@@ -9,6 +9,7 @@ import (
 	"os"
 	"path"
 	"runtime"
+	"runtime/pprof"
 	"slices"
 	"strconv"
 	"strings"
@@ -43,6 +44,16 @@ func main() {
 		start   = time.Now()
 		padding = "\n"
 	)
+
+	f, err := os.Create("cpu.prof")
+	if err != nil {
+		panic(err)
+	}
+	if err := pprof.StartCPUProfile(f); err != nil {
+		panic(err)
+	}
+	defer pprof.StopCPUProfile()
+
 	documents, output, err := run(os.Args)
 	if err != nil {
 		fmt.Fprintf(os.Stderr, "error: %v\n", err)
@@ -95,7 +106,7 @@ func run(args []string) ([]*Document, *os.File, error) {
 
 	// The files need to be processed in order of time, so determine the
 	// timestamp of each file and sort them by time.
-	fileTimes, times, err := orderFiles(dataFilePath)
+	fileTimes, times, err := orderFiles(dataFilePath, shouldReportStatus)
 	if err != nil {
 		return nil, nil, err
 	}
@@ -308,29 +319,16 @@ func (dm *DocumentManager) maybeAssociateFileWithDocument(workItem WorkItem, wor
 // on the number of line-centric differences between the contents of the two
 // files.
 func (dm *DocumentManager) compareFiles(f1Number, f2Number int) (float64, error) {
-	f1, err := dm.fcc.GetFileContents(f1Number)
+	f1, err := dm.fcc.GetFileLineID(f1Number)
 	if err != nil {
 		return 0, fmt.Errorf("file %d: %w", f1Number, err)
 	}
-	f2, err := dm.fcc.GetFileContents(f2Number)
+	f2, err := dm.fcc.GetFileLineID(f2Number)
 	if err != nil {
 		return 0, fmt.Errorf("file %d: %w", f2Number, err)
 	}
 
-	histogram := make(map[string]int)
+	similarity := compareFileLineIDs(f1, f2)
-	for _, line := range f1 {
-		histogram[line]++
-	}
-	for _, line := range f2 {
-		histogram[line]--
-	}
-
-	var differences float64
-	for _, v := range histogram {
-		differences += math.Abs(float64(v))
-	}
-
-	similarity := 1 - (differences / float64(len(f1)+len(f2)))
 	return similarity, nil
 }
 
@@ -382,6 +380,7 @@ func (d *Document) SortAssociatedFiles() {
 type FileContentsCache struct {
 	BaseDir string
 	cache   sync.Map
+	lineIDs sync.Map
 }
 
 // GetFileContents returns the contents of a file, excluding the first timestamp
@@ -418,6 +417,69 @@ func (fcc *FileContentsCache) GetFileContents(fileNumber int) ([]string, error)
 	return lines, nil
 }
 
+func (fcc *FileContentsCache) GetFileLineID(fileNumber int) ([]uint64, error) {
+	if cachedLineIDs, ok := fcc.lineIDs.Load(fileNumber); ok {
+		return cachedLineIDs.([]uint64), nil
+	}
+
+	lines, err := fcc.GetFileContents(fileNumber)
+	if err != nil {
+		return nil, fmt.Errorf("adding line IDs for file %d: %w", fileNumber, err)
+	}
+
+	lineIDs := makeFileLinesID(lines)
+	fcc.lineIDs.Store(fileNumber, lineIDs)
+	return lineIDs, nil
+}
+
+var (
+	fileLinesCache   sync.Map
+	fileLineIDSource atomic.Uint64
+)
+
+func makeFileLinesID(fileLines []string) []uint64 {
+	fileLineIDs := make([]uint64, 0, len(fileLines))
+	for _, line := range fileLines {
+		lineID, ok := fileLinesCache.Load(line)
+		if !ok {
+			newID := fileLineIDSource.Add(1)
+			fileLinesCache.Store(line, newID)
+			fileLineIDs = append(fileLineIDs, newID)
+		} else {
+			fileLineIDs = append(fileLineIDs, lineID.(uint64))
+		}
+	}
+	slices.Sort(fileLineIDs)
+	return fileLineIDs
+}
+
+func compareFileLineIDs(f1, f2 []uint64) float64 {
+	var (
+		i, j       = 0, 0
+		count      int
+		similarity int
+	)
+	for i < len(f1) && j < len(f2) {
+		count++
+		if f1[i] == f2[j] {
+			similarity++
+			i++
+			j++
+		} else if f1[i] < f2[j] {
+			i++
+		} else {
+			j++
+		}
+	}
+
+	count += len(f1) - i + len(f2) - j
+	if count == 0 {
+		return 0
+	}
+
+	return float64(similarity) / float64(count)
+}
+
 // ClearFilesExcept removes the contents of the fileContentsCache except for the
 // provided file numbers. This helps conserve memory by removing the contents of
 // files that are no longer of interest, which we can be sure of since we are
@@ -468,14 +530,20 @@ func readFileTime(filepath string) (int, error) {
 // the map can be iterated in order of time. This allows stepping through the
 // history of the files from the beginning. Using this, we can construct a
 // "chain" of evolution for a given document.
-func orderFiles(dir string) (map[int][]int, []int, error) {
+func orderFiles(dir string, shouldReportStatus bool) (map[int][]int, []int, error) {
+	if shouldReportStatus {
+		defer fmt.Fprintln(os.Stderr)
+	}
 	timeMap := make(map[int][]int)
 
 	dirEntries, err := os.ReadDir(dir)
 	if err != nil {
 		return nil, nil, fmt.Errorf("reading directory %s: %w", dir, err)
 	}
-	for _, entry := range dirEntries {
+	for i, entry := range dirEntries {
+		if shouldReportStatus {
+			fmt.Fprintf(os.Stderr, "\rReading directory %d of %d...", i+1, len(dirEntries))
+		}
 		if entry.IsDir() || !strings.HasSuffix(entry.Name(), ".txt") {
 			continue
 		}