Replace O(n^2) interval merging with sort-then-merge (#15)

mergeIntervals used a nested loop to find merge candidates, making
union-heavy operations cubic when combined with iterative Union calls
in npm || and hex or parsing. Now sorts intervals by lower bound
and merges in a single linear pass. The npm and hex parsers also
collect all intervals first instead of merging incrementally.

Author: andrew

parser.go

@@ -260,23 +260,30 @@ func (p *Parser) parseNpmRange(s string) (*Range, error) {
 		return Unbounded(), nil
 	}
 
-	// Handle || (OR)
+	// Handle || (OR) -- collect all parts, then merge once
 	if strings.Contains(s, "||") {
 		parts := strings.Split(s, "||")
-		var result *Range
+		var allIntervals []Interval
+		var allExclusions []string
+		var allRaw []Interval
 		for _, part := range parts {
-			// Each OR part may contain AND constraints, so recurse
 			r, err := p.parseNpmRange(strings.TrimSpace(part))
 			if err != nil {
 				return nil, err
 			}
-			if result == nil {
-				result = r
+			allIntervals = append(allIntervals, r.Intervals...)
+			allExclusions = append(allExclusions, r.Exclusions...)
+			if len(r.RawConstraints) > 0 {
+				allRaw = append(allRaw, r.RawConstraints...)
 			} else {
-				result = result.Union(r)
+				allRaw = append(allRaw, r.Intervals...)
 			}
 		}
-		return result, nil
+		return &Range{
+			Intervals:      mergeIntervals(allIntervals),
+			Exclusions:     allExclusions,
+			RawConstraints: allRaw,
+		}, nil
 	}
 
 	// Handle space-separated AND constraints
@@ -661,22 +668,27 @@ func (p *Parser) parseGoRange(s string) (*Range, error) {
 func (p *Parser) parseHexRange(s string) (*Range, error) {
 	s = strings.TrimSpace(s)
 
-	// Handle "or" disjunction first
+	// Handle "or" disjunction -- collect all parts, then merge once
 	if strings.Contains(s, " or ") {
 		parts := strings.Split(s, " or ")
-		var result *Range
+		var allIntervals []Interval
+		var allRaw []Interval
 		for _, part := range parts {
 			r, err := p.parseHexSingleRange(strings.TrimSpace(part))
 			if err != nil {
 				return nil, err
 			}
-			if result == nil {
-				result = r
+			allIntervals = append(allIntervals, r.Intervals...)
+			if len(r.RawConstraints) > 0 {
+				allRaw = append(allRaw, r.RawConstraints...)
 			} else {
-				result = result.Union(r)
+				allRaw = append(allRaw, r.Intervals...)
 			}
 		}
-		return result, nil
+		return &Range{
+			Intervals:      mergeIntervals(allIntervals),
+			RawConstraints: allRaw,
+		}, nil
 	}
 
 	return p.parseHexSingleRange(s)

parser_test.go

@@ -1,6 +1,10 @@
 package vers
 
-import "testing"
+import (
+	"fmt"
+	"strings"
+	"testing"
+)
 
 func TestParseVersURI(t *testing.T) {
 	tests := []struct {
@@ -530,3 +534,22 @@ func TestPublicAPISatisfies(t *testing.T) {
 		})
 	}
 }
+
+func TestNpmParseManyClauses(t *testing.T) {
+	// Build a constraint with many || parts. Before the fix, this was O(n^3)
+	// and would take seconds or minutes. Now it should complete quickly.
+	parts := make([]string, 500)
+	for i := range parts {
+		parts[i] = fmt.Sprintf(">=%d.0.0 <%d.0.0", i, i+1)
+	}
+	input := strings.Join(parts, " || ")
+
+	p := NewParser()
+	r, err := p.ParseNative(input, "npm")
+	if err != nil {
+		t.Fatalf("Parse failed: %v", err)
+	}
+	if r == nil || r.IsEmpty() {
+		t.Error("expected non-empty range")
+	}
+}

range.go

@@ -1,6 +1,9 @@
 package vers
 
-import "strings"
+import (
+	"sort"
+	"strings"
+)
 
 // Range represents a version range as a collection of intervals.
 // Multiple intervals represent a union (OR) of ranges.
@@ -198,24 +201,39 @@ func mergeIntervals(intervals []Interval) []Interval {
 		return intervals
 	}
 
-	// Simple implementation: try to merge each pair
-	result := make([]Interval, 0, len(intervals))
-
-	for _, interval := range intervals {
-		if interval.IsEmpty() {
-			continue
+	// Filter empty intervals and sort by lower bound
+	sorted := make([]Interval, 0, len(intervals))
+	for _, iv := range intervals {
+		if !iv.IsEmpty() {
+			sorted = append(sorted, iv)
 		}
+	}
+	if len(sorted) == 0 {
+		return nil
+	}
 
-		merged := false
-		for i, existing := range result {
-			if union := existing.Union(interval); union != nil {
-				result[i] = *union
-				merged = true
-				break
-			}
+	sort.Slice(sorted, func(i, j int) bool {
+		a, b := sorted[i], sorted[j]
+		if a.Min == "" && b.Min != "" {
+			return true // unbounded lower comes first
+		}
+		if a.Min != "" && b.Min == "" {
+			return false
+		}
+		cmp := CompareVersions(a.Min, b.Min)
+		if cmp != 0 {
+			return cmp < 0
 		}
-		if !merged {
-			result = append(result, interval)
+		return a.MinInclusive && !b.MinInclusive
+	})
+
+	result := []Interval{sorted[0]}
+	for _, iv := range sorted[1:] {
+		last := &result[len(result)-1]
+		if union := last.Union(iv); union != nil {
+			*last = *union
+		} else {
+			result = append(result, iv)
 		}
 	}