Merge pull request #2979 from rockwotj/microbench

internal/impl/snowflake/streaming/compat.go

@@ -22,6 +22,7 @@ import (
 	"strconv"
 	"strings"
 	"time"
+	"unicode/utf8"
 
 	"github.com/redpanda-data/connect/v4/internal/impl/snowflake/streaming/int128"
 )
@@ -147,7 +148,38 @@ func normalizeColumnName(name string) string {
 		}
 		// fallthrough
 	}
-	return strings.ToUpper(strings.ReplaceAll(name, `\ `, ` `))
+	// Add a fast path if there is no escaping note that this is an optimized version of
+	//   strings.ToUpper(strings.ReplaceAll(name, `\ `, ` `))
+	// which indeed we fallback to that if we get unicode or any escaped spaces.
+
+	// First check to see if the name is already normalized, in that case we can save
+	// an alloc however most strings I assume are in snake or camel casing so those
+	// will likely just check the first byte in this loop then bail, so this extra
+	// loop allows for still optimizing performance over just calling into the stdlib.
+	hasLower := false
+	for _, c := range []byte(name) {
+		if 'a' <= c && c <= 'z' {
+			hasLower = true
+			break // must alloc
+		} else if c >= utf8.RuneSelf || c == '\\' {
+			// Fallback
+			return strings.ToUpper(strings.ReplaceAll(name, `\ `, ` `))
+		}
+	}
+	if !hasLower {
+		return name
+	}
+	transformed := []byte(name)
+	for i, c := range transformed {
+		if 'a' <= c && c <= 'z' {
+			c -= 'a' - 'A'
+			transformed[i] = c
+		} else if c >= utf8.RuneSelf || c == '\\' {
+			// Fallback
+			return strings.ToUpper(strings.ReplaceAll(name, `\ `, ` `))
+		}
+	}
+	return string(transformed)
 }
 
 // quoteColumnName escapes an object identifier according to the

internal/impl/snowflake/streaming/compat_test.go

@@ -15,6 +15,7 @@ import (
 	"encoding/base64"
 	"encoding/hex"
 	"slices"
+	"strings"
 	"testing"
 	"time"
 
@@ -127,6 +128,25 @@ func TestColumnNormalization(t *testing.T) {
 	require.Equal(t, `foo" bar "baz`, normalizeColumnName(`"foo"" bar ""baz"`))
 }
 
+func BenchmarkColumnNormalization(b *testing.B) {
+	makeBench := func(name string) func(b *testing.B) {
+		return func(b *testing.B) {
+			var normalized string
+			for i := 0; i < b.N; i++ {
+				normalized = normalizeColumnName(name)
+			}
+			b.SetBytes(int64(len(normalized)))
+		}
+	}
+	b.Run("snake_case", makeBench("foo_bar"))
+	b.Run("camelCase", makeBench("fooBar"))
+	b.Run("upper", makeBench("FOOBAR"))
+	b.Run("small", makeBench("a"))
+	b.Run("large", makeBench(strings.Repeat("a", 128)))
+	// Appently this is German for "fuel oil recoil absorber"
+	b.Run("unicode", makeBench("heizölrückstoßabdämpfung"))
+}
+
 func TestColumnQuoting(t *testing.T) {
 	require.Equal(t, `""`, quoteColumnName(""))
 	require.Equal(t, `"FOO"`, quoteColumnName("foo"))