This is an automated email from the ASF dual-hosted git repository.
zeroshade pushed a commit to branch main
in repository https://gitbox.apache.org/repos/asf/iceberg-go.git
The following commit(s) were added to refs/heads/main by this push:
new 914040c0b perf(transforms): skip UTF-8 scan for unchanged strings
(#2222)
914040c0b is described below
commit 914040c0baad3fe09b343ab28ecd202e19b2f325
Author: Minh Vu <[email protected]>
AuthorDate: Mon Oct 19 23:54:21 2026 +0200
perf(transforms): skip UTF-8 scan for unchanged strings (#2222)
* perf(transforms): skip UTF-8 scan for unchanged truncated strings
Signed-off-by: Minh Vu <[email protected]>
* test(perf): benchmark unchanged and truncated UTF-8 strings
Signed-off-by: Minh Vu <[email protected]>
* perf(transforms): fast-path ASCII prefixes during string truncation
Signed-off-by: Minh Vu <[email protected]>
* style(transforms): satisfy nlreturn in benchmark helper
Signed-off-by: Minh Vu <[email protected]>
---------
Signed-off-by: Minh Vu <[email protected]>
---
transforms.go | 22 +++++-
transforms_string_truncate_fastpath_test.go | 110 ++++++++++++++++++++++++++++
2 files changed, 128 insertions(+), 4 deletions(-)
diff --git a/transforms.go b/transforms.go
index 5fea782eb..92895949f 100644
--- a/transforms.go
+++ b/transforms.go
@@ -804,12 +804,26 @@ func (t TruncateTransform) Transformer(src Type)
(func(any) any, error) {
ErrInvalidArgument, src)
}
-// Range yields byte indexes at UTF-8 code point boundaries, so slicing on idx
-// truncates by Unicode characters without copying the whole string to []rune.
+// truncateString truncates by UTF-8 code points without allocating.
+// Strings shorter than the width in bytes need no scan. ASCII prefixes
+// can be sliced directly, while other prefixes still use rune boundaries.
func truncateString(s string, width int) string {
- for idx := range s {
+ if width >= len(s) {
+ return s
+ }
+
+ idx := 0
+ for idx < width && s[idx] < utf8.RuneSelf {
+ idx++
+ }
+ if idx == width {
+ return s[:idx]
+ }
+
+ width -= idx
+ for offset := range s[idx:] {
if width == 0 {
- return s[:idx]
+ return s[:idx+offset]
}
width--
diff --git a/transforms_string_truncate_fastpath_test.go
b/transforms_string_truncate_fastpath_test.go
new file mode 100644
index 000000000..a5948d917
--- /dev/null
+++ b/transforms_string_truncate_fastpath_test.go
@@ -0,0 +1,110 @@
+// Licensed to the Apache Software Foundation (ASF) under one
+// or more contributor license agreements. See the NOTICE file
+// distributed with this work for additional information
+// regarding copyright ownership. The ASF licenses this file
+// to you under the Apache License, Version 2.0 (the
+// "License"); you may not use this file except in compliance
+// with the License. You may obtain a copy of the License at
+//
+// http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing,
+// software distributed under the License is distributed on an
+// "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+// KIND, either express or implied. See the License for the
+// specific language governing permissions and limitations
+// under the License.
+
+package iceberg
+
+import (
+ "strings"
+ "testing"
+)
+
+// Keep the previous implementation here to make before/after comparisons
+// reproducible on the same toolchain and machine.
+func truncateStringBefore(s string, width int) string {
+ for idx := range s {
+ if width == 0 {
+ return s[:idx]
+ }
+ width--
+ }
+
+ return s
+}
+
+func TestTruncateStringFastPath(t *testing.T) {
+ for _, tc := range []struct {
+ name, input string
+ width int
+ want string
+ }{
+ {"short_ascii_unchanged", "us-east-1", 32, "us-east-1"},
+ {"short_ascii_truncated", "us-east-1", 4, "us-e"},
+ {"unicode_unchanged", "東京🧊Berlin", 32, "東京🧊Berlin"},
+ {"unicode_truncated", "東京🧊Berlin", 4, "東京🧊B"},
+ {"utf8_bytes_exceed_width", "éé", 3, "éé"},
+ {"combining_codepoints", "e\u0301clair", 1, "e"},
+ {"invalid_utf8_prefix", "a\xffb", 2, "a\xff"},
+ {"negative_width", "iceberg", -1, "iceberg"},
+ {"exact_width", "iceberg", 7, "iceberg"},
+ {"empty", "", 1, ""},
+ } {
+ t.Run(tc.name, func(t *testing.T) {
+ if got := truncateString(tc.input, tc.width); got !=
tc.want {
+ t.Fatalf("truncateString(%q, %d) = %q, want
%q", tc.input, tc.width, got, tc.want)
+ }
+ })
+ }
+
+ // Compare every width around both byte and code-point boundaries. The
+ // invalid UTF-8 input verifies this optimization preserves the existing
+ // behavior even for strings that cannot appear in spec-compliant data.
+ for _, input := range []string{
+ "", "us-east-1", "東京🧊Berlin", "éé", "a\xffb",
strings.Repeat("é", 64),
+ } {
+ for width := 0; width <= len(input)+2; width++ {
+ if got, want := truncateString(input, width),
truncateStringBefore(input, width); got != want {
+ t.Fatalf("truncateString(%q, %d) = %q, before =
%q", input, width, got, want)
+ }
+ }
+ }
+}
+
+var truncateStringBenchmarkSink string
+
+func BenchmarkTruncateStringFastPath(b *testing.B) {
+ for _, tc := range []struct {
+ name, input string
+ width int
+ }{
+ {"ascii_short_noop", "us-east-1", 32},
+ {"ascii_short_truncate", "us-east-1", 4},
+ {"ascii_long_noop", strings.Repeat("a", 128), 256},
+ {"unicode_short_noop", "東京🧊Berlin", 32},
+ {"unicode_short_truncate", "東京🧊Berlin", 4},
+ {"unicode_long_noop", strings.Repeat("é", 64), 256},
+ {"unicode_long_truncate", strings.Repeat("é", 64), 16},
+ {"utf8_bytes_exceed_width", "éé", 3},
+ } {
+ b.Run(tc.name, func(b *testing.B) {
+ for _, implementation := range []struct {
+ name string
+ run func(string, int) string
+ }{
+ {"before", truncateStringBefore},
+ {"after", truncateString},
+ } {
+ b.Run(implementation.name, func(b *testing.B) {
+ b.ReportAllocs()
+ b.ResetTimer()
+ for range b.N {
+ truncateStringBenchmarkSink =
implementation.run(tc.input, tc.width)
+ }
+ })
+ }
+ })
+ }
+}
