url: speed up URLPattern by anonrig · Pull Request #65364 · nodejs/node

anonrig · 2026-08-18T01:50:22Z

This speeds up WHATWG URLPattern constructor, test(), and exec() without changing observable behavior.

Independent of the new URL() parse PR (#65361) and the URLSearchParams PR (#65363).

What changed

C++ only (src/node_url_pattern.cc):

String input: copy one-byte ASCII V8 strings with ValueView + memcpy into a stack buffer and pass a string_view into Ada. Drop the extra std::string copy that BufferValue/Utf8Value::ToString() used to make. Non-ASCII still goes through Utf8Value.
Object init: look up the nine interned component keys and assign fields by index. Skip UTF-8-converting each key name and the string-compare chain. Non-string members are still ignored.
Regexp provider: convert pattern/input strings with ToV8Value and use isolate->GetCurrentContext() instead of Environment::GetCurrent.
test / exec: share argument parsing so both paths keep the same WebIDL null/undefined/"null" baseURL behavior.

Ada still owns its parsed state; string_views only live for the duration of the C++ call. ValueView is destroyed before any V8 heap allocation.

Tests

test/parallel/test-urlpattern.js, test-urlpattern-types.js, test-urlpattern-invalidthis.js, and test-urlpattern-fast-path.js all pass
WPT test/wpt/test-urlpattern.js: 743 passed, 0 unexpected failures

Review follow-up

Dropped the local NewStringFromUtf8View helper in favor of ToV8Value
CopyV8StringToBuffer takes a pointer, matching project convention for mutating arguments
Author and Signed-off-by are Yagiz Nizipli (human DCO)

Assisted-by: Cursor

nodejs-github-bot · 2026-08-18T01:50:27Z

Review requested:

@nodejs/url

jasnell

AI agents are not permitted to use Signed-off-by

codecov · 2026-08-18T03:36:57Z

Codecov Report

❌ Patch coverage is 91.15044% with 10 lines in your changes missing coverage. Please review.
✅ Project coverage is 90.13%. Comparing base (f83e7df) to head (1ba9273).
⚠️ Report is 122 commits behind head on main.

Files with missing lines	Patch %	Lines
src/node_url_pattern.cc	91.15%	1 Missing and 9 partials ⚠️

Additional details and impacted files

@@ Coverage Diff @@## main #65364 +/- ##
==========================================
- Coverage 90.31% 90.13% -0.18% 
==========================================
Files 751 752 +1 Lines 249956 251861 +1905 Branches 47204 47350 +146 ==========================================
+ Hits 225745 227022 +1277 - Misses 15612 16177 +565 - Partials 8599 8662 +63

Files with missing lines	Coverage Δ
src/node_url_pattern.cc	`83.84% <91.15%> (+0.12%)`	⬆️

... and 96 files with indirect coverage changes

🚀 New features to boost your workflow:

❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

anonrig · 2026-08-18T11:40:12Z

Removed the Signed-off-by trailer from the commit(s). An AI agent cannot attest the DCO.

Avoid extra UTF-8 copies on constructor, test, and exec; convert regexp inputs with ToV8Value; and assign URLPatternInit fields by interned key instead of string-comparing each component name. Signed-off-by: Yagiz Nizipli <yagiz@nizipli.com> Assisted-by: Cursor

mertcanaltin

LGTM

lemire · 2026-08-19T20:48:12Z

- } else if (key == "baseURL") {
- init.base_url = std::string(value);
- }
+ std::optional<std::string>* fields[] = {


@anonrig I don't have my AI with me, but this seems overly complicated.
The structure is just made of...
std::optional<std::string> protocol{}; // If present, must be valid UTF-8. std::optional<std::string> username{}; // If present, must be valid UTF-8. std::optional<std::string> password{}; // If present, must be valid UTF-8. std::optional<std::string> hostname{}; // If present, must be valid UTF-8. std::optional<std::string> port{}; // If present, must be valid UTF-8. std::optional<std::string> pathname{}; // If present, must be valid UTF-8. std::optional<std::string> search{}; // If present, must be valid UTF-8. std::optional<std::string> hash{}; // If present, must be valid UTF-8. std::optional<std::string> base_url{};
The standard will lay them out one after the other without padding, I expect. So you can do (&init.protocol + index) to get the desired pointer, without materializing this fields array.
My idea won't impact the performance, probably, but it would require less code.
Of course, an even better option would be to
std::span<std::optional<std::string>> as_span() { return {&protocol, 9}; }
to the init structure... and then you could do as_span()[i] and that would be that.

Resolved

mcollina

lgtm

nodejs-github-bot · 2026-08-21T10:36:56Z

CI: https://ci.nodejs.org/job/node-test-pull-request/76110/

nodejs-github-bot · 2026-08-21T12:37:18Z

CI: https://ci.nodejs.org/job/node-test-pull-request/76136/

nodejs-github-bot · 2026-08-21T19:55:23Z

CI: https://ci.nodejs.org/job/node-test-pull-request/76180/

anonrig · 2026-08-22T17:28:10Z

@lemire can you review?

anonrig · 2026-08-18T01:50:22Z

This speeds up WHATWG URLPattern constructor, test(), and exec() without changing observable behavior.

Independent of the new URL() parse PR (#65361) and the URLSearchParams PR (#65363).

What changed

C++ only (src/node_url_pattern.cc):

String input: copy one-byte ASCII V8 strings with ValueView + memcpy into a stack buffer and pass a string_view into Ada. Drop the extra std::string copy that BufferValue/Utf8Value::ToString() used to make. Non-ASCII still goes through Utf8Value.
Object init: look up the nine interned component keys and assign fields by index. Skip UTF-8-converting each key name and the string-compare chain. Non-string members are still ignored.
Regexp provider: convert pattern/input strings with ToV8Value and use isolate->GetCurrentContext() instead of Environment::GetCurrent.
test / exec: share argument parsing so both paths keep the same WebIDL null/undefined/"null" baseURL behavior.

Ada still owns its parsed state; string_views only live for the duration of the C++ call. ValueView is destroyed before any V8 heap allocation.

Tests

test/parallel/test-urlpattern.js, test-urlpattern-types.js, test-urlpattern-invalidthis.js, and test-urlpattern-fast-path.js all pass
WPT test/wpt/test-urlpattern.js: 743 passed, 0 unexpected failures

Review follow-up

Dropped the local NewStringFromUtf8View helper in favor of ToV8Value
CopyV8StringToBuffer takes a pointer, matching project convention for mutating arguments
Author and Signed-off-by are Yagiz Nizipli (human DCO)

Assisted-by: Cursor

nodejs-github-bot · 2026-08-18T01:50:27Z

Review requested:

@nodejs/url

jasnell

AI agents are not permitted to use Signed-off-by

codecov · 2026-08-18T03:36:57Z

Codecov Report

❌ Patch coverage is 91.15044% with 10 lines in your changes missing coverage. Please review.
✅ Project coverage is 90.13%. Comparing base (f83e7df) to head (1ba9273).
⚠️ Report is 122 commits behind head on main.

Files with missing lines	Patch %	Lines
src/node_url_pattern.cc	91.15%	1 Missing and 9 partials ⚠️

Additional details and impacted files

@@ Coverage Diff @@## main #65364 +/- ##
==========================================
- Coverage 90.31% 90.13% -0.18% 
==========================================
Files 751 752 +1 Lines 249956 251861 +1905 Branches 47204 47350 +146 ==========================================
+ Hits 225745 227022 +1277 - Misses 15612 16177 +565 - Partials 8599 8662 +63

Files with missing lines	Coverage Δ
src/node_url_pattern.cc	`83.84% <91.15%> (+0.12%)`	⬆️

... and 96 files with indirect coverage changes

🚀 New features to boost your workflow:

❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

anonrig · 2026-08-18T11:40:12Z

Removed the Signed-off-by trailer from the commit(s). An AI agent cannot attest the DCO.

Avoid extra UTF-8 copies on constructor, test, and exec; convert regexp inputs with ToV8Value; and assign URLPatternInit fields by interned key instead of string-comparing each component name. Signed-off-by: Yagiz Nizipli <yagiz@nizipli.com> Assisted-by: Cursor

mertcanaltin

LGTM

lemire · 2026-08-19T20:48:12Z

- } else if (key == "baseURL") {
- init.base_url = std::string(value);
- }
+ std::optional<std::string>* fields[] = {


@anonrig I don't have my AI with me, but this seems overly complicated.
The structure is just made of...
std::optional<std::string> protocol{}; // If present, must be valid UTF-8. std::optional<std::string> username{}; // If present, must be valid UTF-8. std::optional<std::string> password{}; // If present, must be valid UTF-8. std::optional<std::string> hostname{}; // If present, must be valid UTF-8. std::optional<std::string> port{}; // If present, must be valid UTF-8. std::optional<std::string> pathname{}; // If present, must be valid UTF-8. std::optional<std::string> search{}; // If present, must be valid UTF-8. std::optional<std::string> hash{}; // If present, must be valid UTF-8. std::optional<std::string> base_url{};
The standard will lay them out one after the other without padding, I expect. So you can do (&init.protocol + index) to get the desired pointer, without materializing this fields array.
My idea won't impact the performance, probably, but it would require less code.
Of course, an even better option would be to
std::span<std::optional<std::string>> as_span() { return {&protocol, 9}; }
to the init structure... and then you could do as_span()[i] and that would be that.

Resolved

mcollina

lgtm

nodejs-github-bot · 2026-08-21T10:36:56Z

CI: https://ci.nodejs.org/job/node-test-pull-request/76110/

nodejs-github-bot · 2026-08-21T12:37:18Z

CI: https://ci.nodejs.org/job/node-test-pull-request/76136/

nodejs-github-bot · 2026-08-21T19:55:23Z

CI: https://ci.nodejs.org/job/node-test-pull-request/76180/

anonrig · 2026-08-22T17:28:10Z

@lemire can you review?

anonrig · 2026-08-18T01:50:22Z

This speeds up WHATWG URLPattern constructor, test(), and exec() without changing observable behavior.

Independent of the new URL() parse PR (#65361) and the URLSearchParams PR (#65363).

What changed

C++ only (src/node_url_pattern.cc):

String input: copy one-byte ASCII V8 strings with ValueView + memcpy into a stack buffer and pass a string_view into Ada. Drop the extra std::string copy that BufferValue/Utf8Value::ToString() used to make. Non-ASCII still goes through Utf8Value.
Object init: look up the nine interned component keys and assign fields by index. Skip UTF-8-converting each key name and the string-compare chain. Non-string members are still ignored.
Regexp provider: convert pattern/input strings with ToV8Value and use isolate->GetCurrentContext() instead of Environment::GetCurrent.
test / exec: share argument parsing so both paths keep the same WebIDL null/undefined/"null" baseURL behavior.

Ada still owns its parsed state; string_views only live for the duration of the C++ call. ValueView is destroyed before any V8 heap allocation.

Tests

test/parallel/test-urlpattern.js, test-urlpattern-types.js, test-urlpattern-invalidthis.js, and test-urlpattern-fast-path.js all pass
WPT test/wpt/test-urlpattern.js: 743 passed, 0 unexpected failures

Review follow-up

Dropped the local NewStringFromUtf8View helper in favor of ToV8Value
CopyV8StringToBuffer takes a pointer, matching project convention for mutating arguments
Author and Signed-off-by are Yagiz Nizipli (human DCO)

Assisted-by: Cursor

nodejs-github-bot · 2026-08-18T01:50:27Z

Review requested:

@nodejs/url

jasnell

AI agents are not permitted to use Signed-off-by

codecov · 2026-08-18T03:36:57Z

Codecov Report

❌ Patch coverage is 91.15044% with 10 lines in your changes missing coverage. Please review.
✅ Project coverage is 90.13%. Comparing base (f83e7df) to head (1ba9273).
⚠️ Report is 122 commits behind head on main.

Files with missing lines	Patch %	Lines
src/node_url_pattern.cc	91.15%	1 Missing and 9 partials ⚠️

Additional details and impacted files

@@ Coverage Diff @@## main #65364 +/- ##
==========================================
- Coverage 90.31% 90.13% -0.18% 
==========================================
Files 751 752 +1 Lines 249956 251861 +1905 Branches 47204 47350 +146 ==========================================
+ Hits 225745 227022 +1277 - Misses 15612 16177 +565 - Partials 8599 8662 +63

Files with missing lines	Coverage Δ
src/node_url_pattern.cc	`83.84% <91.15%> (+0.12%)`	⬆️

... and 96 files with indirect coverage changes

🚀 New features to boost your workflow:

❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

anonrig · 2026-08-18T11:40:12Z

Removed the Signed-off-by trailer from the commit(s). An AI agent cannot attest the DCO.

Avoid extra UTF-8 copies on constructor, test, and exec; convert regexp inputs with ToV8Value; and assign URLPatternInit fields by interned key instead of string-comparing each component name. Signed-off-by: Yagiz Nizipli <yagiz@nizipli.com> Assisted-by: Cursor

mertcanaltin

LGTM

lemire · 2026-08-19T20:48:12Z

- } else if (key == "baseURL") {
- init.base_url = std::string(value);
- }
+ std::optional<std::string>* fields[] = {


@anonrig I don't have my AI with me, but this seems overly complicated.
The structure is just made of...
std::optional<std::string> protocol{}; // If present, must be valid UTF-8. std::optional<std::string> username{}; // If present, must be valid UTF-8. std::optional<std::string> password{}; // If present, must be valid UTF-8. std::optional<std::string> hostname{}; // If present, must be valid UTF-8. std::optional<std::string> port{}; // If present, must be valid UTF-8. std::optional<std::string> pathname{}; // If present, must be valid UTF-8. std::optional<std::string> search{}; // If present, must be valid UTF-8. std::optional<std::string> hash{}; // If present, must be valid UTF-8. std::optional<std::string> base_url{};
The standard will lay them out one after the other without padding, I expect. So you can do (&init.protocol + index) to get the desired pointer, without materializing this fields array.
My idea won't impact the performance, probably, but it would require less code.
Of course, an even better option would be to
std::span<std::optional<std::string>> as_span() { return {&protocol, 9}; }
to the init structure... and then you could do as_span()[i] and that would be that.

Resolved

mcollina

lgtm

nodejs-github-bot · 2026-08-21T10:36:56Z

CI: https://ci.nodejs.org/job/node-test-pull-request/76110/

nodejs-github-bot · 2026-08-21T12:37:18Z

CI: https://ci.nodejs.org/job/node-test-pull-request/76136/

nodejs-github-bot · 2026-08-21T19:55:23Z

CI: https://ci.nodejs.org/job/node-test-pull-request/76180/

anonrig · 2026-08-22T17:28:10Z

@lemire can you review?

anonrig · 2026-08-18T01:50:22Z

This speeds up WHATWG URLPattern constructor, test(), and exec() without changing observable behavior.

Independent of the new URL() parse PR (#65361) and the URLSearchParams PR (#65363).

What changed

C++ only (src/node_url_pattern.cc):

String input: copy one-byte ASCII V8 strings with ValueView + memcpy into a stack buffer and pass a string_view into Ada. Drop the extra std::string copy that BufferValue/Utf8Value::ToString() used to make. Non-ASCII still goes through Utf8Value.
Object init: look up the nine interned component keys and assign fields by index. Skip UTF-8-converting each key name and the string-compare chain. Non-string members are still ignored.
Regexp provider: convert pattern/input strings with ToV8Value and use isolate->GetCurrentContext() instead of Environment::GetCurrent.
test / exec: share argument parsing so both paths keep the same WebIDL null/undefined/"null" baseURL behavior.

Ada still owns its parsed state; string_views only live for the duration of the C++ call. ValueView is destroyed before any V8 heap allocation.

Tests

test/parallel/test-urlpattern.js, test-urlpattern-types.js, test-urlpattern-invalidthis.js, and test-urlpattern-fast-path.js all pass
WPT test/wpt/test-urlpattern.js: 743 passed, 0 unexpected failures

Review follow-up

Dropped the local NewStringFromUtf8View helper in favor of ToV8Value
CopyV8StringToBuffer takes a pointer, matching project convention for mutating arguments
Author and Signed-off-by are Yagiz Nizipli (human DCO)

Assisted-by: Cursor

nodejs-github-bot · 2026-08-18T01:50:27Z

Review requested:

@nodejs/url

jasnell

AI agents are not permitted to use Signed-off-by

codecov · 2026-08-18T03:36:57Z

Codecov Report

❌ Patch coverage is 91.15044% with 10 lines in your changes missing coverage. Please review.
✅ Project coverage is 90.13%. Comparing base (f83e7df) to head (1ba9273).
⚠️ Report is 122 commits behind head on main.

Files with missing lines	Patch %	Lines
src/node_url_pattern.cc	91.15%	1 Missing and 9 partials ⚠️

Additional details and impacted files

@@ Coverage Diff @@## main #65364 +/- ##
==========================================
- Coverage 90.31% 90.13% -0.18% 
==========================================
Files 751 752 +1 Lines 249956 251861 +1905 Branches 47204 47350 +146 ==========================================
+ Hits 225745 227022 +1277 - Misses 15612 16177 +565 - Partials 8599 8662 +63

Files with missing lines	Coverage Δ
src/node_url_pattern.cc	`83.84% <91.15%> (+0.12%)`	⬆️

... and 96 files with indirect coverage changes

🚀 New features to boost your workflow:

❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

anonrig · 2026-08-18T11:40:12Z

Removed the Signed-off-by trailer from the commit(s). An AI agent cannot attest the DCO.

Avoid extra UTF-8 copies on constructor, test, and exec; convert regexp inputs with ToV8Value; and assign URLPatternInit fields by interned key instead of string-comparing each component name. Signed-off-by: Yagiz Nizipli <yagiz@nizipli.com> Assisted-by: Cursor

mertcanaltin

LGTM

lemire · 2026-08-19T20:48:12Z

- } else if (key == "baseURL") {
- init.base_url = std::string(value);
- }
+ std::optional<std::string>* fields[] = {


@anonrig I don't have my AI with me, but this seems overly complicated.
The structure is just made of...
std::optional<std::string> protocol{}; // If present, must be valid UTF-8. std::optional<std::string> username{}; // If present, must be valid UTF-8. std::optional<std::string> password{}; // If present, must be valid UTF-8. std::optional<std::string> hostname{}; // If present, must be valid UTF-8. std::optional<std::string> port{}; // If present, must be valid UTF-8. std::optional<std::string> pathname{}; // If present, must be valid UTF-8. std::optional<std::string> search{}; // If present, must be valid UTF-8. std::optional<std::string> hash{}; // If present, must be valid UTF-8. std::optional<std::string> base_url{};
The standard will lay them out one after the other without padding, I expect. So you can do (&init.protocol + index) to get the desired pointer, without materializing this fields array.
My idea won't impact the performance, probably, but it would require less code.
Of course, an even better option would be to
std::span<std::optional<std::string>> as_span() { return {&protocol, 9}; }
to the init structure... and then you could do as_span()[i] and that would be that.

Resolved

mcollina

lgtm

nodejs-github-bot · 2026-08-21T10:36:56Z

CI: https://ci.nodejs.org/job/node-test-pull-request/76110/

nodejs-github-bot · 2026-08-21T12:37:18Z

CI: https://ci.nodejs.org/job/node-test-pull-request/76136/

nodejs-github-bot · 2026-08-21T19:55:23Z

CI: https://ci.nodejs.org/job/node-test-pull-request/76180/

anonrig · 2026-08-22T17:28:10Z

@lemire can you review?

anonrig · 2026-08-18T01:50:22Z

This speeds up WHATWG URLPattern constructor, test(), and exec() without changing observable behavior.

Independent of the new URL() parse PR (#65361) and the URLSearchParams PR (#65363).

What changed

C++ only (src/node_url_pattern.cc):

String input: copy one-byte ASCII V8 strings with ValueView + memcpy into a stack buffer and pass a string_view into Ada. Drop the extra std::string copy that BufferValue/Utf8Value::ToString() used to make. Non-ASCII still goes through Utf8Value.
Object init: look up the nine interned component keys and assign fields by index. Skip UTF-8-converting each key name and the string-compare chain. Non-string members are still ignored.
Regexp provider: convert pattern/input strings with ToV8Value and use isolate->GetCurrentContext() instead of Environment::GetCurrent.
test / exec: share argument parsing so both paths keep the same WebIDL null/undefined/"null" baseURL behavior.

Ada still owns its parsed state; string_views only live for the duration of the C++ call. ValueView is destroyed before any V8 heap allocation.

Tests

test/parallel/test-urlpattern.js, test-urlpattern-types.js, test-urlpattern-invalidthis.js, and test-urlpattern-fast-path.js all pass
WPT test/wpt/test-urlpattern.js: 743 passed, 0 unexpected failures

Review follow-up

Dropped the local NewStringFromUtf8View helper in favor of ToV8Value
CopyV8StringToBuffer takes a pointer, matching project convention for mutating arguments
Author and Signed-off-by are Yagiz Nizipli (human DCO)

Assisted-by: Cursor

nodejs-github-bot · 2026-08-18T01:50:27Z

Review requested:

@nodejs/url

jasnell

AI agents are not permitted to use Signed-off-by

codecov · 2026-08-18T03:36:57Z

Codecov Report

❌ Patch coverage is 91.15044% with 10 lines in your changes missing coverage. Please review.
✅ Project coverage is 90.13%. Comparing base (f83e7df) to head (1ba9273).
⚠️ Report is 122 commits behind head on main.

Files with missing lines	Patch %	Lines
src/node_url_pattern.cc	91.15%	1 Missing and 9 partials ⚠️

Additional details and impacted files

@@ Coverage Diff @@## main #65364 +/- ##
==========================================
- Coverage 90.31% 90.13% -0.18% 
==========================================
Files 751 752 +1 Lines 249956 251861 +1905 Branches 47204 47350 +146 ==========================================
+ Hits 225745 227022 +1277 - Misses 15612 16177 +565 - Partials 8599 8662 +63

Files with missing lines	Coverage Δ
src/node_url_pattern.cc	`83.84% <91.15%> (+0.12%)`	⬆️

... and 96 files with indirect coverage changes

🚀 New features to boost your workflow:

❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

anonrig · 2026-08-18T11:40:12Z

Removed the Signed-off-by trailer from the commit(s). An AI agent cannot attest the DCO.

Avoid extra UTF-8 copies on constructor, test, and exec; convert regexp inputs with ToV8Value; and assign URLPatternInit fields by interned key instead of string-comparing each component name. Signed-off-by: Yagiz Nizipli <yagiz@nizipli.com> Assisted-by: Cursor

mertcanaltin

LGTM

lemire · 2026-08-19T20:48:12Z

- } else if (key == "baseURL") {
- init.base_url = std::string(value);
- }
+ std::optional<std::string>* fields[] = {


@anonrig I don't have my AI with me, but this seems overly complicated.
The structure is just made of...
std::optional<std::string> protocol{}; // If present, must be valid UTF-8. std::optional<std::string> username{}; // If present, must be valid UTF-8. std::optional<std::string> password{}; // If present, must be valid UTF-8. std::optional<std::string> hostname{}; // If present, must be valid UTF-8. std::optional<std::string> port{}; // If present, must be valid UTF-8. std::optional<std::string> pathname{}; // If present, must be valid UTF-8. std::optional<std::string> search{}; // If present, must be valid UTF-8. std::optional<std::string> hash{}; // If present, must be valid UTF-8. std::optional<std::string> base_url{};
The standard will lay them out one after the other without padding, I expect. So you can do (&init.protocol + index) to get the desired pointer, without materializing this fields array.
My idea won't impact the performance, probably, but it would require less code.
Of course, an even better option would be to
std::span<std::optional<std::string>> as_span() { return {&protocol, 9}; }
to the init structure... and then you could do as_span()[i] and that would be that.

Resolved

mcollina

lgtm

nodejs-github-bot · 2026-08-21T10:36:56Z

CI: https://ci.nodejs.org/job/node-test-pull-request/76110/

nodejs-github-bot · 2026-08-21T12:37:18Z

CI: https://ci.nodejs.org/job/node-test-pull-request/76136/

nodejs-github-bot · 2026-08-21T19:55:23Z

CI: https://ci.nodejs.org/job/node-test-pull-request/76180/

anonrig · 2026-08-22T17:28:10Z

@lemire can you review?

anonrig · 2026-08-18T01:50:22Z

This speeds up WHATWG URLPattern constructor, test(), and exec() without changing observable behavior.

Independent of the new URL() parse PR (#65361) and the URLSearchParams PR (#65363).

What changed

C++ only (src/node_url_pattern.cc):

String input: copy one-byte ASCII V8 strings with ValueView + memcpy into a stack buffer and pass a string_view into Ada. Drop the extra std::string copy that BufferValue/Utf8Value::ToString() used to make. Non-ASCII still goes through Utf8Value.
Object init: look up the nine interned component keys and assign fields by index. Skip UTF-8-converting each key name and the string-compare chain. Non-string members are still ignored.
Regexp provider: convert pattern/input strings with ToV8Value and use isolate->GetCurrentContext() instead of Environment::GetCurrent.
test / exec: share argument parsing so both paths keep the same WebIDL null/undefined/"null" baseURL behavior.

Ada still owns its parsed state; string_views only live for the duration of the C++ call. ValueView is destroyed before any V8 heap allocation.

Tests

test/parallel/test-urlpattern.js, test-urlpattern-types.js, test-urlpattern-invalidthis.js, and test-urlpattern-fast-path.js all pass
WPT test/wpt/test-urlpattern.js: 743 passed, 0 unexpected failures

Review follow-up

Dropped the local NewStringFromUtf8View helper in favor of ToV8Value
CopyV8StringToBuffer takes a pointer, matching project convention for mutating arguments
Author and Signed-off-by are Yagiz Nizipli (human DCO)

Assisted-by: Cursor

nodejs-github-bot · 2026-08-18T01:50:27Z

Review requested:

@nodejs/url

jasnell

AI agents are not permitted to use Signed-off-by

codecov · 2026-08-18T03:36:57Z

Codecov Report

❌ Patch coverage is 91.15044% with 10 lines in your changes missing coverage. Please review.
✅ Project coverage is 90.13%. Comparing base (f83e7df) to head (1ba9273).
⚠️ Report is 122 commits behind head on main.

Files with missing lines	Patch %	Lines
src/node_url_pattern.cc	91.15%	1 Missing and 9 partials ⚠️

Additional details and impacted files

@@ Coverage Diff @@## main #65364 +/- ##
==========================================
- Coverage 90.31% 90.13% -0.18% 
==========================================
Files 751 752 +1 Lines 249956 251861 +1905 Branches 47204 47350 +146 ==========================================
+ Hits 225745 227022 +1277 - Misses 15612 16177 +565 - Partials 8599 8662 +63

Files with missing lines	Coverage Δ
src/node_url_pattern.cc	`83.84% <91.15%> (+0.12%)`	⬆️

... and 96 files with indirect coverage changes

🚀 New features to boost your workflow:

❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

anonrig · 2026-08-18T11:40:12Z

Removed the Signed-off-by trailer from the commit(s). An AI agent cannot attest the DCO.

Avoid extra UTF-8 copies on constructor, test, and exec; convert regexp inputs with ToV8Value; and assign URLPatternInit fields by interned key instead of string-comparing each component name. Signed-off-by: Yagiz Nizipli <yagiz@nizipli.com> Assisted-by: Cursor

mertcanaltin

LGTM

lemire · 2026-08-19T20:48:12Z

- } else if (key == "baseURL") {
- init.base_url = std::string(value);
- }
+ std::optional<std::string>* fields[] = {


@anonrig I don't have my AI with me, but this seems overly complicated.
The structure is just made of...
std::optional<std::string> protocol{}; // If present, must be valid UTF-8. std::optional<std::string> username{}; // If present, must be valid UTF-8. std::optional<std::string> password{}; // If present, must be valid UTF-8. std::optional<std::string> hostname{}; // If present, must be valid UTF-8. std::optional<std::string> port{}; // If present, must be valid UTF-8. std::optional<std::string> pathname{}; // If present, must be valid UTF-8. std::optional<std::string> search{}; // If present, must be valid UTF-8. std::optional<std::string> hash{}; // If present, must be valid UTF-8. std::optional<std::string> base_url{};
The standard will lay them out one after the other without padding, I expect. So you can do (&init.protocol + index) to get the desired pointer, without materializing this fields array.
My idea won't impact the performance, probably, but it would require less code.
Of course, an even better option would be to
std::span<std::optional<std::string>> as_span() { return {&protocol, 9}; }
to the init structure... and then you could do as_span()[i] and that would be that.

Resolved

mcollina

lgtm

nodejs-github-bot · 2026-08-21T10:36:56Z

CI: https://ci.nodejs.org/job/node-test-pull-request/76110/

nodejs-github-bot · 2026-08-21T12:37:18Z

CI: https://ci.nodejs.org/job/node-test-pull-request/76136/

nodejs-github-bot · 2026-08-21T19:55:23Z

CI: https://ci.nodejs.org/job/node-test-pull-request/76180/

anonrig · 2026-08-22T17:28:10Z

@lemire can you review?

anonrig · 2026-08-18T01:50:22Z

This speeds up WHATWG URLPattern constructor, test(), and exec() without changing observable behavior.

Independent of the new URL() parse PR (#65361) and the URLSearchParams PR (#65363).

What changed

C++ only (src/node_url_pattern.cc):

String input: copy one-byte ASCII V8 strings with ValueView + memcpy into a stack buffer and pass a string_view into Ada. Drop the extra std::string copy that BufferValue/Utf8Value::ToString() used to make. Non-ASCII still goes through Utf8Value.
Object init: look up the nine interned component keys and assign fields by index. Skip UTF-8-converting each key name and the string-compare chain. Non-string members are still ignored.
Regexp provider: convert pattern/input strings with ToV8Value and use isolate->GetCurrentContext() instead of Environment::GetCurrent.
test / exec: share argument parsing so both paths keep the same WebIDL null/undefined/"null" baseURL behavior.

Ada still owns its parsed state; string_views only live for the duration of the C++ call. ValueView is destroyed before any V8 heap allocation.

Tests

test/parallel/test-urlpattern.js, test-urlpattern-types.js, test-urlpattern-invalidthis.js, and test-urlpattern-fast-path.js all pass
WPT test/wpt/test-urlpattern.js: 743 passed, 0 unexpected failures

Review follow-up

Dropped the local NewStringFromUtf8View helper in favor of ToV8Value
CopyV8StringToBuffer takes a pointer, matching project convention for mutating arguments
Author and Signed-off-by are Yagiz Nizipli (human DCO)

Assisted-by: Cursor

nodejs-github-bot · 2026-08-18T01:50:27Z

Review requested:

@nodejs/url

jasnell

AI agents are not permitted to use Signed-off-by

codecov · 2026-08-18T03:36:57Z

Codecov Report

❌ Patch coverage is 91.15044% with 10 lines in your changes missing coverage. Please review.
✅ Project coverage is 90.13%. Comparing base (f83e7df) to head (1ba9273).
⚠️ Report is 122 commits behind head on main.

Files with missing lines	Patch %	Lines
src/node_url_pattern.cc	91.15%	1 Missing and 9 partials ⚠️

Additional details and impacted files

@@ Coverage Diff @@## main #65364 +/- ##
==========================================
- Coverage 90.31% 90.13% -0.18% 
==========================================
Files 751 752 +1 Lines 249956 251861 +1905 Branches 47204 47350 +146 ==========================================
+ Hits 225745 227022 +1277 - Misses 15612 16177 +565 - Partials 8599 8662 +63

Files with missing lines	Coverage Δ
src/node_url_pattern.cc	`83.84% <91.15%> (+0.12%)`	⬆️

... and 96 files with indirect coverage changes

🚀 New features to boost your workflow:

❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

anonrig · 2026-08-18T11:40:12Z

Removed the Signed-off-by trailer from the commit(s). An AI agent cannot attest the DCO.

Avoid extra UTF-8 copies on constructor, test, and exec; convert regexp inputs with ToV8Value; and assign URLPatternInit fields by interned key instead of string-comparing each component name. Signed-off-by: Yagiz Nizipli <yagiz@nizipli.com> Assisted-by: Cursor

mertcanaltin

LGTM

lemire · 2026-08-19T20:48:12Z

- } else if (key == "baseURL") {
- init.base_url = std::string(value);
- }
+ std::optional<std::string>* fields[] = {


@anonrig I don't have my AI with me, but this seems overly complicated.
The structure is just made of...
std::optional<std::string> protocol{}; // If present, must be valid UTF-8. std::optional<std::string> username{}; // If present, must be valid UTF-8. std::optional<std::string> password{}; // If present, must be valid UTF-8. std::optional<std::string> hostname{}; // If present, must be valid UTF-8. std::optional<std::string> port{}; // If present, must be valid UTF-8. std::optional<std::string> pathname{}; // If present, must be valid UTF-8. std::optional<std::string> search{}; // If present, must be valid UTF-8. std::optional<std::string> hash{}; // If present, must be valid UTF-8. std::optional<std::string> base_url{};
The standard will lay them out one after the other without padding, I expect. So you can do (&init.protocol + index) to get the desired pointer, without materializing this fields array.
My idea won't impact the performance, probably, but it would require less code.
Of course, an even better option would be to
std::span<std::optional<std::string>> as_span() { return {&protocol, 9}; }
to the init structure... and then you could do as_span()[i] and that would be that.

Resolved

mcollina

lgtm

nodejs-github-bot · 2026-08-21T10:36:56Z

CI: https://ci.nodejs.org/job/node-test-pull-request/76110/

nodejs-github-bot · 2026-08-21T12:37:18Z

CI: https://ci.nodejs.org/job/node-test-pull-request/76136/

nodejs-github-bot · 2026-08-21T19:55:23Z

CI: https://ci.nodejs.org/job/node-test-pull-request/76180/

anonrig · 2026-08-22T17:28:10Z

@lemire can you review?

anonrig · 2026-08-18T01:50:22Z

This speeds up WHATWG URLPattern constructor, test(), and exec() without changing observable behavior.

Independent of the new URL() parse PR (#65361) and the URLSearchParams PR (#65363).

What changed

C++ only (src/node_url_pattern.cc):

String input: copy one-byte ASCII V8 strings with ValueView + memcpy into a stack buffer and pass a string_view into Ada. Drop the extra std::string copy that BufferValue/Utf8Value::ToString() used to make. Non-ASCII still goes through Utf8Value.
Object init: look up the nine interned component keys and assign fields by index. Skip UTF-8-converting each key name and the string-compare chain. Non-string members are still ignored.
Regexp provider: convert pattern/input strings with ToV8Value and use isolate->GetCurrentContext() instead of Environment::GetCurrent.
test / exec: share argument parsing so both paths keep the same WebIDL null/undefined/"null" baseURL behavior.

Ada still owns its parsed state; string_views only live for the duration of the C++ call. ValueView is destroyed before any V8 heap allocation.

Tests

test/parallel/test-urlpattern.js, test-urlpattern-types.js, test-urlpattern-invalidthis.js, and test-urlpattern-fast-path.js all pass
WPT test/wpt/test-urlpattern.js: 743 passed, 0 unexpected failures

Review follow-up

Dropped the local NewStringFromUtf8View helper in favor of ToV8Value
CopyV8StringToBuffer takes a pointer, matching project convention for mutating arguments
Author and Signed-off-by are Yagiz Nizipli (human DCO)

Assisted-by: Cursor

nodejs-github-bot · 2026-08-18T01:50:27Z

Review requested:

@nodejs/url

jasnell

AI agents are not permitted to use Signed-off-by

codecov · 2026-08-18T03:36:57Z

Codecov Report

❌ Patch coverage is 91.15044% with 10 lines in your changes missing coverage. Please review.
✅ Project coverage is 90.13%. Comparing base (f83e7df) to head (1ba9273).
⚠️ Report is 122 commits behind head on main.

Files with missing lines	Patch %	Lines
src/node_url_pattern.cc	91.15%	1 Missing and 9 partials ⚠️

Additional details and impacted files

@@ Coverage Diff @@## main #65364 +/- ##
==========================================
- Coverage 90.31% 90.13% -0.18% 
==========================================
Files 751 752 +1 Lines 249956 251861 +1905 Branches 47204 47350 +146 ==========================================
+ Hits 225745 227022 +1277 - Misses 15612 16177 +565 - Partials 8599 8662 +63

Files with missing lines	Coverage Δ
src/node_url_pattern.cc	`83.84% <91.15%> (+0.12%)`	⬆️

... and 96 files with indirect coverage changes

🚀 New features to boost your workflow:

❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

anonrig · 2026-08-18T11:40:12Z

Removed the Signed-off-by trailer from the commit(s). An AI agent cannot attest the DCO.

Avoid extra UTF-8 copies on constructor, test, and exec; convert regexp inputs with ToV8Value; and assign URLPatternInit fields by interned key instead of string-comparing each component name. Signed-off-by: Yagiz Nizipli <yagiz@nizipli.com> Assisted-by: Cursor

mertcanaltin

LGTM

lemire · 2026-08-19T20:48:12Z

- } else if (key == "baseURL") {
- init.base_url = std::string(value);
- }
+ std::optional<std::string>* fields[] = {


@anonrig I don't have my AI with me, but this seems overly complicated.
The structure is just made of...
std::optional<std::string> protocol{}; // If present, must be valid UTF-8. std::optional<std::string> username{}; // If present, must be valid UTF-8. std::optional<std::string> password{}; // If present, must be valid UTF-8. std::optional<std::string> hostname{}; // If present, must be valid UTF-8. std::optional<std::string> port{}; // If present, must be valid UTF-8. std::optional<std::string> pathname{}; // If present, must be valid UTF-8. std::optional<std::string> search{}; // If present, must be valid UTF-8. std::optional<std::string> hash{}; // If present, must be valid UTF-8. std::optional<std::string> base_url{};
The standard will lay them out one after the other without padding, I expect. So you can do (&init.protocol + index) to get the desired pointer, without materializing this fields array.
My idea won't impact the performance, probably, but it would require less code.
Of course, an even better option would be to
std::span<std::optional<std::string>> as_span() { return {&protocol, 9}; }
to the init structure... and then you could do as_span()[i] and that would be that.

Resolved

mcollina

lgtm

nodejs-github-bot · 2026-08-21T10:36:56Z

CI: https://ci.nodejs.org/job/node-test-pull-request/76110/

nodejs-github-bot · 2026-08-21T12:37:18Z

CI: https://ci.nodejs.org/job/node-test-pull-request/76136/

nodejs-github-bot · 2026-08-21T19:55:23Z

CI: https://ci.nodejs.org/job/node-test-pull-request/76180/

anonrig · 2026-08-22T17:28:10Z

@lemire can you review?

nodejs-github-bot added c++ Issues and PRs that require attention from people who are familiar with C++. needs-ci PRs that need a full CI run. whatwg-url Issues and PRs related to the WHATWG URL implementation. labels Aug 18, 2026

jasnell previously requested changes Aug 18, 2026
View reviewed changes

jasnell reviewed Aug 18, 2026
View reviewed changes

Comment threadsrc/node_url_pattern.cc Outdated

jasnell reviewed Aug 18, 2026
View reviewed changes

Comment threadsrc/node_url_pattern.cc Outdated

cursorBotforce-pushed the cursor/urlpattern-performance-603e branch from 351c14d to 52a2a71Compare August 18, 2026 11:39

cursorBotforce-pushed the cursor/urlpattern-performance-603e branch from 52a2a71 to a2e81deCompare August 18, 2026 12:49

url: speed up URLPattern
1ba9273
Avoid extra UTF-8 copies on constructor, test, and exec; convert regexp inputs with ToV8Value; and assign URLPatternInit fields by interned key instead of string-comparing each component name. Signed-off-by: Yagiz Nizipli <yagiz@nizipli.com> Assisted-by: Cursor

cursorBotforce-pushed the cursor/urlpattern-performance-603e branch 5 times, most recently from ed512d0 to 1ba9273Compare August 18, 2026 18:43

mertcanaltin approved these changes Aug 19, 2026
View reviewed changes

lemire reviewed Aug 19, 2026
View reviewed changes

anonrig requested review from jasnell, joyeecheung and lemire August 20, 2026 18:56

mcollina approved these changes Aug 21, 2026
View reviewed changes

mcollina added the request-ci Add this label to start a Jenkins CI on a PR. label Aug 21, 2026

github-actionsBot removed the request-ci Add this label to start a Jenkins CI on a PR. label Aug 21, 2026

anonrig requested a review from mcollina August 23, 2026 03:15

panva approved these changes Aug 25, 2026
View reviewed changes