Make EventTarget compatible with the existing implementation of ReadOnlyNode #48427

rubennorte · 2025-01-02T10:22:44Z

Summary:
Changelog: [internal]

The ReactNativeElement class was refactored for performance reasons, and the current implementation does NOT call super(), and it inlines the parent constructor instead.

When it eventually extends EventTarget, things won't work as expected because the existing EventTarget implementation has constructor dependencies.

This refactors the current implementation of EventTarget to eliminate those constructor side-effects, and eliminates the constructor altogether.

This breaks encapsulation, but it has some positive side-effects on performance:

Creating EventTarget instances is faster because it has no constructor logic.
Improves memory by not creating maps to hold the event listeners if no event listeners are ever added to the target (which is very common).
Improves the overall runtime performance of the methods in the class by migrating away from private methods (which are known to be slow on the babel transpiled version we're currently using).

Differential Revision: D67758408

facebook-github-bot · 2025-01-02T10:23:12Z

This pull request was exported from Phabricator. Differential Revision: D67758408

…n of ReadOnlyNode (facebook#48427) Summary: Changelog: [internal] The `ReactNativeElement` class was refactored for performance reasons, and the current implementation does **NOT** call `super()`, and it inlines the parent constructor instead. When it eventually extends `EventTarget`, things won't work as expected because the existing `EventTarget` implementation has constructor dependencies. This refactors the current implementation of `EventTarget` to eliminate those constructor side-effects, and eliminates the constructor altogether. This breaks encapsulation, but it has some positive side-effects on performance: 1. Creating `EventTarget` instances is faster because it has no constructor logic. 2. Improves memory by not creating maps to hold the event listeners if no event listeners are ever added to the target (which is very common). 3. Improves the overall runtime performance of the methods in the class by migrating away from private methods (which are known to be slow on the babel transpiled version we're currently using). Extra: it also simplifies making window/the global scope implement the EventTarget interface :) Differential Revision: D67758408

…nlyNode (facebook#48427) Summary: Changelog: [internal] The `ReactNativeElement` class was refactored for performance reasons, and the current implementation does **NOT** call `super()`, and it inlines the parent constructor instead. When it eventually extends `EventTarget`, things won't work as expected because the existing `EventTarget` implementation has constructor dependencies. This refactors the current implementation of `EventTarget` to eliminate those constructor side-effects, and eliminates the constructor altogether. This breaks encapsulation, but it has some positive side-effects on performance: 1. Creating `EventTarget` instances is faster because it has no constructor logic. 2. Improves memory by not creating maps to hold the event listeners if no event listeners are ever added to the target (which is very common). 3. Improves the overall runtime performance of the methods in the class by migrating away from private methods (which are known to be slow on the babel transpiled version we're currently using). Extra: it also simplifies making window/the global scope implement the EventTarget interface :) Differential Revision: D67758408

…n of ReadOnlyNode (facebook#48427) Summary: Changelog: [internal] The `ReactNativeElement` class was refactored for performance reasons, and the current implementation does **NOT** call `super()`, and it inlines the parent constructor instead. When it eventually extends `EventTarget`, things won't work as expected because the existing `EventTarget` implementation has constructor dependencies. This refactors the current implementation of `EventTarget` to eliminate those constructor side-effects, and eliminates the constructor altogether. This breaks encapsulation, but it has some positive side-effects on performance: 1. Creating `EventTarget` instances is faster because it has no constructor logic. 2. Improves memory by not creating maps to hold the event listeners if no event listeners are ever added to the target (which is very common). 3. Improves the overall runtime performance of the methods in the class by migrating away from private methods (which are known to be slow on the babel transpiled version we're currently using). Extra: it also simplifies making window/the global scope implement the EventTarget interface :) Differential Revision: D67758408

…nlyNode (facebook#48427) Summary: Changelog: [internal] The `ReactNativeElement` class was refactored for performance reasons, and the current implementation does **NOT** call `super()`, and it inlines the parent constructor instead. When it eventually extends `EventTarget`, things won't work as expected because the existing `EventTarget` implementation has constructor dependencies. This refactors the current implementation of `EventTarget` to eliminate those constructor side-effects, and eliminates the constructor altogether. This breaks encapsulation, but it has some positive side-effects on performance: 1. Creating `EventTarget` instances is faster because it has no constructor logic. 2. Improves memory by not creating maps to hold the event listeners if no event listeners are ever added to the target (which is very common). 3. Improves the overall runtime performance of the methods in the class by migrating away from private methods (which are known to be slow on the babel transpiled version we're currently using). Extra: it also simplifies making window/the global scope implement the EventTarget interface :) Differential Revision: D67758408

facebook-github-bot · 2025-01-05T18:42:31Z

This pull request was exported from Phabricator. Differential Revision: D67758408

…nlyNode (facebook#48427) Summary: Changelog: [internal] The `ReactNativeElement` class was refactored for performance reasons, and the current implementation does **NOT** call `super()`, and it inlines the parent constructor instead. When it eventually extends `EventTarget`, things won't work as expected because the existing `EventTarget` implementation has constructor dependencies. This refactors the current implementation of `EventTarget` to eliminate those constructor side-effects, and eliminates the constructor altogether. This breaks encapsulation, but it has some positive side-effects on performance: 1. Creating `EventTarget` instances is faster because it has no constructor logic. 2. Improves memory by not creating maps to hold the event listeners if no event listeners are ever added to the target (which is very common). 3. Improves the overall runtime performance of the methods in the class by migrating away from private methods (which are known to be slow on the babel transpiled version we're currently using). Extra: it also simplifies making window/the global scope implement the EventTarget interface :) Differential Revision: D67758408

…n of ReadOnlyNode (facebook#48427) Summary: Changelog: [internal] The `ReactNativeElement` class was refactored for performance reasons, and the current implementation does **NOT** call `super()`, and it inlines the parent constructor instead. When it eventually extends `EventTarget`, things won't work as expected because the existing `EventTarget` implementation has constructor dependencies. This refactors the current implementation of `EventTarget` to eliminate those constructor side-effects, and eliminates the constructor altogether. This breaks encapsulation, but it has some positive side-effects on performance: 1. Creating `EventTarget` instances is faster because it has no constructor logic. 2. Improves memory by not creating maps to hold the event listeners if no event listeners are ever added to the target (which is very common). 3. Improves the overall runtime performance of the methods in the class by migrating away from private methods (which are known to be slow on the babel transpiled version we're currently using). Extra: it also simplifies making window/the global scope implement the EventTarget interface :) Differential Revision: D67758408

facebook-github-bot · 2025-01-06T15:21:07Z

This pull request was exported from Phabricator. Differential Revision: D67758408

…n of ReadOnlyNode (facebook#48427) Summary: Changelog: [internal] The `ReactNativeElement` class was refactored for performance reasons, and the current implementation does **NOT** call `super()`, and it inlines the parent constructor instead. When it eventually extends `EventTarget`, things won't work as expected because the existing `EventTarget` implementation has constructor dependencies. This refactors the current implementation of `EventTarget` to eliminate those constructor side-effects, and eliminates the constructor altogether. This breaks encapsulation, but it has some positive side-effects on performance: 1. Creating `EventTarget` instances is faster because it has no constructor logic. 2. Improves memory by not creating maps to hold the event listeners if no event listeners are ever added to the target (which is very common). 3. Improves the overall runtime performance of the methods in the class by migrating away from private methods (which are known to be slow on the babel transpiled version we're currently using). Extra: it also simplifies making window/the global scope implement the EventTarget interface :) Differential Revision: D67758408

…nlyNode (facebook#48427) Summary: Changelog: [internal] The `ReactNativeElement` class was refactored for performance reasons, and the current implementation does **NOT** call `super()`, and it inlines the parent constructor instead. When it eventually extends `EventTarget`, things won't work as expected because the existing `EventTarget` implementation has constructor dependencies. This refactors the current implementation of `EventTarget` to eliminate those constructor side-effects, and eliminates the constructor altogether. This breaks encapsulation, but it has some positive side-effects on performance: 1. Creating `EventTarget` instances is faster because it has no constructor logic. 2. Improves memory by not creating maps to hold the event listeners if no event listeners are ever added to the target (which is very common). 3. Improves the overall runtime performance of the methods in the class by migrating away from private methods (which are known to be slow on the babel transpiled version we're currently using). Extra: it also simplifies making window/the global scope implement the EventTarget interface :) Differential Revision: D67758408

facebook-github-bot · 2025-01-06T16:54:31Z

This pull request was exported from Phabricator. Differential Revision: D67758408

…n of ReadOnlyNode (facebook#48427) Summary: Changelog: [internal] The `ReactNativeElement` class was refactored for performance reasons, and the current implementation does **NOT** call `super()`, and it inlines the parent constructor instead. When it eventually extends `EventTarget`, things won't work as expected because the existing `EventTarget` implementation has constructor dependencies. This refactors the current implementation of `EventTarget` to eliminate those constructor side-effects, and eliminates the constructor altogether. This breaks encapsulation, but it has some positive side-effects on performance: 1. Creating `EventTarget` instances is faster because it has no constructor logic. 2. Improves memory by not creating maps to hold the event listeners if no event listeners are ever added to the target (which is very common). 3. Improves the overall runtime performance of the methods in the class by migrating away from private methods (which are known to be slow on the babel transpiled version we're currently using). Extra: it also simplifies making window/the global scope implement the EventTarget interface :) Differential Revision: D67758408

…nlyNode (facebook#48427) Summary: Changelog: [internal] The `ReactNativeElement` class was refactored for performance reasons, and the current implementation does **NOT** call `super()`, and it inlines the parent constructor instead. When it eventually extends `EventTarget`, things won't work as expected because the existing `EventTarget` implementation has constructor dependencies. This refactors the current implementation of `EventTarget` to eliminate those constructor side-effects, and eliminates the constructor altogether. This breaks encapsulation, but it has some positive side-effects on performance: 1. Creating `EventTarget` instances is faster because it has no constructor logic. 2. Improves memory by not creating maps to hold the event listeners if no event listeners are ever added to the target (which is very common). 3. Improves the overall runtime performance of the methods in the class by migrating away from private methods (which are known to be slow on the babel transpiled version we're currently using). Extra: it also simplifies making window/the global scope implement the EventTarget interface :) Differential Revision: D67758408

facebook-github-bot · 2025-01-06T17:27:01Z

This pull request was exported from Phabricator. Differential Revision: D67758408

…nlyNode (facebook#48427) Summary: Changelog: [internal] The `ReactNativeElement` class was refactored for performance reasons, and the current implementation does **NOT** call `super()`, and it inlines the parent constructor instead. When it eventually extends `EventTarget`, things won't work as expected because the existing `EventTarget` implementation has constructor dependencies. This refactors the current implementation of `EventTarget` to eliminate those constructor side-effects, and eliminates the constructor altogether. This breaks encapsulation, but it has some positive side-effects on performance: 1. Creating `EventTarget` instances is faster because it has no constructor logic. 2. Improves memory by not creating maps to hold the event listeners if no event listeners are ever added to the target (which is very common). 3. Improves the overall runtime performance of the methods in the class by migrating away from private methods (which are known to be slow on the babel transpiled version we're currently using). Extra: it also simplifies making window/the global scope implement the EventTarget interface :) Differential Revision: D67758408

facebook-github-bot · 2025-01-07T10:33:25Z

This pull request was exported from Phabricator. Differential Revision: D67758408

…n of ReadOnlyNode (facebook#48427) Summary: Changelog: [internal] The `ReactNativeElement` class was refactored for performance reasons, and the current implementation does **NOT** call `super()`, and it inlines the parent constructor instead. When it eventually extends `EventTarget`, things won't work as expected because the existing `EventTarget` implementation has constructor dependencies. This refactors the current implementation of `EventTarget` to eliminate those constructor side-effects, and eliminates the constructor altogether. This breaks encapsulation, but it has some positive side-effects on performance: 1. Creating `EventTarget` instances is faster because it has no constructor logic. 2. Improves memory by not creating maps to hold the event listeners if no event listeners are ever added to the target (which is very common). 3. Improves the overall runtime performance of the methods in the class by migrating away from private methods (which are known to be slow on the babel transpiled version we're currently using). Extra: it also simplifies making window/the global scope implement the EventTarget interface :) ## Benchmark results Before: | Latency average (ns) | Latency median (ns) | Samples | Task name | Throughput average (ops/s) | Throughput median (ops/s) | | ---|--- |--- |--- |---|---| | 8234.22 ± 0.27% | 8132.00 | 121445 | dispatchEvent, no bubbling, no listeners | 122323 ± 0.02% | 122971 | | 9001.22 ± 0.41% | 8883.00 | 111097 | dispatchEvent, no bubbling, single listener | 111981 ± 0.02% | 112575 | | 51777.94 ± 0.58% | 51247.00 | 19314 | dispatchEvent, no bubbling, multiple listeners | 19393 ± 0.04% | 19513 | | 8256.65 ± 0.29% | 8152.00 | 121115 | dispatchEvent, bubbling, no listeners | 122031 ± 0.02% | 122669 | | 9064.32 ± 0.44% | 8933.00 | 110323 | dispatchEvent, bubbling, single listener per target | 111265 ± 0.02% | 111944 | | 51879.66 ± 0.27% | 51447.00 | 19276 | dispatchEvent, bubbling, multiple listeners per target | 19325 ± 0.04% | 19437 | After: | Latency average (ns) | Latency median (ns) | Samples | Task name | Throughput average (ops/s) | Throughput median (ops/s)| | ---------------------|---------------------|---------|--------------------------------------------------------|----------------------------|--------------------------| | 5664.62 ± 0.50% | 5588.00 | 176535 | dispatchEvent, no bubbling, no listeners | 178219 ± 0.02% | 178955 | | 7232.86 ± 0.50% | 7131.00 | 138258 | dispatchEvent, no bubbling, single listener | 139540 ± 0.02% | 140233 | | 50957.51 ± 0.71% | 50336.00 | 19625 | dispatchEvent, no bubbling, multiple listeners | 19751 ± 0.04% | 19866 | | 5692.36 ± 0.50% | 5618.00 | 175675 | dispatchEvent, bubbling, no listeners | 177315 ± 0.02% | 177999 | | 7277.82 ± 0.38% | 7181.00 | 137404 | dispatchEvent, bubbling, single listener per target | 138560 ± 0.02% | 139256 | | 50493.64 ± 0.28% | 50105.00 | 19805 | dispatchEvent, bubbling, multiple listeners per target | 19855 ± 0.04% | 19958 | Reviewed By: yungsters Differential Revision: D67758408

facebook-github-bot · 2025-01-27T10:29:53Z

This pull request was exported from Phabricator. Differential Revision: D67758408

…n of ReadOnlyNode (facebook#48427) Summary: Changelog: [internal] The `ReactNativeElement` class was refactored for performance reasons, and the current implementation does **NOT** call `super()`, and it inlines the parent constructor instead. When it eventually extends `EventTarget`, things won't work as expected because the existing `EventTarget` implementation has constructor dependencies. This refactors the current implementation of `EventTarget` to eliminate those constructor side-effects, and eliminates the constructor altogether. This breaks encapsulation, but it has some positive side-effects on performance: 1. Creating `EventTarget` instances is faster because it has no constructor logic. 2. Improves memory by not creating maps to hold the event listeners if no event listeners are ever added to the target (which is very common). 3. Improves the overall runtime performance of the methods in the class by migrating away from private methods (which are known to be slow on the babel transpiled version we're currently using). Extra: it also simplifies making window/the global scope implement the EventTarget interface :) ## Benchmark results Before: | Latency average (ns) | Latency median (ns) | Samples | Task name | Throughput average (ops/s) | Throughput median (ops/s) | | ---|--- |--- |--- |---|---| | 8234.22 ± 0.27% | 8132.00 | 121445 | dispatchEvent, no bubbling, no listeners | 122323 ± 0.02% | 122971 | | 9001.22 ± 0.41% | 8883.00 | 111097 | dispatchEvent, no bubbling, single listener | 111981 ± 0.02% | 112575 | | 51777.94 ± 0.58% | 51247.00 | 19314 | dispatchEvent, no bubbling, multiple listeners | 19393 ± 0.04% | 19513 | | 8256.65 ± 0.29% | 8152.00 | 121115 | dispatchEvent, bubbling, no listeners | 122031 ± 0.02% | 122669 | | 9064.32 ± 0.44% | 8933.00 | 110323 | dispatchEvent, bubbling, single listener per target | 111265 ± 0.02% | 111944 | | 51879.66 ± 0.27% | 51447.00 | 19276 | dispatchEvent, bubbling, multiple listeners per target | 19325 ± 0.04% | 19437 | After: | Latency average (ns) | Latency median (ns) | Samples | Task name | Throughput average (ops/s) | Throughput median (ops/s)| | ---------------------|---------------------|---------|--------------------------------------------------------|----------------------------|--------------------------| | 5664.62 ± 0.50% | 5588.00 | 176535 | dispatchEvent, no bubbling, no listeners | 178219 ± 0.02% | 178955 | | 7232.86 ± 0.50% | 7131.00 | 138258 | dispatchEvent, no bubbling, single listener | 139540 ± 0.02% | 140233 | | 50957.51 ± 0.71% | 50336.00 | 19625 | dispatchEvent, no bubbling, multiple listeners | 19751 ± 0.04% | 19866 | | 5692.36 ± 0.50% | 5618.00 | 175675 | dispatchEvent, bubbling, no listeners | 177315 ± 0.02% | 177999 | | 7277.82 ± 0.38% | 7181.00 | 137404 | dispatchEvent, bubbling, single listener per target | 138560 ± 0.02% | 139256 | | 50493.64 ± 0.28% | 50105.00 | 19805 | dispatchEvent, bubbling, multiple listeners per target | 19855 ± 0.04% | 19958 | Reviewed By: yungsters Differential Revision: D67758408

…nlyNode (facebook#48427) Summary: Changelog: [internal] The `ReactNativeElement` class was refactored for performance reasons, and the current implementation does **NOT** call `super()`, and it inlines the parent constructor instead. When it eventually extends `EventTarget`, things won't work as expected because the existing `EventTarget` implementation has constructor dependencies. This refactors the current implementation of `EventTarget` to eliminate those constructor side-effects, and eliminates the constructor altogether. This breaks encapsulation, but it has some positive side-effects on performance: 1. Creating `EventTarget` instances is faster because it has no constructor logic. 2. Improves memory by not creating maps to hold the event listeners if no event listeners are ever added to the target (which is very common). 3. Improves the overall runtime performance of the methods in the class by migrating away from private methods (which are known to be slow on the babel transpiled version we're currently using). Extra: it also simplifies making window/the global scope implement the EventTarget interface :) ## Benchmark results Before: | Latency average (ns) | Latency median (ns) | Samples | Task name | Throughput average (ops/s) | Throughput median (ops/s) | | ---|--- |--- |--- |---|---| | 8234.22 ± 0.27% | 8132.00 | 121445 | dispatchEvent, no bubbling, no listeners | 122323 ± 0.02% | 122971 | | 9001.22 ± 0.41% | 8883.00 | 111097 | dispatchEvent, no bubbling, single listener | 111981 ± 0.02% | 112575 | | 51777.94 ± 0.58% | 51247.00 | 19314 | dispatchEvent, no bubbling, multiple listeners | 19393 ± 0.04% | 19513 | | 8256.65 ± 0.29% | 8152.00 | 121115 | dispatchEvent, bubbling, no listeners | 122031 ± 0.02% | 122669 | | 9064.32 ± 0.44% | 8933.00 | 110323 | dispatchEvent, bubbling, single listener per target | 111265 ± 0.02% | 111944 | | 51879.66 ± 0.27% | 51447.00 | 19276 | dispatchEvent, bubbling, multiple listeners per target | 19325 ± 0.04% | 19437 | After: | Latency average (ns) | Latency median (ns) | Samples | Task name | Throughput average (ops/s) | Throughput median (ops/s)| | ---------------------|---------------------|---------|--------------------------------------------------------|----------------------------|--------------------------| | 5664.62 ± 0.50% | 5588.00 | 176535 | dispatchEvent, no bubbling, no listeners | 178219 ± 0.02% | 178955 | | 7232.86 ± 0.50% | 7131.00 | 138258 | dispatchEvent, no bubbling, single listener | 139540 ± 0.02% | 140233 | | 50957.51 ± 0.71% | 50336.00 | 19625 | dispatchEvent, no bubbling, multiple listeners | 19751 ± 0.04% | 19866 | | 5692.36 ± 0.50% | 5618.00 | 175675 | dispatchEvent, bubbling, no listeners | 177315 ± 0.02% | 177999 | | 7277.82 ± 0.38% | 7181.00 | 137404 | dispatchEvent, bubbling, single listener per target | 138560 ± 0.02% | 139256 | | 50493.64 ± 0.28% | 50105.00 | 19805 | dispatchEvent, bubbling, multiple listeners per target | 19855 ± 0.04% | 19958 | Reviewed By: yungsters Differential Revision: D67758408

…n of ReadOnlyNode (facebook#48427) Summary: Changelog: [internal] The `ReactNativeElement` class was refactored for performance reasons, and the current implementation does **NOT** call `super()`, and it inlines the parent constructor instead. When it eventually extends `EventTarget`, things won't work as expected because the existing `EventTarget` implementation has constructor dependencies. This refactors the current implementation of `EventTarget` to eliminate those constructor side-effects, and eliminates the constructor altogether. This breaks encapsulation, but it has some positive side-effects on performance: 1. Creating `EventTarget` instances is faster because it has no constructor logic. 2. Improves memory by not creating maps to hold the event listeners if no event listeners are ever added to the target (which is very common). 3. Improves the overall runtime performance of the methods in the class by migrating away from private methods (which are known to be slow on the babel transpiled version we're currently using). Extra: it also simplifies making window/the global scope implement the EventTarget interface :) ## Benchmark results Before: | Latency average (ns) | Latency median (ns) | Samples | Task name | Throughput average (ops/s) | Throughput median (ops/s) | | ---|--- |--- |--- |---|---| | 8234.22 ± 0.27% | 8132.00 | 121445 | dispatchEvent, no bubbling, no listeners | 122323 ± 0.02% | 122971 | | 9001.22 ± 0.41% | 8883.00 | 111097 | dispatchEvent, no bubbling, single listener | 111981 ± 0.02% | 112575 | | 51777.94 ± 0.58% | 51247.00 | 19314 | dispatchEvent, no bubbling, multiple listeners | 19393 ± 0.04% | 19513 | | 8256.65 ± 0.29% | 8152.00 | 121115 | dispatchEvent, bubbling, no listeners | 122031 ± 0.02% | 122669 | | 9064.32 ± 0.44% | 8933.00 | 110323 | dispatchEvent, bubbling, single listener per target | 111265 ± 0.02% | 111944 | | 51879.66 ± 0.27% | 51447.00 | 19276 | dispatchEvent, bubbling, multiple listeners per target | 19325 ± 0.04% | 19437 | After: | Latency average (ns) | Latency median (ns) | Samples | Task name | Throughput average (ops/s) | Throughput median (ops/s)| | ---------------------|---------------------|---------|--------------------------------------------------------|----------------------------|--------------------------| | 5664.62 ± 0.50% | 5588.00 | 176535 | dispatchEvent, no bubbling, no listeners | 178219 ± 0.02% | 178955 | | 7232.86 ± 0.50% | 7131.00 | 138258 | dispatchEvent, no bubbling, single listener | 139540 ± 0.02% | 140233 | | 50957.51 ± 0.71% | 50336.00 | 19625 | dispatchEvent, no bubbling, multiple listeners | 19751 ± 0.04% | 19866 | | 5692.36 ± 0.50% | 5618.00 | 175675 | dispatchEvent, bubbling, no listeners | 177315 ± 0.02% | 177999 | | 7277.82 ± 0.38% | 7181.00 | 137404 | dispatchEvent, bubbling, single listener per target | 138560 ± 0.02% | 139256 | | 50493.64 ± 0.28% | 50105.00 | 19805 | dispatchEvent, bubbling, multiple listeners per target | 19855 ± 0.04% | 19958 | Reviewed By: yungsters Differential Revision: D67758408

…nlyNode (facebook#48427) Summary: Changelog: [internal] The `ReactNativeElement` class was refactored for performance reasons, and the current implementation does **NOT** call `super()`, and it inlines the parent constructor instead. When it eventually extends `EventTarget`, things won't work as expected because the existing `EventTarget` implementation has constructor dependencies. This refactors the current implementation of `EventTarget` to eliminate those constructor side-effects, and eliminates the constructor altogether. This breaks encapsulation, but it has some positive side-effects on performance: 1. Creating `EventTarget` instances is faster because it has no constructor logic. 2. Improves memory by not creating maps to hold the event listeners if no event listeners are ever added to the target (which is very common). 3. Improves the overall runtime performance of the methods in the class by migrating away from private methods (which are known to be slow on the babel transpiled version we're currently using). Extra: it also simplifies making window/the global scope implement the EventTarget interface :) ## Benchmark results Before: | Latency average (ns) | Latency median (ns) | Samples | Task name | Throughput average (ops/s) | Throughput median (ops/s) | | ---|--- |--- |--- |---|---| | 8234.22 ± 0.27% | 8132.00 | 121445 | dispatchEvent, no bubbling, no listeners | 122323 ± 0.02% | 122971 | | 9001.22 ± 0.41% | 8883.00 | 111097 | dispatchEvent, no bubbling, single listener | 111981 ± 0.02% | 112575 | | 51777.94 ± 0.58% | 51247.00 | 19314 | dispatchEvent, no bubbling, multiple listeners | 19393 ± 0.04% | 19513 | | 8256.65 ± 0.29% | 8152.00 | 121115 | dispatchEvent, bubbling, no listeners | 122031 ± 0.02% | 122669 | | 9064.32 ± 0.44% | 8933.00 | 110323 | dispatchEvent, bubbling, single listener per target | 111265 ± 0.02% | 111944 | | 51879.66 ± 0.27% | 51447.00 | 19276 | dispatchEvent, bubbling, multiple listeners per target | 19325 ± 0.04% | 19437 | After: | Latency average (ns) | Latency median (ns) | Samples | Task name | Throughput average (ops/s) | Throughput median (ops/s)| | ---------------------|---------------------|---------|--------------------------------------------------------|----------------------------|--------------------------| | 5664.62 ± 0.50% | 5588.00 | 176535 | dispatchEvent, no bubbling, no listeners | 178219 ± 0.02% | 178955 | | 7232.86 ± 0.50% | 7131.00 | 138258 | dispatchEvent, no bubbling, single listener | 139540 ± 0.02% | 140233 | | 50957.51 ± 0.71% | 50336.00 | 19625 | dispatchEvent, no bubbling, multiple listeners | 19751 ± 0.04% | 19866 | | 5692.36 ± 0.50% | 5618.00 | 175675 | dispatchEvent, bubbling, no listeners | 177315 ± 0.02% | 177999 | | 7277.82 ± 0.38% | 7181.00 | 137404 | dispatchEvent, bubbling, single listener per target | 138560 ± 0.02% | 139256 | | 50493.64 ± 0.28% | 50105.00 | 19805 | dispatchEvent, bubbling, multiple listeners per target | 19855 ± 0.04% | 19958 | Reviewed By: yungsters Differential Revision: D67758408

…n of ReadOnlyNode (facebook#48427) Summary: Changelog: [internal] The `ReactNativeElement` class was refactored for performance reasons, and the current implementation does **NOT** call `super()`, and it inlines the parent constructor instead. When it eventually extends `EventTarget`, things won't work as expected because the existing `EventTarget` implementation has constructor dependencies. This refactors the current implementation of `EventTarget` to eliminate those constructor side-effects, and eliminates the constructor altogether. This breaks encapsulation, but it has some positive side-effects on performance: 1. Creating `EventTarget` instances is faster because it has no constructor logic. 2. Improves memory by not creating maps to hold the event listeners if no event listeners are ever added to the target (which is very common). 3. Improves the overall runtime performance of the methods in the class by migrating away from private methods (which are known to be slow on the babel transpiled version we're currently using). Extra: it also simplifies making window/the global scope implement the EventTarget interface :) ## Benchmark results Before: | Latency average (ns) | Latency median (ns) | Samples | Task name | Throughput average (ops/s) | Throughput median (ops/s) | | ---|--- |--- |--- |---|---| | 8234.22 ± 0.27% | 8132.00 | 121445 | dispatchEvent, no bubbling, no listeners | 122323 ± 0.02% | 122971 | | 9001.22 ± 0.41% | 8883.00 | 111097 | dispatchEvent, no bubbling, single listener | 111981 ± 0.02% | 112575 | | 51777.94 ± 0.58% | 51247.00 | 19314 | dispatchEvent, no bubbling, multiple listeners | 19393 ± 0.04% | 19513 | | 8256.65 ± 0.29% | 8152.00 | 121115 | dispatchEvent, bubbling, no listeners | 122031 ± 0.02% | 122669 | | 9064.32 ± 0.44% | 8933.00 | 110323 | dispatchEvent, bubbling, single listener per target | 111265 ± 0.02% | 111944 | | 51879.66 ± 0.27% | 51447.00 | 19276 | dispatchEvent, bubbling, multiple listeners per target | 19325 ± 0.04% | 19437 | After: | Latency average (ns) | Latency median (ns) | Samples | Task name | Throughput average (ops/s) | Throughput median (ops/s)| | ---------------------|---------------------|---------|--------------------------------------------------------|----------------------------|--------------------------| | 5664.62 ± 0.50% | 5588.00 | 176535 | dispatchEvent, no bubbling, no listeners | 178219 ± 0.02% | 178955 | | 7232.86 ± 0.50% | 7131.00 | 138258 | dispatchEvent, no bubbling, single listener | 139540 ± 0.02% | 140233 | | 50957.51 ± 0.71% | 50336.00 | 19625 | dispatchEvent, no bubbling, multiple listeners | 19751 ± 0.04% | 19866 | | 5692.36 ± 0.50% | 5618.00 | 175675 | dispatchEvent, bubbling, no listeners | 177315 ± 0.02% | 177999 | | 7277.82 ± 0.38% | 7181.00 | 137404 | dispatchEvent, bubbling, single listener per target | 138560 ± 0.02% | 139256 | | 50493.64 ± 0.28% | 50105.00 | 19805 | dispatchEvent, bubbling, multiple listeners per target | 19855 ± 0.04% | 19958 | Reviewed By: yungsters Differential Revision: D67758408

…nlyNode (facebook#48427) Summary: Changelog: [internal] The `ReactNativeElement` class was refactored for performance reasons, and the current implementation does **NOT** call `super()`, and it inlines the parent constructor instead. When it eventually extends `EventTarget`, things won't work as expected because the existing `EventTarget` implementation has constructor dependencies. This refactors the current implementation of `EventTarget` to eliminate those constructor side-effects, and eliminates the constructor altogether. This breaks encapsulation, but it has some positive side-effects on performance: 1. Creating `EventTarget` instances is faster because it has no constructor logic. 2. Improves memory by not creating maps to hold the event listeners if no event listeners are ever added to the target (which is very common). 3. Improves the overall runtime performance of the methods in the class by migrating away from private methods (which are known to be slow on the babel transpiled version we're currently using). Extra: it also simplifies making window/the global scope implement the EventTarget interface :) ## Benchmark results Before: | Latency average (ns) | Latency median (ns) | Samples | Task name | Throughput average (ops/s) | Throughput median (ops/s) | | ---|--- |--- |--- |---|---| | 8234.22 ± 0.27% | 8132.00 | 121445 | dispatchEvent, no bubbling, no listeners | 122323 ± 0.02% | 122971 | | 9001.22 ± 0.41% | 8883.00 | 111097 | dispatchEvent, no bubbling, single listener | 111981 ± 0.02% | 112575 | | 51777.94 ± 0.58% | 51247.00 | 19314 | dispatchEvent, no bubbling, multiple listeners | 19393 ± 0.04% | 19513 | | 8256.65 ± 0.29% | 8152.00 | 121115 | dispatchEvent, bubbling, no listeners | 122031 ± 0.02% | 122669 | | 9064.32 ± 0.44% | 8933.00 | 110323 | dispatchEvent, bubbling, single listener per target | 111265 ± 0.02% | 111944 | | 51879.66 ± 0.27% | 51447.00 | 19276 | dispatchEvent, bubbling, multiple listeners per target | 19325 ± 0.04% | 19437 | After: | Latency average (ns) | Latency median (ns) | Samples | Task name | Throughput average (ops/s) | Throughput median (ops/s)| | ---------------------|---------------------|---------|--------------------------------------------------------|----------------------------|--------------------------| | 5664.62 ± 0.50% | 5588.00 | 176535 | dispatchEvent, no bubbling, no listeners | 178219 ± 0.02% | 178955 | | 7232.86 ± 0.50% | 7131.00 | 138258 | dispatchEvent, no bubbling, single listener | 139540 ± 0.02% | 140233 | | 50957.51 ± 0.71% | 50336.00 | 19625 | dispatchEvent, no bubbling, multiple listeners | 19751 ± 0.04% | 19866 | | 5692.36 ± 0.50% | 5618.00 | 175675 | dispatchEvent, bubbling, no listeners | 177315 ± 0.02% | 177999 | | 7277.82 ± 0.38% | 7181.00 | 137404 | dispatchEvent, bubbling, single listener per target | 138560 ± 0.02% | 139256 | | 50493.64 ± 0.28% | 50105.00 | 19805 | dispatchEvent, bubbling, multiple listeners per target | 19855 ± 0.04% | 19958 | Reviewed By: yungsters Differential Revision: D67758408

Summary: Changelog: [internal] This implements a (mostly) spec-compliant version of the [`Event`](https://dom.spec.whatwg.org/#interface-event) and [`EventTarget`](https://dom.spec.whatwg.org/#interface-eventtarget) Web interfaces. It does not implement legacy methods in either of the interfaces, and ignores the parts of the spec that are related to Web-specific quirks (shadow roots, re-mapping of animation events with webkit prefixes, etc.). IMPORTANT: This only creates the interfaces and does not expose them externally yet (no `Event` or `EventTarget` in the global scope). Reviewed By: yungsters Differential Revision: D67738145

Summary: Changelog: [internal] Creates a benchmarks to measure the performance of `EventTarget`. Reviewed By: javache Differential Revision: D67750677

Summary: Changelog: [internal] Adds a regression test to make sure we implement the correct spec-compliant behavior for a possible bug in ~~the Web spec~~ __Chrome__: whatwg/dom#1346 Edit: the bug is in the Chrome implementation, not in the spec. Reviewed By: javache Differential Revision: D67758702

…n of ReadOnlyNode (facebook#48427) Summary: Changelog: [internal] The `ReactNativeElement` class was refactored for performance reasons, and the current implementation does **NOT** call `super()`, and it inlines the parent constructor instead. When it eventually extends `EventTarget`, things won't work as expected because the existing `EventTarget` implementation has constructor dependencies. This refactors the current implementation of `EventTarget` to eliminate those constructor side-effects, and eliminates the constructor altogether. This breaks encapsulation, but it has some positive side-effects on performance: 1. Creating `EventTarget` instances is faster because it has no constructor logic. 2. Improves memory by not creating maps to hold the event listeners if no event listeners are ever added to the target (which is very common). 3. Improves the overall runtime performance of the methods in the class by migrating away from private methods (which are known to be slow on the babel transpiled version we're currently using). Extra: it also simplifies making window/the global scope implement the EventTarget interface :) ## Benchmark results Before: | Latency average (ns) | Latency median (ns) | Samples | Task name | Throughput average (ops/s) | Throughput median (ops/s) | | ---|--- |--- |--- |---|---| | 8234.22 ± 0.27% | 8132.00 | 121445 | dispatchEvent, no bubbling, no listeners | 122323 ± 0.02% | 122971 | | 9001.22 ± 0.41% | 8883.00 | 111097 | dispatchEvent, no bubbling, single listener | 111981 ± 0.02% | 112575 | | 51777.94 ± 0.58% | 51247.00 | 19314 | dispatchEvent, no bubbling, multiple listeners | 19393 ± 0.04% | 19513 | | 8256.65 ± 0.29% | 8152.00 | 121115 | dispatchEvent, bubbling, no listeners | 122031 ± 0.02% | 122669 | | 9064.32 ± 0.44% | 8933.00 | 110323 | dispatchEvent, bubbling, single listener per target | 111265 ± 0.02% | 111944 | | 51879.66 ± 0.27% | 51447.00 | 19276 | dispatchEvent, bubbling, multiple listeners per target | 19325 ± 0.04% | 19437 | After: | Latency average (ns) | Latency median (ns) | Samples | Task name | Throughput average (ops/s) | Throughput median (ops/s)| | ---------------------|---------------------|---------|--------------------------------------------------------|----------------------------|--------------------------| | 5664.62 ± 0.50% | 5588.00 | 176535 | dispatchEvent, no bubbling, no listeners | 178219 ± 0.02% | 178955 | | 7232.86 ± 0.50% | 7131.00 | 138258 | dispatchEvent, no bubbling, single listener | 139540 ± 0.02% | 140233 | | 50957.51 ± 0.71% | 50336.00 | 19625 | dispatchEvent, no bubbling, multiple listeners | 19751 ± 0.04% | 19866 | | 5692.36 ± 0.50% | 5618.00 | 175675 | dispatchEvent, bubbling, no listeners | 177315 ± 0.02% | 177999 | | 7277.82 ± 0.38% | 7181.00 | 137404 | dispatchEvent, bubbling, single listener per target | 138560 ± 0.02% | 139256 | | 50493.64 ± 0.28% | 50105.00 | 19805 | dispatchEvent, bubbling, multiple listeners per target | 19855 ± 0.04% | 19958 | Reviewed By: yungsters Differential Revision: D67758408

facebook-github-bot · 2025-01-27T14:45:57Z

This pull request was exported from Phabricator. Differential Revision: D67758408

facebook-github-bot · 2025-01-27T14:46:48Z

This pull request was exported from Phabricator. Differential Revision: D67758408

facebook-github-bot · 2025-01-27T15:54:11Z

This pull request has been merged in 47e490f.

…nlyNode (facebook#48427) Summary: Pull Request resolved: facebook#48427 Changelog: [internal] The `ReactNativeElement` class was refactored for performance reasons, and the current implementation does **NOT** call `super()`, and it inlines the parent constructor instead. When it eventually extends `EventTarget`, things won't work as expected because the existing `EventTarget` implementation has constructor dependencies. This refactors the current implementation of `EventTarget` to eliminate those constructor side-effects, and eliminates the constructor altogether. This breaks encapsulation, but it has some positive side-effects on performance: 1. Creating `EventTarget` instances is faster because it has no constructor logic. 2. Improves memory by not creating maps to hold the event listeners if no event listeners are ever added to the target (which is very common). 3. Improves the overall runtime performance of the methods in the class by migrating away from private methods (which are known to be slow on the babel transpiled version we're currently using). Extra: it also simplifies making window/the global scope implement the EventTarget interface :) ## Benchmark results Before: | Latency average (ns) | Latency median (ns) | Samples | Task name | Throughput average (ops/s) | Throughput median (ops/s) | | ---|--- |--- |--- |---|---| | 8234.22 ± 0.27% | 8132.00 | 121445 | dispatchEvent, no bubbling, no listeners | 122323 ± 0.02% | 122971 | | 9001.22 ± 0.41% | 8883.00 | 111097 | dispatchEvent, no bubbling, single listener | 111981 ± 0.02% | 112575 | | 51777.94 ± 0.58% | 51247.00 | 19314 | dispatchEvent, no bubbling, multiple listeners | 19393 ± 0.04% | 19513 | | 8256.65 ± 0.29% | 8152.00 | 121115 | dispatchEvent, bubbling, no listeners | 122031 ± 0.02% | 122669 | | 9064.32 ± 0.44% | 8933.00 | 110323 | dispatchEvent, bubbling, single listener per target | 111265 ± 0.02% | 111944 | | 51879.66 ± 0.27% | 51447.00 | 19276 | dispatchEvent, bubbling, multiple listeners per target | 19325 ± 0.04% | 19437 | After: | Latency average (ns) | Latency median (ns) | Samples | Task name | Throughput average (ops/s) | Throughput median (ops/s)| | ---------------------|---------------------|---------|--------------------------------------------------------|----------------------------|--------------------------| | 5664.62 ± 0.50% | 5588.00 | 176535 | dispatchEvent, no bubbling, no listeners | 178219 ± 0.02% | 178955 | | 7232.86 ± 0.50% | 7131.00 | 138258 | dispatchEvent, no bubbling, single listener | 139540 ± 0.02% | 140233 | | 50957.51 ± 0.71% | 50336.00 | 19625 | dispatchEvent, no bubbling, multiple listeners | 19751 ± 0.04% | 19866 | | 5692.36 ± 0.50% | 5618.00 | 175675 | dispatchEvent, bubbling, no listeners | 177315 ± 0.02% | 177999 | | 7277.82 ± 0.38% | 7181.00 | 137404 | dispatchEvent, bubbling, single listener per target | 138560 ± 0.02% | 139256 | | 50493.64 ± 0.28% | 50105.00 | 19805 | dispatchEvent, bubbling, multiple listeners per target | 19855 ± 0.04% | 19958 | Reviewed By: yungsters Differential Revision: D67758408 fbshipit-source-id: f8da1788251c9e21377de5ab730875bcc7610361

facebook-github-bot added CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. p: Facebook Partner: Facebook Partner labels Jan 2, 2025

facebook-github-bot added the fb-exported label Jan 2, 2025

rubennorte force-pushed the export-D67758408 branch from 0972f3c to 0c25e5a Compare January 5, 2025 18:42

rubennorte force-pushed the export-D67758408 branch from 0c25e5a to 5e6513c Compare January 6, 2025 15:20

rubennorte force-pushed the export-D67758408 branch from 5e6513c to c8e4820 Compare January 6, 2025 16:54

rubennorte force-pushed the export-D67758408 branch from c8e4820 to 624bd0e Compare January 6, 2025 17:26

rubennorte force-pushed the export-D67758408 branch from 624bd0e to 47883ab Compare January 7, 2025 10:33

rubennorte force-pushed the export-D67758408 branch from 1943b40 to a1c79b4 Compare January 27, 2025 10:29

rubennorte force-pushed the export-D67758408 branch from a1c79b4 to b12a22a Compare January 27, 2025 10:31

rubennorte force-pushed the export-D67758408 branch from b12a22a to 8ae4f48 Compare January 27, 2025 14:45

rubennorte added 4 commits January 27, 2025 06:45

[skip ci] Add benchmark for EventTarget (facebook#48886)

e136430

Summary: Changelog: [internal] Creates a benchmarks to measure the performance of `EventTarget`. Reviewed By: javache Differential Revision: D67750677

rubennorte force-pushed the export-D67758408 branch from 8ae4f48 to 3dea235 Compare January 27, 2025 14:46

facebook-github-bot closed this in 47e490f Jan 27, 2025

facebook-github-bot added the Merged This PR has been merged. label Jan 27, 2025

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Make EventTarget compatible with the existing implementation of ReadOnlyNode #48427

Make EventTarget compatible with the existing implementation of ReadOnlyNode #48427

Uh oh!

rubennorte commented Jan 2, 2025

Uh oh!

facebook-github-bot commented Jan 2, 2025

Uh oh!

facebook-github-bot commented Jan 5, 2025

Uh oh!

facebook-github-bot commented Jan 6, 2025

Uh oh!

facebook-github-bot commented Jan 6, 2025

Uh oh!

facebook-github-bot commented Jan 6, 2025

Uh oh!

facebook-github-bot commented Jan 7, 2025

Uh oh!

facebook-github-bot commented Jan 27, 2025

Uh oh!

facebook-github-bot commented Jan 27, 2025

Uh oh!

facebook-github-bot commented Jan 27, 2025

Uh oh!

facebook-github-bot commented Jan 27, 2025

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Make EventTarget compatible with the existing implementation of ReadOnlyNode #48427

Make EventTarget compatible with the existing implementation of ReadOnlyNode #48427

Uh oh!

Conversation

rubennorte commented Jan 2, 2025

Uh oh!

facebook-github-bot commented Jan 2, 2025

Uh oh!

facebook-github-bot commented Jan 5, 2025

Uh oh!

facebook-github-bot commented Jan 6, 2025

Uh oh!

facebook-github-bot commented Jan 6, 2025

Uh oh!

facebook-github-bot commented Jan 6, 2025

Uh oh!

facebook-github-bot commented Jan 7, 2025

Uh oh!

facebook-github-bot commented Jan 27, 2025

Uh oh!

facebook-github-bot commented Jan 27, 2025

Uh oh!

facebook-github-bot commented Jan 27, 2025

Uh oh!

facebook-github-bot commented Jan 27, 2025

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants