Performance Improvements #222
Conversation
| @@ -997,513 +1048,45 @@ | |||
| } | |||
| } | |||
|
|
|||
| switch (parser.state) { | |||
There was a problem hiding this comment.
Each case in this switch has been moved into its own function.
| @@ -962,6 +980,42 @@ | |||
| return result | |||
| } | |||
|
|
|||
| var parsers = {} | |||
There was a problem hiding this comment.
This creates a mapping from each state to the correct function. Because of the number of branches in the original switch statement, this approach leads to performance gains due to significantly fewer compares/jumps.
| @@ -1562,4 +1145,460 @@ | |||
| } | |||
| }()) | |||
| } | |||
|
|
|||
There was a problem hiding this comment.
This is the beginning of the functions that were extracted from the original switch statement. For the most part they are just copied/pasted from their original location w/ little to no alteration (other then changing continue to return in a few places)
|
@isaacs I see there are a few PR's on this repo that date back a ways. Is this project still actively maintained? |
This PR includes purely performance related improvements. As such, there are no new tests because there is no new functionality. I hope that's ok. If there are some tests you would like to see added to cover these changes, please let me know what you'd like to see.
The primary changes include
nameStartandnameBodyswitchstatement and replacing with a look up tableif/else ifstatements to make the more probable case come firstAlso, it seems that by default
trackPositionends up being true. Turing it off gives another ~10% boost. However it wasn't completely clear to me if this default could safely be changed. If the default can't be changed maybe it could at least be mentioned in the docs/examples?Please see the attached files for a simple test case. In my testing this PR improved performance by ~%33. Unfortunately I primarily tested against only the one xml file included. Due to some time constraints on my end I didn't create a very broad test suite :(. If you'd like to see the performance proven against a wider array of example files, let me know and I'll try to put something together.
tests.zip