One of the challenges of working with JavaScript and libxml is that they employ different string encodings. JavaScript uses UTF-16/UCS-2, while libxml uses UTF-8. Consequently, we encounter the need to convert strings every time we invoke a function from libxml.
To mitigate this overhead, this library provides two types of parsing functions: String API and Buffer API. The String API accepts a JavaScript string as input and converts it into a UTF-8 buffer before processing. Conversely, the Buffer API takes a UTF-8 buffer as input and directly processes it. The Buffer API proves to be faster, particularly for large XML documents. Therefore, we strongly recommend using the Buffer API whenever feasible to enhance the performance of your code.
The benchmark report below evaluates the performance of DOM tree parsing among the following libraries:
To execute the benchmark, ensure that the lib is built first:
npm ci && npm run build
Then, navigate to the benchmark directory and execute the following commands:
cd benchmark && npm ci
npm test
Environment: NodeJs v18.19.0 on Darwin arm64 Apple M3 Max
Running "fixtures/small.xml: 780 bytes" suite...
Progress: 100%
libxml2-wasm:
120 578 ops/s, ±0.12% | 28% slower
libxml2-wasm(buffer api):
167 479 ops/s, ±0.37% | fastest
libxmljs2:
78 688 ops/s, ±1.05% | 53.02% slower
@rgrove/parse-xml:
78 790 ops/s, ±0.19% | 52.96% slower
fast-xml-parser:
52 077 ops/s, ±0.27% | 68.91% slower
xmldoc:
41 125 ops/s, ±0.42% | slowest, 75.44% slower
Running "fixtures/medium.xml: 35562 bytes" suite...
Progress: 100%
libxml2-wasm:
6 013 ops/s, ±0.10% | 64.87% slower
libxml2-wasm(buffer api):
17 114 ops/s, ±0.14% | fastest
libxmljs2:
12 414 ops/s, ±1.82% | 27.46% slower
@rgrove/parse-xml:
5 049 ops/s, ±0.20% | 70.5% slower
fast-xml-parser:
2 708 ops/s, ±0.46% | 84.18% slower
xmldoc:
1 912 ops/s, ±0.41% | slowest, 88.83% slower
Running "fixtures/large.xml: 2337522 bytes" suite...
Progress: 100%
libxml2-wasm:
43 ops/s, ±0.11% | 30.65% slower
libxml2-wasm(buffer api):
62 ops/s, ±0.27% | fastest
libxmljs2:
50 ops/s, ±1.41% | 19.35% slower
@rgrove/parse-xml:
19 ops/s, ±0.93% | 69.35% slower
fast-xml-parser:
16 ops/s, ±0.37% | 74.19% slower
xmldoc:
11 ops/s, ±1.13% | slowest, 82.26% slower