Kikobeats / simple-wappalyzer / 40593324518

100%

push

github

perf(parse): read meta and script tags with htmlparser2 instead of happy-dom (#128)

* perf(load): compile technologies on the first analyze call

setTechnologies compiled every pattern of 2.8MB of technology data at
require time (~70ms plus ~20ms JSON parse). Loading it on the first call
keeps require cheap for processes that import the module but never call
it, or call it later.

A/B, node benchmark/index.js, 9 alternating runs per side, median (range):

| case                  | base                  | change                | delta  |
| --------------------- | --------------------- | --------------------- | ------ |
| cold require          | 204.3ms (201.0-210.7) | 121.1ms (117.9-122.3) | -40.7% |
| analyze kikobeats.com | 51.8ms (50.1-53.6)    | 53.0ms (50.8-54.8)    | noise  |
| analyze microlink.io  | 136.1ms (134.2-138.0) | 137.2ms (135.5-141.4) | noise  |
| analyze theverge.com  | 216.5ms (211.6-223.8) | 217.0ms (215.3-223.0) | noise  |
| analyze vercel.com    | 144.8ms (138.8-146.8) | 143.7ms (142.5-146.7) | noise  |

require plus the first analyze call is unchanged (~281ms vs ~284ms):
the cost moves to the first call, it is not removed.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* perf(parse): read meta and script tags with htmlparser2 instead of happy-dom

happy-dom built a full DOM for every page, plus a Window, only to read
<meta name|property content> and <script src>. htmlparser2's streaming
Parser collects those two tags and builds nothing else. happy-dom is
dropped from dependencies (21.9MB, 7 packages, 113ms to require).

A/B against the previous commit, node benchmark/index.js, 7 alternating
runs per side, median (range):

| case                  | before                | after              | delta  |
| --------------------- | --------------------- | ------------------ | ------ |
| analyze kikobeats.com | 51.4ms (50.6-52.4)    | 34.1ms (33.9-36.5) | -33.7% |
| analyze microlink.io  | 131.2ms (130.7-132.4) | 104.9ms (1... (continued)

50 of 50 branches covered (100.0%)

Branch coverage included in aggregate %.

74 of 74 new or added lines in 1 file covered. (100.0%)

114 of 114 relevant lines covered (100.0%)

418.83 hits per line