Use htmlparser2 for tolerant & webworker-friendly parsing about togeojson HOT 4 CLOSED

Okay, I've encountered this again in the wild, and have a plan. Basically the options are:

• Remove namespaces from the XML as a string, using a regex, so that DOMParser can handle it. This is iffy, because regexes + XML, plus the performance effects.
• Swap out the XML datastructure and parser for something that is more tolerant. This comes with the bonus that toGeoJSON will work in a webworker.

Hence: renaming. Going to rewrite togeojson to use htmlparser2 (which can parse XML).

from togeojson.

Can you post the file itself so I can help debug this?

from togeojson.

@tmcw sure! It was a monster, so I've removed all but one <Placemark> so I could fit in a gist, but you can still see the issue:

huc6-no-xsi.kml ❌

huc6-with-xsi.kml ✅

Here's a Codepen where you can see 0 vs 1 feature in browser: https://codepen.io/davetapley/pen/JjJRdBv?editors=0011

But with CLI both return the feature:

$ curl -s https://gist.githubusercontent.com/davetapley/c426f7ad68027bc292f5a03837ab50e4/raw/f5974f2e47c3af8c39b02116a160759624b0c0cd/huc6-no-xsi.kml | npx -q @tmcw/togeojson-cli | jq -e '.features | length'
1


$ curl -s https://gist.githubusercontent.com/davetapley/fa6f8e2e96b03f2911c27f20f4123107/raw/7f0b4295e367be50309be47e70f1fda3845b91bd/huc6-with-xsi.kml | npx -q @tmcw/togeojson-cli | jq -e '.features | length'
1

from togeojson.

Okay, I implemented this in #56 but I think that's the wrong solution. htmlparser2 is surprisingly a larger dependency than @xmldom/xmldom, and requires a lot more shifted-around code.

I think the best path here is actually just to recommend using @xmldom/xmldom as a parser, even on the web.

from togeojson.