Issues with UTF-8 encoded href attributes

[rustyconover]

Hi Team,

Assume you have an HTML file that is UTF-8 encoded, and there is an emoji included in an href attribute of the <a> tag. The emoji is not escaped or anything else it is just a properly formed UTF-8 glyph. Such as:

<a href="https://rusty.today/🏈">Test link</a>

When attempting to move that asset this error is thrown:

URIError: moveAssets transform: URI malformed
    at encodeURIComponent (<anonymous>)
    at String.replace (<anonymous>)
    at AssetGraph.addAsset (/Users/rusty/Development/assetgraph/lib/AssetGraph.js:241:45)
    at Html._vivifyRelation (/Users/rusty/Development/assetgraph/lib/assets/Asset.js:1090:50)
    at /Users/rusty/Development/assetgraph/lib/assets/Asset.js:1108:34
    at Array.map (<anonymous>)
    at Html.get outgoingRelations [as outgoingRelations] (/Users/rusty/Development/assetgraph/lib/assets/Asset.js:1107:73)
    at gatherExternalRelations (/Users/rusty/Development/assetgraph/lib/assets/Asset.js:1180:44)
    at Html.get externalRelations [as externalRelations] (/Users/rusty/Development/assetgraph/lib/assets/Asset.js:1187:7)
    at Html.set url [as url] (/Users/rusty/Development/assetgraph/lib/assets/Asset.js:1022:37)
    at /Users/rusty/Development/assetgraph/lib/util/assetMover.js:21:17
    at moveAssets (/Users/rusty/Development/assetgraph/lib/transforms/moveAssets.js:10:7)
    at Immediate._onImmediate (/Users/rusty/Development/assetgraph/lib/AssetGraph.js:653:25)

This is because this code in addAsset() is called:

          assetConfig.url = assetConfig.url.replace(
            /[^\x21-\x7f]/g,
            encodeURIComponent
          );

That code mangles the nicely formed emoji character.

Rusty

[papandreou]

To be fair, unencoded emojis are invalid per https://tools.ietf.org/html/rfc3986 as they aren't ascii. But it would probably be good to try to be as forgiving as browsers are.