Jsoup upgrade

[borkdude]

No description provided.

[danielcompton]

@borkdude note that jsoup 1.20.1 introduces some breaking changes around self-closing tags: jhy/jsoup#2322, https://github.com/jhy/jsoup/releases/tag/jsoup-1.20.1.

Possibly it might be good to add a parse-xml/parse-xml-fragment function to hickory.core to match the API for parse/parse-fragment for those who are using Hickory to parse SVGs (like we were).

In the meantime, I've created this function for our own usage:

(defn parse-xml-fragment
  ;; Version of hickory/parse-fragment specialised for XML.
  ;; https://github.com/jhy/jsoup/discussions/2322
  "Parse an XML fragment (some group of tags that might be at home somewhere
   in the tag hierarchy) into a list of elements that can
   each be passed as input to as-hiccup or as-hickory."
  [s]
  (into [] (Parser/parseXmlFragment s "https://example.com")))

[borkdude]

@danielcompton Thanks for the note! I didn't notice any breakage since no tests were breaking. Does Jsoup break for HTML with svg elements in them? Or is that only a concern when you're using Jsoup to parse an XML fragment standalone?

[danielcompton]

Here is an example test I wrote with two self-closing tags side-by-side.

  (testing "parse-svg returns consistent hiccup structure"
    (let [svg-input "<rect x=\"10.5\" y=\"2\" width=\"3\" height=\"14\" rx=\"1\" ry=\"1\" fill=\"#444\"/><circle data-color=\"color-2\" cx=\"12\" cy=\"20\" r=\"2\" fill=\"#444\"/>"
          result (parse/parse-svg svg-input)
          expected-result [[:rect {:x "10.5" :y "2" :width "3" :height "14" :rx "1" :ry "1" :fill "#444"}]
                           [:circle {:data-color "color-2" :cx "12" :cy "20" :r "2" :fill "#444"}]]]
      (is (= expected-result result)
          "SVG parsing should produce stable structure across jsoup/hickory upgrades")))

Using the new version of jsoup, the result will have the circle nested inside the rect

[[:rect {:x "10.5" :y "2" :width "3" :height "14" :rx "1" :ry "1" :fill "#444"}
  [:circle {:data-color "color-2" :cx "12" :cy "20" :r "2" :fill "#444"}]]]

I would expect that jsoup would handle SVG inside HTML correctly; the issue is that I'm parsing an SVG/XML fragment as HTML. That previously worked by accident, but was never a valid use-case.

[borkdude]

Yeah it seems wrapping the input in an svg element does the right thing:

user=> (c/as-hiccup (first (c/parse-fragment "<svg><rect x=\"10.5\" y=\"2\" width=\"3\" height=\"14\" rx=\"1\" ry=\"1\" fill=\"#444\"/><circle data-color=\"color-2\" cx=\"12\" cy=\"20\" r=\"2\" fill=\"#444\"/></svg>")))
[:svg {} [:rect {:x "10.5", :y "2", :width "3", :height "14", :rx "1", :ry "1", :fill "#444"}] [:circle {:data-color "color-2", :cx "12", :cy "20", :r "2", :fill "#444"}]]

so maybe that's a good workaround for your use case as well.