Parsing <tr> as a document fragment doesn't work

[davetron5000]

Nokogiri::HTML5::DocumentFragment.parse("<tr><td>Hi</td></tr>")
=> #(DocumentFragment:0x3f6a8 { name = "#document-fragment", children = [ #(Text "Hi")] })

I can't understand why the HTML isn't a valid fragment and thus Nokogiri is stripping out the <tr> and <td> tags. Mostly I guess I'm not sure what is and is not a valid fragment.

I can make this work by using context:, but my use-case is more generic so I don't have an easy way to know if I need to use context: "tbody" or not.

[stevecheckoway]

Hi @davetron5000,

If you ask Nokogiri for errors, you'll see what's going on:

1:1: ERROR: Start tag 'tr' isn't allowed here. Currently open tags: html.
<tr><td>Hi</td></tr>
^
1:5: ERROR: Start tag 'td' isn't allowed here. Currently open tags: html.
<tr><td>Hi</td></tr>
    ^
1:11: ERROR: End tag 'td' isn't allowed here. Currently open tags: html.
<tr><td>Hi</td></tr>
          ^
1:16: ERROR: End tag 'tr' isn't allowed here. Currently open tags: html.
<tr><td>Hi</td></tr>
               ^

I think the easiest thing to do here (and something @flavorjones and I have discussed making the default but I don't recall what we decided about that to be honest) is to make template be your context.

irb(main):012:0> frag = Nokogiri::HTML5::DocumentFragment.parse("<tr><td>Hi</td></tr>", context: 'template', max_errors:
 100)
=>
#(DocumentFragment:0x55a00 {
...
irb(main):013:0> frag.errors.empty?
=> true
irb(main):014:0> frag.to_s
=> "<tr><td>Hi</td></tr>"

[davetron5000]

Oh, template is a good idea! I hadn't thought of that, but it should work for my use case.

Also, for anyone finding this who just learned about errors (which, uh, I definitely knew about before posting this :), none come back unless you set max_errors to greater than 0:

irb(main):008> x = Nokogiri::HTML5::fragment("<tr><td>Hi</td></tr>", max_errors: 100)
=> #(DocumentFragment:0x12620 { name = "#document-fragment", children = [ #(Text "Hi")] })
irb(main):009> x.errors
=> 
[#<Nokogiri::XML::SyntaxError:"1:1: ERROR: Start tag 'tr' isn't allowed here. Currently open tags: html.\n<tr><td>Hi</td></tr>\n^">,
 #<Nokogiri::XML::SyntaxError:"1:5: ERROR: Start tag 'td' isn't allowed here. Currently open tags: html.\n<tr><td>Hi</td></tr>\n    ^">,
 #<Nokogiri::XML::SyntaxError:"1:11: ERROR: End tag 'td' isn't allowed here. Currently open tags: html.\n<tr><td>Hi</td></tr>\n          ^">,
 #<Nokogiri::XML::SyntaxError:"1:16: ERROR: End tag 'tr' isn't allowed here. Currently open tags: html.\n<tr><td>Hi</td></tr>\n               ^">]

[stevecheckoway]

The max_errors behavior is something Nokogiri inherited from the Nokogumbo project. The option is described here. There are tradeoffs associated with making it default to 0 errors vs. some other default. See rubys/nokogumbo#78 for some old discussion of it.

[flavorjones]

@davetron5000 You may want to take a look at https://github.com/flavorjones/nokogiri-html5-inference if you're not sure exactly what you're parsing. But the TL;DR of that gem is that it uses <template> for fragments, yeah.

@stevecheckoway Should we make template the default? Is there a case when using <template> as the context tag would break someone who's parsing "in body" today?