Incorrect conversion of tag-inline ERB

[aprescott]

Reproduced this on master:

$ cat /tmp/blah.erb 
<%
  attrs = 'data-a="1" data-b="2"'
  class = "some-class"
%>

<div class="panel <%= class -%>" <%= raw attrs -%>>
</div>

$ html2haml /tmp/blah.erb /tmp/blah.haml

$ cat /tmp/blah.haml
- attrs = 'data-a="1" data-b="2"'
- class = "some-class"
.panel
= class
" #{raw attrs}>

This should instead be

- attrs = 'data-a="1" data-b="2"'
- class = "some-class"
<div class="panel #{class}" #{raw attrs}>

html2haml 1.0.1 generates the expected HAML, but the latest version doesn't.

[snatchev]

@aprescott I am not sure if html2haml could ever produce the expected output in the case. The issue is that it will not create an html <div> tag like that. The second issue is that there is no haml representation of raw attrs as your have shown in the expected output. They need to be a key-value pair of some kind.

[aprescott]

@snatchev html2haml 1.0.1 handles this case correctly, though.

[snatchev]

@aprescott Good point. I will look into how it handled it. Thanks for reporting.

[bvogel]

<div class="panel #{class}" #{raw attrs}> does not look like haml, I'd expect something like
.panel{class: class}, however I'm unsure how the raw attrs should be handled correctly. I'd expect something like .panel{class: class, data: attrs} with attrs = {a: "1", b: "2"}

[aprescott]

Oh, that's a good point. Sorry @snatchev, getting my HTML and HAML confused. So yes, 1.0.1 "handles" this case for some value of "handle", but I guess any ERB that inserts attributes directly as with #{raw ...} here — which could just as well be #{anything} — should ideally be handled. Not sure if that's possible though with HAML?

[snatchev]

@aprescott this test case is really difficult. For one, using a variable called class is not valid ruby, so html2haml will not treat it like a code block. Secondly, as mentioned above, there is no HAML equivalent to #{raw attrs} so the only possible solution would be for html2haml to not convert the line and keep it as HTML. This would still be valid Haml.