function `pretty_codeblock` leaves an end tag for `code` element which is not open

[milmazz]

The current pretty_codeblock function in html.ex generates invalid HTML code, this function leaves an end tag for the code element which is not open.

iex> bin = "<pre><code>Hello world</code></pre><pre><code class=\"\">Hello world</code></pre>"
"<pre><code>Hello world</code></pre><pre><code class=\"\">Hello world</code></pre>"
iex> Regex.replace(~r/<pre><code\s*(class=\"\")?>/, bin, "<pre class=\"codeblock\">")
"<pre class=\"codeblock\">Hello world</code></pre><pre class=\"codeblock\">Hello world</code></pre>"

One way to solve this is updating the regex:

iex> Regex.replace(~r/<pre><code\s*(class=\"\")?>(.+?)<\/code><\/pre>/, bin, "<pre class=\"codeblock\">\\2</pre>")
"<pre class=\"codeblock\">Hello world</pre><pre class=\"codeblock\">Hello world</pre>"

But, if we look further, right now there is no rule for the .codeblock CSS class. So, I suggest to delete this pretty_codeblock and try to control the look of the <pre> and <code> tags in the CSS files, because there is where the rules that are in control of the style of HTML blocks belongs.

But, I really want to know your opinion about this matter before I propose a Pull Request.

[josevalim]

/cc @eksperimental

[milmazz]

Hello @tonini, I really want to know your opinion about this issue. Do you remember the pretty_codeblock function purpose?

[josevalim]

@milmazz the idea about pretty_codeblocks is that "" by defaults adds `<code class="">` and that is not going to be formatted as Elixir. Now that we are enabling "" everywhere, we should fix it by somehow setting the default language for such code blocks to Elixir (it can be with a regex) or by changing highlight.js to use elixir by default when another language was not specified.

Does it make sense?

[milmazz]

@josevalim Got it!, I'll try to work on this.

[milmazz]

This comment is just a reference of the different outputs per Markdown parser.

$ cat ~/test/demo.md
line 1
```elixir
iex> max(4, 5)
5
```
line 2
```
iex> max(4, 5)
5

Output with earmark ~/test/demo.md:

<p>line 1</p>
<pre><code class="elixir">iex&gt; max(4, 5)
5</code></pre>
<p>line 2</p>
<pre><code class="">iex&gt; max(4, 5)
5</code></pre>

Output with pandoc -f markdown+fenced_code_blocks -t html ~/test/demo.md:

<p>line 1</p>
<pre class="elixir"><code>iex&gt; max(4, 5)
5
</code></pre>
<p>line 2</p>
<pre><code>iex&gt; max(4, 5)
5</code></pre>

Output with hoedown --html --fenced-code ~/test/demo.md:

<p>line 1</p>

<pre><code class="language-elixir">iex&gt; max(4, 5)
5

</code></pre>

<p>line 2</p>

<pre><code>iex&gt; max(4, 5)
5
</code></pre>

/me Take this into account when working on the PR 👈

[milmazz]

This issue was fixed with #217 👍