Below is the generated HTML for with line numbers and without line numbers, respectively.
```html
<!-- Without line numbers -->
<pre class="sourceCode scala">
<code class="sourceCode scala">
<span class="kw">trait</span>
Monad[M[_]]
</code>
</pre>
<!-- With line numbers -->
<table class="sourceCode scala numberLines">
<tr class="sourceCode">
<td class="lineNumbers">
<pre>1</pre>
</td>
<td class="sourceCode">
<pre>
<code class="sourceCode scala">
<span class="kw">trait</span>
Monad[M[_]]
</code>
</pre>
</td>
</tr>
</table>
```
The markdown used is:
```
~~~~ { .scala }
trait Monad[M[_]]
~~~~
~~~~ { .scala .numberLines}
trait Monad[M[_]]
~~~~
```
Notice that without line numbers the outer `pre` has a `sourceCode` class whereas with line numbers the `sourceCode` class is only put in `code` and table-related tags.
In the original `syntax.css` file synax highlighting was only triggered if the `kw` and friends tags were within a `pre.sourceCode` - that means it did not trigger with line numbers. This change removes the requirement for the `pre` prefix so that highlighting is triggered with or without line numbers, so long as the code is within *some* `.sourceCode` tag.
I had to prepend some Rules to global Rules set. This might be possible
to replaced by a correct Store.set call.
I also had to prepend some Compile rules.
Pandoc 1.12 decouples citeproc-hs from itself, so there is no longer a
Text.Pandoc.Biblio module in Pandoc. Further, citeproc-hs depends on
pandoc-types 1.10 but Pandoc 1.12 depends on pandoc-types 1.12. To
alleviate these issues, pandoc-citeproc was created which includes a
copy of the citeproc-hs source made to be compatible, since the
developer of citeproc-hs is apparently MIA.
pandoc-citeproc is a separate module that handles the mixture of
citeproc-hs and Pandoc. It includes `processCites` in Text.CSL.Pandoc,
which is the new name of what used to be `processBiblio` from
Text.Pandoc.Biblio
Most of these changes are seamless, consisting of simple name changes in
both functions and modules. However, a more direct change in the
Hakyll API itself is that `readPandocBiblio`'s second parameter, the
CSL, is now mandatory, i.e. not of type Maybe. This is to reflect the
same change in the underlying processing function from Text.CSL.Pandoc,
`processCites`, where the Style argument is now mandatory, and the style
is derived from the CSL.
See the old function:
processBiblio :: Maybe Style -> [Reference] -> Pandoc -> Pandoc
Compared to the new one:
processCites :: Style -> [Reference] -> Pandoc -> Pandoc
Sources:
* http://hackage.haskell.org/packages/archive/pandoc/1.11.1/doc/html/Text-Pandoc-Biblio.html
* http://hackage.haskell.org/packages/archive/pandoc-citeproc/0.1/doc/html/Text-CSL-Pandoc.html
Similarly, there is no longer a `readerReferences` field in the reader
options structure.
toSiteRoot uses splitPath, joinPath, and takeDirectory from the
System.FilePath module. On Windows systems, the implementation of
joinPath will build up a path using the Windows path separator "\\".
We don't want this behavior since the paths we are working with
are always URLs, so we force POSIX behavior for System.FilePath.
