rst code blocks with language tag not converting correctly to gfm

[mMerlin]

Fedora 27 4.18.19-100.fc27.x86_64
pandoc 2.5 «installed from tarbell»
Compiled with pandoc-types 1.17.5.4, texmath 0.11.1.2, skylighting 0.7.4

Searching through existing issues #721, #4523 look to be related. Sorry about the length of this, but the variation details could important.

Are there multiple flavours of rst, as there for markdown? This is a simplified extract from an rst document from a bitbucket wiki page.

.. sourcecode:: python

    import thonny
    thonny.THONNY_USER_DIR = "H:\\home\\.thonny"

passing that through pandoc -s --to=gfm --from=rst -o test.md test.rst give test.md as

<div class="sourcecode">

python

import thonny thonny.THONNY\_USER\_DIR = "H:\\home\\.thonny"

</div>

I would expect

```python
import thonny
thonny.THONNY\_USER\_DIR = "H:\\home\\.thonny"
```

rst is not a format I have worked with, but doing some research, I could not find an rst structure that pandoc would convert to the expected markdown. Here is a bit of an extended case in sourcecode.rst

.. sourcecode:: python

    import sys

.. code:: python

    import sys

.. code-block:: python

    import sys

.. codeblock:: python

    import sys

and pandoc -s --to=gfm --from=rst -o sourcecode.md sourcecode.rst give sourcecode.md as

<div class="sourcecode">

python

import sys

</div>

``` sourceCode python
import sys
```

``` sourceCode python
import sys
```

<div class="codeblock">

python

import sys

</div>

none of which looks like the expected markdown block. In fact, round trip does not work. Starting from block.md

```python
import sys
import os
```

``` python
import sys
import os
```

converting with pandoc -s --to=rst --from=gfm -o block.rst block.md gives block.rst as

.. code:: python

   import sys
   import os

.. code:: python

   import sys
   import os

and pandoc -s --to=gfm --from=rst -o block1.md block.rst gives block1.md as

``` sourceCode python
import sys
import os
```

``` sourceCode python
import sys
import os
```

Am I missing something obvious? rst is new to me, but gfm I have done a lot of work with.

[mb21]

You can also use -t native to see the intermediate representation.

But basically it seems to boil down to whether the RST reader should add the sourceCode class to the intermediate AST. I would say the answer is no (for both .. code:: and .. sourcecode::), because the markdown reader doesn't do it either.

[mMerlin]

Plus .. codeblock:: and ..code-block::?

Or is it that the markdown (gfm) writer should be ignoring the class? That would fix 2 of the 4 sample cases as well. The other 2 are coming through as embedded html, to they have more problems.

Assuming they are valid rst of course.

[jgm]

Looks like sourcecode is an undocumented synonym for code directive:
https://github.com/docutils-mirror/docutils/blob/e88c5fb08d5cdfa8b4ac1020dd6f7177778d5990/docutils/parsers/rst/languages/en.py

[jgm]

As for sourceCode class:

% ack sourceCode src
src/Text/Pandoc/Readers/LaTeX.hs
1826:       (codeBlockWith ("",["sourceCode","literate","haskell"],[]) <$>

src/Text/Pandoc/Readers/Markdown.hs
751:  (return . B.codeBlockWith ("",["sourceCode","literate","haskell"],[]) <$>
753:    <|> (return . B.codeBlockWith ("",["sourceCode","haskell"],[]) <$>

src/Text/Pandoc/Readers/RST.hs
424:  return $ B.codeBlockWith ("", ["sourceCode", "literate", "haskell"], [])
998:          classes' = "sourceCode" : lang
1417:    "code" -> return $ B.codeWith (addClass "sourceCode" attr) contents

src/Text/Pandoc/Writers/RST.hs
291:                     c `notElem` ["sourceCode","literate","numberLines",

Note that skylighting adds the sourceCode class in HTML. I think we should modify the readers not to add it (and we could tell the RST writer not to ignore it).