Fix Markdown br bug

[boscoh]

In the :markdown filter, the markdown for implicit <br> tag is broken.

The reason is that markdown recognises a line with <br> by two trailing spaces.

However, trailing spaces are stripped when PlaintextNode.haml is read.

To fix: when the text is reconstructed to send to the Markdown filter, instead of using the PlaintextNode.haml field, PlaintextNode.raw_haml.lstrip() is used, which keeps the trailing spaces.

[noelbush-xx]

Thanks very much! Any chance of writing a quick unit test for this? Would be much appreciated.

[boscoh]

I rewrote the markdown test in the markdown_br_fix branch to add the
check.

[noelbush-xx]

Thanks for updating the test. It doesn't seem to be passing for me, though:

FAIL: test_filters_markdown (hamlpy.test.template_compare_test.TestTemplateCompare)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/home/nsh/projects/HamlPy/hamlpy/test/template_compare_test.py", line 47, in test_filters_markdown
self._compare_test_files('filtersMarkdown')
  File "/home/nsh/projects/HamlPy/hamlpy/test/template_compare_test.py", line 124, in _compare_test_files
eq_(parsed, html)
AssertionError: u'<p>hello</p>\n<p>no paragraph</p>\n<p>line break  </p>\n<p>follow</p>\n<p>New paragraph</p>\n<pre><code>code block\n</code></pre>' != '<p>hello\nno paragraph</p>\n<p>line break<br />\nfollow</p>\n<p>New paragraph</p>\n<pre><code>code block\n</code></pre>'
-------------------- >> begin captured stdout << ---------------------

HTML (actual): 
1. <p>hello</p>
2. <p>no paragraph</p>
3. <p>line break  </p>
4. <p>follow</p>
5. <p>New paragraph</p>
6. <pre><code>code block
7. </code></pre>
Difference begins at line 1 column 9
HTML (actual, len=12)   : <p>hello</p>
HTML (expected, len= 8) : <p>hello
Character code (actual)  : 60 (<)
Character code (expected): 10 (
)

--------------------- >> end captured stdout << ----------------------

[boscoh]

The origin of that was a bit subtle. Turns out hamlpy.py reads the input file using read().splitlines(), whereas the template tests uses readlines(). This gives different line endings w.r.t. to '\n'. Now the markdown filter must explicitly check for the line endings, to remove it.