Erlmarkdown uses generated tests. There are 2 levels of generation:
./tests/index.html uses generate_tests.js to take a set of input test cases
and output an EUnit test file for Erlang.
Some of the tests in generate_tests.js are hand written and some have been generated from the MarkdownSharp test suites.
The MarkdownSharp test suites were originally downloaded from:
http://code.google.com/p/markdownsharp/
They can be found in ./tests/markdowndownsharp/
The code that generates tests from them in in the escript file:
./priv/make_tests
Bug fix for "<>"
Bug Fix for tags with a couple more shonky unfixed whitespace bugs
Shonk Alert! A couple of whitespace bugs remain unfixed
Wasn't handling special white space inserted by the PARSER (don't ask!)
Some blocktag/html errrors fixed
URl's were not html encoded.
Failing tests of 1.1.5 not addressed in this release
Can now specify a UTF 8 entry point - this addresses the non-breaking space/ascii 160 problem of Version 1.1.3
Failing tests of 1.1.5 not addressed in this release
Integration of additional tests from markdownsharp: http://code.google.com/p/markdownsharp/source/browse/trunk/MarkdownSharpTests/#MarkdownSharpTests/testfiles%3Fstate%3Dclosed
17 currently failing which is why this is an interim release
Fixes 4 sets of bugs:
- you can now put an image inside an href
(eg [](http://example.com/dir/) - text in ordered and unordered lists now renders correctly eg bold and italic etc
- the 'you can't have 2 code segments in a single line' bug is fixed
Under certain circumstances non-breaking spaces (ascii 160) could wedge the server.
fuzz.erl rejigged to make for better diagnostics and run with 1,000,000 random characters against markdown.
CAVEAT: having serious doubts as to how erlmarkdown will handle unicode
A number of failures have been noticed in production of bits of markdown failing to convert.
A fuzz generator has been added to produce large amounts or random characters with particular emphasis on characters that are 'structural' in markdown. 50,000 random character strings found 3 or 4 new bugs which have been fixed in this release.
Fixed a bug involving underscores/formatting characters inside inline references and images
In the beginning erlmarkdown was an implementation of markdown written to spec
- the spec being the daring-fireball syntax page.
BUT it soon transpired that there were significant differences between our implementation of the spec and that of showdown (from attacklabs.net). Given that the client-side markdown and the server-side markdown need to work together we switched the erlang code from being an implementation of the spec to being a server-side implementation of showdown.
BUT the WMD dialog box produces (as input) code which is 'off spec' (but which showdown supports), so...
Version 1.1 is tested for compatibility with:
- showdown
- the markdown input produced by the WMD dialog box editor
We are tracking the release of WMD and showdown from this version on github: http://github.com/derobins/wmd/commit/980f68797307d28f0541868c740974cb2eeb1209
We are no longer tracking the showdown code on attacklabs.net
As a result of this is that the list of ERATTA AND KNOWN BUGS is a bit longer
This is a major rewrite. There is little point writing servers-side markdown without a client-side markdown to preview it.
This markdown will now track showdown from Attack Labs as its twin implementation.
http://attacklab.net/showdown/
To that end the test suite has been repurposed to test compatibility with showdown. The current release of showdown used is V0.9
The directory /tests now contains an webpage index.html which generates
the tests (ie the file markdown_tests.erl in /src).
index.html loads a javascript file generate_tests.js which holds a list of
strings which it generates the tests from (ie testing if markdown produces the
same output as showdown...)
There are a number of places where markdown is not whitespace compatible with showdown - you can inspect these by looking for commented out tests in
generate_tests.js. There are also a number of showdown bugs or other
inconsistencies which also show up as commented out tests in
generate_tests.js.
The biggest single difference is that showdown doesn't escape any html tags
by default whereas markdown has a whitelist of blocklevel tags which it doesn't
escape (see the function markdown:is_block_tag/1).
(Other) Known Bugs and Errata: as per the previous versions (see bottom of document)
There are 261 Unit Tests.
Various bug fixes:
- supports short inline URL references of the form 'some text [id] some more'
- fixes bugs with lines containing markdown only like '#', '>', '=' and '-'
- fixes a bug where a terminating '\n\n' would mean each line came back reversed
- fixes various 'atx header' bugs
- fixes bugs where lines start with an emphasis or strong marker (thanks to Tom McNulty)
Unit tests bumped up to 255.
This version comes with 226 Unit Tests and is being released for pre-production testing.
It also has 1 System Test - this document :)
[id]: http://sub.dom.tld:1234/some/path\\t\\n(Title)
The title wrap to a new line is not supported in this release.
I'm not entirely clear how character escaping should work for:
\ backslash
\` backtick
* asterisk
_ underscore
{} curly braces
[] square brackets
() parentheses
# hash mark
+ plus sign
- minus sign (hyphen)
. dot
! exclamation mark
I escape them only in context.
The official release obfuscated e-mail addresses, this don't :(
Showdown supports nested
- 's,
- 's and
's - erlmarkdown doesnt (yet)...This is an Erlang implementation of Markdown.
See http://daringfireball.net/projects/markdown/ for details
The license is the same as the original Markdown
'My Song' when I wrote this was: http://open.spotify.com/track/6mqdunuFFSODHKcpDTFvAj
'Generic Song' to play when you find a bug in my code: http://open.spotify.com/track/5YLa8jWik5OqgBoSix3NUp
'Generic Song' about how I feel about myself when you find a bug: http://open.spotify.com/track/2RbJj7D5pRff82NtDDSwah