Test Suite

Python-Markdown comes with a test suite which uses the Nose testing framework and YAML. The test suite primarily serves to ensure that new bugs are not introduced as existing bugs are patched or new features are added. It also allows Python-Markdown to be tested with the tests from other implementations such as John Gruber’s Perl implementation or Michel Fortin’s PHP implementation.

The test suite can be run by calling the run_tests.py command at the root of the distribution tarball or by calling the nosetests command directly. Either way, Nose will need to be installed on your system first (run easy_install nose). Any standard nosetests configuration options can be passed in on the command line (i.e.: verbosity level or use of a plugin like coverage).

Additionally, a nicely formatted HTML report of all output is written to a temporary file in test-output.html. Open the file in a browser to view the report.

A tox.ini file is also provided, so tox can be used to automatically create virtual environments, install all testing dependencies and run the tests on each supported Python version. See the wiki for instructions on setting up a testing environment to use tox.

The test suite contains two kinds of tests: Markdown Syntax Tests and Unit Tests.

Markdown Syntax Tests

The Syntax Tests are in the various directories contained within the ‘tests’ directory of the packaged tarball. Each test consists of a matching pair of text and HTML files. The text file contains a snippet of Markdown source text formatted for a specific syntax feature and the HTML file contains the expected HTML output of that snippet. When the test suite is run, each text file is run through Markdown and the output is compared with the HTML file as a separate Unit Test.

In fact, this is the primary reason for using Nose, it gives us an easy way to treat each of these tests as a separate unit test which is reported on separately. Additionally, with the help of a couple custom Nose plugins which are included with the Markdown Test Suite, we are able to get back an easy to read diff of the actual output compared to expected output when a test fails.

Here is some sample output with a test that is failing because of some insignificant white space differences:

$ ./run-tests.py
..........................................................M...........
............................SSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSS
SSSSSSSSSS.................S..........................................
.........
======================================================================
MarkdownSyntaxError: TestSyntax: "misc/lists3"
----------------------------------------------------------------------
MarkdownSyntaxError: Output from "/home/waylan/code/python-markdown/te
sts/misc/lists3.txt" failed to match expected output.

--- /home/waylan/code/python-markdown/tests/misc/lists3.html 
+++ actual_output.html 
@@ -1,5 +1,5 @@ 
 <ul> 
 <li>blah blah blah 
-sdf asdf asdf asdf asdf 
-asda asdf asdfasd</li> 
+    sdf asdf asdf asdf asdf 
+    asda asdf asdfasd</li> 
 </ul>

---------------------------------------------------------------------- 
Ran 219 tests in 7.698s

FAILED (MarkdownSyntaxError=1, SKIP=53)

Note that 219 tests were run, one of which failed with a MarkdownSyntaxError. Only Markdown Syntax Tests should fail with a MarkdownSyntaxError. Nose then formats the error reports for MarkdownSyntaxErrors so that they only include useful information. Namely the text file which failed and a unified diff showing the failure. Without the plugin, you would also get a useless traceback showing how the code stepped through the test framework, but nothing about how Markdown actually ran.

If, on the other hand, a Syntax Test failed because some other exception gets raised by either Markdown or the test suite, then that would be reported as per a normal unit test failure with the appropriate traceback for debugging purposes.

Syntax Test Configuration Settings

The other thing to note about the above example is that 53 tests were skipped. Those tests have been explicitly configured to be skipped as they are primarily tests from either PHP or Perl which are known to fail for various reasons. In fact, a number of different configuration settings can be set for any specific test.

Each Syntax Test directory contains a test.cfg file in the YAML format. The file may contain a separate section for each text file named exactly as the file is named minus the file extension (i.e.; the section for a test in foo.txt would be foo). All settings are optional. Default settings for the entire directory can be set under the DEFAULT section (must be all caps). Any settings under a specific file section will override anything in the DEFAULT section for that specific test only.

Below are the configuration options available and the defaults used when they are not explicitly set.

  • normalize: Switches white space normalization of the test output on or off. Defaults to False (off). Note: This requires that PyTidyLib be installed on the system. Otherwise the test will be skipped, regardless of any other settings.
  • skip: Switches skipping of the test on and off. Defaults to False (off).
  • input_ext: Extension of input file. Defaults to .txt. Useful for tests from other implementations.
  • output_ext: Extension of output file. Defaults to .html. Useful for tests from other implementations.
  • Any keyword argument accepted by the Markdown class. If not set, Markdown’s defaults are used.

Unit Tests

Unit Tests are used as regression tests for Python-Markdown’s API. All Unit Tests shipped with Python-Markdown are standard Python Unit Tests and are all contained in tests/test_apis.py and tests/test_extensions.py. Standard discovery methods are used to find and run the tests. Therefore, when writing new tests, those standards and naming conventions should be followed.