Use dtf for Documentation Consistency Testing

dtf provides tools to define testable parameters for documentation and verify that documentation is correct and well formed. To test documentation, users define two components:

  • a “case”, or Python module that defines the parameters of the test, and implements the test itself.
  • a “test” or YAML file/document that defines inputs to the test, including the files, and any other required meta data.

Users can define multiple tests and cases as needed, though typically users will define a handful of cases and a much larger number of tests. Goals of dtf are:

  • to make conditions (i.e. tests) easy to define for technical writers and documentation maintainers, without needing to write code.
  • to help documentation teams encode a style and verify adherence automatically without relying on individuals to police all aspects of editorial style.
  • to automate checking of various internal dependencies to help make larger multi-file resources manageable.

Continue reading for a description of the use and implementation of dtf in your projects.


Install using pip. The process resembles the following:

cd ~/
git clone git://
cd ~/dtf
python bdist
sudo pip install dist/dtf-*

dtf should now be accessible in your system path, as dtf.

In your project you will want to create two directories, named case/, for case modules, and test/, for test definitions.

Command Line Operation

This section describes various invocations of the dtf program:

Running A Collection of Tests

By default, dtf will look for case modules in the case/ sub-directory of the current directory and test definition in the test/ sub-directory of the current directory, and run all available tests. In this mode, simply issue the following command:


You can change the default directory by passing the following options on the command line:

dtf --casedir buildscripts/dtf/ --testdir t/

Alternately, you can use the following short form options:

dtf -c buildscripts/dtf/ -t t/

Running A Single Test

Optionally, you can run a single test case. Toggle this mode by passing --single or -s to dtf. Consider the following invocations:

dtf --single --casedef buildscripts/dtf/ --yamltest t/equality_test0.yaml

In short form, you can express this as:

dtf -s -d buildscripts/dtf/ -y t/equality_test0.yaml

Additional Options

Combine any of the above dtf invocations with the following three behavioral options:

--vebose, -v

When set, dtf will output the result of all test validation and all passing and failing tests. Off by default.

--fatal, -f

When set, dtf will raise an exception (and terminate operation,) if a test fails. Useful for large test suites and rapid development cycles.

--passing, -p

If implemented, a passing method tests will return a passing document. See in the dtf source repository for an example of this implementation.

Writing Case Modules

All tests have a type field that specifies which case module to use when running the test. The name of your case module should correspond identically to this type declaration.

When running a test, dtf calls the the main() method from the case module, and passes two arguments: the name of the case (string) and the entire case as read from the YAML specification as a dictionary.

Strictly speaking, as long as main() runs a test using these archives, the implementation is not important. However, the DtfCase class provides a number of useful tools for writing cases. You can read the dtf source to get a better idea of the available tools. Consider the following example:

from cases import DtfCase

class DtfEquality(DtfCase):
    def test(self):
        if['value0'] ==['value1']:
            r = True
            msg = ('[%s]: "%s" %s successful! %s equals %s'
                   % (,['name'], 'equality test',['value0'],['value1']))
            r = False
            msg = ('[%s]: "%s" %s failed! %s does not equal %s'
                   % (,['name'], 'equality test',['value0'],['value1']))

        return r, msg

def main(name, case):
    c = DtfEquality(name, case)
    c.required_keys(['name', 'type', 'value0', 'value1'])

The best way to implement a test case is to subclass DtfCase and implement the test() method on this class. test() should return a tuple, that contains a boolean reflecting the test’s success or failure, and a message that dtf should return (if needed) regarding the test’s output. The above main() method:

  • creates an object (c) of the previously defined DtfEquality class, passing in the test name and object.
  • calling the required_keys() method to define what top level fields must appear in the test specifications.

While this example is typical, inside of the main, you may call whatever DtfCase methods you like. The default run() method resembles the following:

def run(self):
    self.validate(verbose=VERBOSE, fatal=FATAL)

    t = self.test()
    self.return_value = t[0]

    self.response(result=t[0], msg=t[1], verbose=VERBOSE, fatal=FATAL)

To summarize, run():

- validates the top level keys, as set by ``required_keys()``
  • calls the test() method, defined in the case itself.
  • passes the return value to the objects return_value instance variable and calls the response() method which returns messages as needed, based on the return value of test().

Specifying dtf Tests

The structure of a test definition depends largely on the case module. Define tests in YAML. For example, a basic equality/inequality test might resemble the following:

name: ‘example equality test’ type: equality value0: 1 value1: 1

The DtfCase.validate() method, at present, does not validate documents recursively.