There are many ways to contribute to dragonfly and the project would certainly benefit from more contributions.

Reporting bugs/other issues

If you come across bugs or other issues with dragonfly, you can open a new issue in the issue tracker.

If the issue is a bug, make sure you describe what you expected to happen as well as what actually happened. Include a full traceback if there was an exception.

Submitting patches/pull requests

If you have changes that can resolve an issue in the issue tracker, you can create a pull request to merge your changes with the master branch.

Documentation changes

There are parts of dragonfly that are undocumented as well as, undoubtedly, documented functionality which is poorly explained, out of date or incorrect. If you notice problems with the documentation, you can open a new issue in the issue tracker.

Dragonfly’s documentation is written in the reStructuredText format. ReStructuredText is similar to the Markdown format. If you are unfamiliar with the format, the reStructuredText primer might be a good starting point.

The Sphinx documentation engine and Read the Docs are used to generate documentation from the .txt files in the documentation/ folder. Docstrings in the source code are included in a semi-automatic way through use of the sphinx.ext.autodoc extension.

To build the documentation locally, install Sphinx and any other documentation requirements:

$ cd documentation
$ pip install -r requirements.txt

Then run the following command on Windows to build the documentation:

$ make.bat html

Or use the Makefile on other systems:

$ make html

If there were no errors during the build process, open the build/html/index.html file in a web browser. Make changes, rebuild the documentation and reload the doc page(s) in your browser as you go.

Improving spoken language support

Dragonfly supports using various languages with speech recognition engines. Language-specific code is located in sub-packages under dragonfly.language and loaded automatically when dragonfly.language is imported.

English is fully supported with mappings from English words to characters, integers (e.g. for IntegerRef) and time/date formats.

Other languages such as German and Dutch only have mappings for using IntegerRef (and similar) elements.

Additional SR engine backends

Dragonfly supports using various speech recognition engines: Dragon NaturallySpeaking (DNS), Windows Speech Recognition (WSR) and CMU Pocket Sphinx.

Adding an additional engine backend is a significant undertaking. Engine implementations should be placed in a sub-package of dragonfly.engines, e.g. backend_natlink. The engine’s sub-package should define an engine class implementing the EngineBase class and a get_engine function to be used in the dragonfly/engines/ file.

Examining the source code under dragonfly/engines/backend_text/ may be a good place to start, as it is currently the simplest engine implementation.

Implementations should define and use sub-classes of the DictationContainer, RecObsManagerBase and TimerManagerBase base classes from dragonfly.engines.base. These are for dictation result containers, managing recognition observers and managing multiplexing timers respectively.

A compiler class to translate dragonfly grammars into a format that the SR engine accepts will probably also be required. Engine backends other than the “text” engine have compiler classes to look at as examples.

If there are additional engine dependencies, they should be placed into the file in the extras_require dictionary. For example:

    "EngineX": ["engine_dependency >= X.Y.Z"],

This allows the engine’s dependencies to be installed using something like:

$ pip install dragonfly2[EngineX]

In addition to the engine implementation, each engine should define its own test suite in the dragonfly/test/ file. For example:

# Define test files to run for the new engine, including common ones.
x_names = [
    # Assume that "" exists. Including the '.py' file
    # extension is not necessary.

    # Include the tests for English integer content.
] + common_names

# Exclude one or more common names if the new engine doesn't [yet]
# support certain dragonfly functionality. Also display a warning.
_log.warning("Excluding 'test_timer' for engine X because multiplexing "
             "timers are not supported (yet).")

# Build a test suite for the engine.
x_suite = build_suite(EngineTestSuite("<engine_name>"), x_names)

The test suite should be runnable using the following (or similar) command:

$ python test --test-suite=dragonfly.test.suites.x_suite

The file should be able to build each engine’s test suite without engine-specific dependencies being available, such as natlink. You should be able to test this by running the default test suite in a virtual environment:

$ python test

The above command should run successfully for Python versions 2.7 and 3.x.

The new engine and its tests should be documented in the engines and test suites documentation pages respectively. If the engine implementation doesn’t work with some of dragonfly’s functionality, such as Dictation elements, it should be mentioned somewhere in the engine’s documentation.