Contributing to Verto

Welcome to the Verto developer community! We have spent many months developing this project, and we would love for you to get involved! The following documentation has been written to help you get a grasp on how Verto is pieced together to make contributing as simple and straight forward as possible. Please feel free to fix bugs and/or suggest new features and improvements to the system (or the docs) by making a pull request.

Verto was created to be used by two much larger projects (the CS Unplugged and CS Field Guide websites) as the Markdown to HTML converter. The tags we chose are designed to allow authors of these two projects to easily write material without technical elements getting in the way. It is therefore important to us that Verto remains as simple and robust as possible, please keep this in mind if you decide to work on Verto with us.

The git repository for Verto can be found here, jump in and take a look around!


The two projects that Verto was developed for are Django projects, so you may come across HTML (in templates, test cases etc) that contains Django syntax.

For example, below is the expected output for a for a image tag test:

 <img alt="Lipsum" class="" src="{% static 'computer-studying-turing-test.png' %}"/>

This does not mean that Verto is only suitable for Django projects, as it’s just a matter of customising the relevant HTML templates.

Issue Reporting and Bug Fixes

If you come across a bug in Verto, please report it on the repo issue tracker.

If you choose to fix the bug for us, consider adding the relevant tests to the Test Suite (detailed further down this page) to help us catch any future bugs.

The Code Base

If you would like to contribute to Verto, create a fork of the repository.


Before reading this section, make sure you have read how to use (or even better, have already used Verto!).


There are a couple of terms we use when describing Verto to become familiar with:

  • Tag

    This refers to the custom markdown syntax that Verto processes.

    For example:

    {comment this will be removed by the converter}
    {image file-path="img/totally-real-image.png" alt="process me"}

    are examples of the comment and image tags in Verto.

  • Processor

    This refers to the class that is responsible for converting a specific tag. For example, RelativeLinkPattern is the processor for internal links.

Project Structure

Below is a basic overview of the project structure:

├── docs/
├── verto/
|   ├── errors/
│   ├── html-templates/
│   ├──
│   ├──
│   ├── processor-info.json
│   ├── processors/
│   ├── tests/
│   └── utils/
├── requirements.txt
├── requirements-dev.txt

The items of interest are:

  • Verto() - The convertor object itself. This is what a user will use to create a Verto converter, and what is used to define a custom processor list, custom html templates and custom Markdown Extensions to use.
  • VertoResult() (found in - The object returned by Verto() containing:
    • Converted html string
    • Title
    • Required files (images, interactives, scratch images, page scripts)
    • Heading tree
    • Required glossary terms
  • VertoExtension() - This is the main class of the project, and inherits the Extension class from Markdown. It loads all of the processor information, loads the template files and clears and populates the attributes to be returned by the VertoResult object.
  • processor-info.json - Every processor is listed in this file, and will at least contain a class determining whether it is custom or generic, where custom processors will have a pattern to match it’s corresponding tag. Most will also define required and optional parameters, these correspond to arguments in the tag’s html template.
  • processors/ - There is a different processor for each tag. A processor uses it’s corresponding description loaded from processor-info.json to find matches in the text, and uses the given arguments in the matched tag to populate and output it’s html template.
  • html-templates/ - The html templates (using the Jinja2 template engine) with variable arguments to be populated by processors.
  • errors/ - Contains all the errors exposed by the Verto module. Where an Error is an exception that is caused by user input. New errors should be created in here inheriting from the base Error class.
  • utils/ - Contains classes and methods not necessarily unique to Verto that are useful in any sub-module. This includes slugify handlers, html parsers and serialisers, and other utilities. The utilities should be used over external libraries as they are purposely built because of: compatibility reasons, licensing restrictions, and/or unavailability of require features.
  • tests/ - explained in the Test Suite section further down the page.

It is important to note that Verto is not just a Markdown Extension, it is a wrapper for Python Markdown. VertoExtension is an extension for Python Markdown. We have created a wrapper because we wanted to not only convert text, but also extract information from the text as it was being converted (recall VertoResult() listed above).

Creating a New Processor

There are two ways distinctly different ways to create a new processor. The simplist way is to make use of the provided generic processors and define the new processor in the processor-info.json file, while more complex processors reqiure additional source code. Complex processors should be considered when custom functionality is required that cannot be achieved with generic processors.

In all cases new processors should:

  • Be thoroughly tested (see the section on testing)
  • Have clear and accurate documentation. See the docs on other processors for the preferred format. Your docs should include:
    • An example of the tag in markdown
    • Required parameters
    • Optional parameters
    • Examples
    • Examples of overriding the html

We recommend writing documentation and test cases before you even write the processor itself as this will give you a clear idea of how a processor in Verto should behave.

Generic Processors

There are two types of generic processors:

  • tags (generic_tag): which match {<processor_name> <args>} in the markdown text replacing with the given html-template.
  • containers (generic_container): which are a pair of tags which capture the content between the tags for the html-template. A generic container’s opening tag specifies the arguments, while the closing tag only has the end argument allowing for the content to contain generic containers.

To create a new processor that uses the generic processors the processor must be added to the processor-info.json file and an associated html-template must be created. The template must only have one root level node after rendering.

How to make a JSON Definition

The json description of a generic processor must contain the attributes:

  • class: Either generic_tag or generic_container for a generic processor.
  • arguments: An object describing arguments passed to the tag.
  • template_parameters: An object describing template parameters.
  • (Optional) template_name: A custom name for the html-template to use. Defaults to the processor name otherwise.
  • (Optional) tag_argument: The text given at the beginning of a tag (e.g. the tag argument for {image file-path="example.png"} is image). This is only necessary for processors with different names sharing the same resources (e.g. both image-container and image-tag share the tag_argument).

The argument parameter is a dictionary (or object) containing argument name, argument-info pairs. Where the argument-info contains the attributes:

  • required: true if the argument must be set or false otherwise.
  • (Optional) dependencies: A list of argument-names that must also be set if this argument is used.
  • (Optional) values: A list of values of which the argument may take. If the argument does not have the value in this list the ArgumentValueError exception is raised.

These arguments are transformed for use in the html-template by the template_parameters attribute. This attribute is similar to the argument attribute by containing parameter name, parameter-info pairs. Where the parameter-info contains the attributes:

  • argument: The name of the argument to retrieve the value of to use/transform into the parameter value.
  • (Optional) default: The value the parameter defaults to if the argument is not given otherwise defaults to None.
  • (Optional) transform: The name of the transform to modify the argument value by or defaults to null for no transformation. The avaliable transforms are detailed below.
  • (Optional) transform_condition: A function that takes the context after parameters are set but before transformation (The transformations are done in order they appear in the json document). If the function returns True then the transformation is applied.

For a generic container type processor the argument of the parameter may be content which is the captured content between the start and end tags.

The set of currently avaliable transformations for the transform attribute are:

  • str.lower: Converts the string into a lowercase version.
  • str.upper: Converts the string into an UPPERCASE version.
  • relative_file_link: Applies the relative-file-link html-template to the argument.

A generic tag processor, is a simple single line tag that uses the given arguments as parameters to an html template. An example of a processor that uses the generic tag processor is the button-link processor which is described in the json as:

"button-link": {
  "class": "generic_tag",
  "arguments": {
    "link": {
      "required": true,
      "dependencies": []
    "text": {
      "required": true,
      "dependencies": []
    "file": {
      "required": false,
      "dependencies": []
  "template_parameters": {
    "file": {
      "argument": "file",
      "transform": "str.lower",
      "default": "no"
    "link": {
      "argument": "link",
      "transform": "relative_file_link",
      "transform_condition": "lambda context: context['file'] == 'yes'"
    "text": {
      "argument": "text",
      "transform": null

And has the following html-template:

<a class='button' href='{{ link }}'>{{ text }}</a>

This enables the following markdown:

{button-link link="" text="Visit Google"}

To generate the output:

<a class="button" href="">Visit Google</a>

A generic container processor, a pair of matching tags where one opens the container and one closes the container. The start tag gives the arguments as parameters to an html template. The end tag is used to capture the content between the tags to be used as an additional parameter to the html template. An example of a processor that uses the generic container processor is the boxed-text processor which is described in the json as:

"boxed-text": {
    "class": "generic_container",
    "arguments": {
      "indented": {
        "required": false,
        "dependencies": []
    "template_name": "boxed-text",
    "template_parameters": {
      "indented": {
        "argument": "indented",
        "transform": "str.lower"
      "text": {
        "argument": "content",
        "transform": null

And has the following html-template:

<div class='boxed-text{% if indented == "yes" %} boxed-text-indented{% endif %}{% if type %} boxed-text-{{ type }}{% endif %}'>
{% autoescape false -%}
{{ text }}
{% endautoescape -%}

This enables the following markdown:


**Computer Science Report for 2.44**

Put your introduction to what bits are here.

{boxed-text end}

To generate the output:

<div class="boxed-text">
<p><strong>Computer Science Report for 2.44</strong></p>
<p>Put your introduction to what bits are here.</p>

Custom Processors

To create a custom processor, the class attribute of the processor in the processor-info.json file must be "custom". A good place to start when programming a new processor is the Extension API page of the Python Markdown docs, and you can also read the source code itself.

There are several different kinds of processors in Python Markdown, each serving a slightly different purpose. We recommend reading the API docs to determine which processor best suits your purpose. Verto currently makes use of preprocessor, blockprocessor, inlinepattern, treeprocessor and postprocessor, but you are welcome to use another type of processor if it better suits the task.

The order of the processors matters and is defined when each processor is added to the OrderedDict in

Each processor should try to be as independent of every other processor as possible. Sometimes this is not possible, and in this case compatibility should occur in the processor that happens last (i.e. the downstream processor). That is output should be consistent based on input, not the other way round (e.g. codehilite and fenced_code).

The logic for each processor belongs in the processors/ directory, and there are several other places where processors details need to be listed. These are:

  • The processor’s relevant information (regex pattern, required parameters etc) should be included in processor-info.json.
  • If it should be a default processor, it should be added to the frozenset of DEFAULT_PROCESSORS in
  • The relevant list in extendMarkdown() in (see OrderedDict in the Markdown API docs for manipulating processor order).
  • The processor’s template should be added to html-templates using the Jinja2 template engine syntax for variable parameters. A valid template will only have one root level node after rendering, if more root nodes are necessary the remove tag can be used as the root node which will be removed later.
  • Any errors should have appropriate classes in the errors\ directory, they should be well described by their class name such that for an expert knows immediately what to do to resolve the issue, otherwise a message should be used to describe the exact causation of the error for a novice.

The Test Suite

To start the test suite:

$ python3 -m verto.tests.start_tests

This will execute the Smoke, System and then Unit tests.

There are several arguments that can be used with this command to skip particular tests (--no_smoke, --no_system and --no_unit).

Test Suite Structure

We are now focusing on our project structure diagram from earlier:

└── verto/
    └── tests/
        ├── assets/

The items of interest are:

  • BaseTest()) - This class is inherited by nearly every other test file, and contains a method to read a given test asset file.
  • ConfigurationTest() - This is the test class for testing different configurations of Verto() (e.g. using a custom list of processors and/or custom html templates). This class inherits the BaseTest class.
  • - This is the class inherited by all processor test classes. It contains several useful methods for testing processors, including those for loading templates and processor info.
  • SmokeDocsTest() and SmokeFileTest() - These are the two classes for smoke testing.
  • - This is the file that is executed in order to run each of the three types of tests (Smoke, System and Unit). Every new test class must be added to the relevant section of this file.
  • assets/ - This directory contains a sub directory for every test class that loads external assets (e.g. test input files).

Adding Tests

When writing a new test function, it is important that the method name is as descriptive as possible. The method name should also be prefixed with test_ as the test suite will only execute methods with this prefix.

If you have added a new processor to Verto, then a corresponding test suite also needs to be added. This test suite should be added to the unit_suite() function in The section below has details on how to write a processor test.

Processor Tests

All processor tests inherit from the ProcessorTest class. Processors should create a Mock() object, which will contain the bare minimum for the processor to be run (it’s HTML template and properties loaded from processor-info.json), i.e. there is no reason for it to know about properties of the other processors.

A test method will typically follow the same sequence of steps:

  1. Retrieve the test string (there is a read_test_file() method provided by the ProcessorTest class)
  2. Confirm there are (not) matches to the regex in the test string
  3. Convert the test string using the verto_extension (provided by the SetUp() method in ProcessorTest)
  4. Load the expected converted result
  5. Check the converted result is the same as the expected result

Testing Assets

Most tests will load an asset file. This file contains example Markdown text (and therefore has a .md extension). For comparing the converted result of this Markdown file with it’s expected output, a corresponding “expected” file should be created. The expected file should have the same name as the corresponding test file, with expected appended to the file name (and has a .html extension).

These asset files should be placed in verto/tests/assets/<processor-name>/.

For example:



  • Asset files should have discriptive names, and in many cases will have the same name as the method they are used in.

Creating a release

This is our current process for creating and publishing a Verto release. This can only be performed by repository administrators.

  1. Create a release branch. Checkout to this branch.
$ git checkout develop
$ git pull
$ git checkout -b release/0.7.0
  1. Update the version number [1] within verto/
  2. Check test suite for errors, and fix any issues that arise, or log an issue.
  3. Detail the changes in docs/source/changelog.rst.
  4. Complete the release branch. Be sure to tag the release with the version number for creating the release on GitHub.
$ git checkout master
$ git pull
$ git merge --no-ff release/0.7.0
$ git tag -a 0.7.0
$ git push
$ git push --tags
  1. Upload the release to PyPI
$ rm -r dist/
$ python3 -m pip install --user --upgrade setuptools wheel
$ python3 sdist bdist_wheel
$ python3 -m pip install --user --upgrade twine
$ twine upload dist/*
  1. Merge the release into develop
$ git checkout develop
$ git pull
$ git merge --no-ff release/0.7.0
  1. Create the release on GitHub on the tagged commit.
  2. Delete the release branch.

Other notes

Why are requirements split across two files?

The requirements.txt file is used to specify required dependencies for Verto (and are automatically installed as dependencies when installed via pip). The requirements-dev.txt file is used to specify dependencies for developing Verto (for example, documentation generators).

Why are all dependency versions pinned?

Yes, this is not considered the best practice by the Python Software Foundation in their packaging guide. However pinning dependencies ensure we control over each Verto release, following the logic of Vincent Driessen.

[1]We follow Semantic Versioning for our numbering system. The number is used by to tell PyPI which version is being uploaded or pip which version is installed, and also used during the documentation build to number the version of Verto it was built from.