groktest 0.3

Literate testing framework based on Python doctest

---

test-type: doctest

Groktest

Groktest is a test framework that's inspired by Python's long standing doctest library.

Similarities to doctest:

• A test suite is a plain text file containing more or more tests
• A test is an expression accompanied by an expected result
• A test is run by evaluating the expression in the applicable language environment to generate a computed result
• A test passes when its computed result matches is expected result and fails otherwise

Differences to doctest:

• Test expressions may be defined in any supported language (currently Python, bash planned, support for other languages over time)
• Expected results may contain format-expressions, which are used to match expected output to a pattern with optional value capture to runtime state variables (see examples below)
• Tests are not sensitive to white-space by default
• Test options do not require a doctest: prefix
• Front-matter in test files may be used to configure a test suite

Why?

Python's doctest paved the way for computational documents like Jupyter Notebooks, R Markdown, and Quarto. It was one of the first of its kind. We owe a debt of gratitude to Tim Peters for the original work and to Jim Fulton for advancing its use.

The rationale for doctest is that examples provide a high-signal, low-noise representation of developer intent.

• Tests are framed within comments (rather than using comments), which encourages a narrative mode of testing
• Examples are often easier to understand (grok) than a series of assertions
• Example output can replace a large number of explicit assertions, providing more test coverage with fewer expressions

doctest however presents some key problems.

• Result comparison can use an ellipsis ... to match any output; while the match is greedy, it can match invalid output and mast errors
• The framework is Python specific and difficult to apply to cases outside Python
• Pattern matching is limited to a single 'match all' pattern
• Output from tests cannot be captured for use in subsequent tests

Groktest addresses these issues while maintaining the original doctest value proposition.

Not supported in Groktest

Groktest is in its early development phases. Its immediate goal is to be a near drop-in replacement for doctest for tests defined in standalone plain text files.

Groktest is not designed to test Python doc strings. Use doctest for that.

While Groktest is designed to be language independent, it will take some time to get this right. In order of priority, Groktest will support:

1. Tests defined using Python expressions (under development)
2. Tests defined as shell commands with bash/sh as the first target (planned)
3. Other runtimes as the project evolves (e.g R, Node, Julia, Erlang, etc.)

Examples

The simplest example is an exact match.

>>> 'The number 42 is overused in examples'
'The number 42 is overused in examples'

In groktest you can provide a format expression, which has the format {[name][:format]}. The following example asserts that the claim contains a series of digits.

>>> 'The number 42 is overused in examples'  # doctest: +SKIP
'The number {:d} is overused in examples'

Matched values can be bound to local variables.

>>> 'The number 42 is overused in examples'  # doctest: +SKIP
'The number {n} is overused in examples'

>>> n == 42  # doctest: +SKIP
True

Output parsing is provided by Richard Jones' outstanding parse library.

To match anything, use {}.

>>> 'Another tedious example is the use of "ham" and "spam"'
... # doctest: +SKIP
'Another tedious example is {}'

Multiline expressions are continued on subsequent lines using ....

>>> (1 +
... 2 + 4)
7

Front matter is used to configure tests.

---
test-config:
  ps1: '> '
  ps2: '+ '
---

> "This should be familiar
+ to R developers"

Aside: In this case the use of > conflicts with Markdown's syntax for block quotes. We recommend indenting tests with four spaces in Markdown files.

Custom format types may be defined in test font matter.

---
test-config:
  match-types:
    id: [a-f0-9]{8}
---

>>> 'Sample id is abcd1234'  # doctest: +SKIP
Sample id is {:id}

What works?

A this point Groktest is in early development and nothing much of use is available.

Refer to Groktest tests to see what works. Run them to verify!

$ groktest *.md docs/*.md

Contributing

Please feel free to reach out if you'd like to contribute to this project. While we don't have a contribution policy we will get there! In the meantime, use the following guidelines:

1. Use the project issue tracker to ask questions, report bugs, and suggest features.

2. Contributions will be accepted via pull requests.

3. All tests in tests.md should pass.

4. All source code is formatted using black - expect any contribution to be reformatted on this basis.

5. Before working on a contribution, we recommend opening an issue to get early feedback.

6. We expect contributors to follow generally accepted standards of respect and kindness to others.