Continuing examples

[TiemenSch]

Would you consider supporting "continuing" examples that a feature text in between code fences?

For example:

foo = "bar"

We set foo to "bar" to indicate bar. We can then print "Quux!"

if foo == "bar":
    print("Quux!")
#> "Quux!"

Currently, I'm trying some sort of hybrid setup with pytest-markdown-docs that supports it, but that more or less has to run every example twice. Also their ```python continuation thing doesn't quite render right in MkDocs.

[TiemenSch]

This might make the formatters and linters quite sad, though.

[samuelcolvin]

Happy to accept a change to fix this if it's not too complicated. How would you implement a fix?

[rmorshea]

The first thing you'd probably need is to add in the ability for EvalExample to preserve execution state each time run() is called. Once that's done a bunch of possibilities open up. The first one is execution all examples in the same file together.

import itertools

import pytest
from pytest_examples import CodeExample
from pytest_examples import EvalExample
from pytest_examples import find_examples

GROUPED_EXAMPLES = []
for _, group in itertools.groupby(find_examples("path/to/examples"), lambda ex: ex.path):
    GROUPED_EXAMPLES.append(list(group))


@pytest.mark.parametrize("examples", GROUPED_EXAMPLES, ids=str)
def test_grouped_examples(examples: list[CodeExample], eval_example: EvalExample):
    for ex in examples:
        eval_example.run(ex)

You can also get clever with CodeExample.prefix_settings() and start grouping examples:

```python { test_group="1" }
x = 1
```

Some text

```python { test_group="1" }
assert x == 1
```

Where you could then gather only the specific examples that need to be run together:

import itertools

import pytest
from pytest_examples import CodeExample
from pytest_examples import EvalExample
from pytest_examples import find_examples

GROUPED_EXAMPLES = []
UNGROUPED_EXAMPLES = []
for group_id, group in itertools.groupby(
    find_examples("path/to/examples"),
    lambda ex: ex.prefix_settings().get("test_group"),
):
    if group_id:
        GROUPED_EXAMPLES.append(list(group))
    else:
        UNGROUPED_EXAMPLES.extend(group)


@pytest.mark.parametrize("examples", GROUPED_EXAMPLES, ids=str)
def test_grouped_examples(examples: list[CodeExample], eval_example: EvalExample):
    for ex in examples:
        eval_example.run(ex)


@pytest.mark.parametrize("example", UNGROUPED_EXAMPLES, ids=str)
def test_ungrouped_examples(example: CodeExample, eval_example: EvalExample):
    eval_example.run(example)

[rmorshea]

For reference I'm already using this strategy to skip testing certain docstrings by marking them with { test="false" }.