A reactive notebook for Python — run reproducible experiments, query with SQL, execute as a script, deploy as an app, and version with git. Stored as pure Python. All in a modern, AI-native editor.
Python artificial-intelligencedagdata-sciencedata-visualizationdataflowdeveloper-toolsmachine-learningnotebookspipelinepythonreactivesqlweb-app
# /// script
# requires-python = ">=3.13"
# dependencies = [
# "jupytext",
# "nbformat",
# ]
#
# [tool.uv]
# exclude-newer = "2025-06-03T16:30:20.082913-04:00"
# ///
"""
Generate `.ipynb` test fixtures for marimo's notebook conversion pipeline.
Each fixture corresponds to a named Jupyter notebook containing one or more code cells.
To add a new fixture, call `create_notebook_fixture(name, sources)`:
- `name`: the output filename (without `.ipynb`)
- `sources`: a list of code cell contents (as strings or `nbformat` cell dicts)
Run this script with `uv run scripts/generate_ipynb_fixtures.py` to regenerate all fixtures.
Output notebooks are written to `tests/_convert/fixtures/`.
"""
from __future__ import annotations
from pathlib import Path
import nbformat.v4.nbbase as nb
import jupytext
SELF_DIR = Path(__file__).parent
FIXTURES_DIR = SELF_DIR / "../tests/_convert/fixtures"
def create_notebook_fixture(name: str, sources: list[str | dict]) -> None:
cells = []
for source in sources:
if isinstance(source, str):
cell = nb.new_code_cell(source)
else:
cell = source
cells.append(cell)
notebook = nb.new_notebook(cells=cells)
for i, cell in enumerate(notebook.cells):
cell.id = str(i) # ensure we always have 1,2,3,4
(FIXTURES_DIR / f"{name}.ipynb").write_text(
jupytext.writes(notebook, fmt="ipynb")
)
def main() -> None:
FIXTURES_DIR.mkdir(exist_ok=True)
create_notebook_fixture(
"multiple_definitions",
[
"x = 1\nprint(x) # print",
"x = 2\nprint(x) # print",
],
)
create_notebook_fixture(
"multiple_definitions_multiline",
[
"K = 2\nnearest_partition = np.argpartition(dist_sq_1, K + 1, axis=1)",
"""plt.scatter(X_1[:, 0], X_1[:, 1], s=100)
K = 2
for i_1 in range(X_1.shape[0]):
for j in nearest_partition[i_1, :K + 1]:
plt.plot(*zip(X_1[j], X_1[i_1]), color='black')\
""",
],
)
create_notebook_fixture(
"duplicate_definitions_and_aug_assign",
[
"x = 1",
"x",
"x += 1",
"x",
],
)
create_notebook_fixture(
"duplicate_definitions_read_before_write",
[
"x = 1",
"x",
"x; x = 2; x",
"x",
],
)
create_notebook_fixture(
"duplicate_definitions_syntax_error",
[
"x ( b 2 d & !",
"x",
],
)
create_notebook_fixture(
"cell_metadata",
[
nb.new_code_cell(
"print('Hello')", metadata={"tags": ["tag1", "tag2"]}
),
nb.new_code_cell("print('World')", metadata={}),
nb.new_code_cell(
"print('Cell 1')",
metadata={"tags": ["important", "data-processing"]},
),
nb.new_code_cell("print('Cell 2')", metadata={"tags": []}),
nb.new_code_cell(
"print('Cell 3')",
metadata={"tags": ["visualization"], "collapsed": True},
),
nb.new_code_cell(
"print('Complex metadata')",
metadata={
"tags": ["tag1", "tag2"],
"collapsed": True,
"scrolled": False,
"custom": {"key": "value"},
},
),
nb.new_code_cell(
"print('hidden cell')",
metadata={
"tags": ["hide-cell"],
},
),
nb.new_code_cell(
"print('hidden cell, with other tags')",
metadata={
"tags": ["hide-cell", "remove-print"],
},
),
],
)
create_notebook_fixture(
"hides_markdown_cells",
[
nb.new_markdown_cell("A markdown cell."),
nb.new_markdown_cell(
"A markdown cell with tags: ['blah'].",
metadata={"tags": ["blah"]},
),
nb.new_markdown_cell(
"A markdown cell with tags: ['blah', 'hide-cell'].",
metadata={"tags": ["blah", "hide-cell"]},
),
],
)
create_notebook_fixture(
"pip_commands",
[
"!pip install transformers",
"!pip install pandas numpy matplotlib",
"# Mixed cell with pip and other commands\n!pip install scikit-learn\nimport numpy as np\n!pip install seaborn",
"# Non-pip exclamation commands should remain unchanged\n!ls -la\n!echo 'Hello World'",
"# Magic pip command should also be handled\n%pip install requests",
],
)
if __name__ == "__main__":
main()