execnb

Install

Either:

pip install execnb

or if you use conda:

conda install -c fastai execnb

(You can replace conda with mamba in the line above if you have mamba installed.)

How to use

Use CaptureShell to run Jupyter code and capture notebook outputs, without running a Jupyter server (or even having it installed). The API is async, so await the run from a notebook or async code, or wrap it in asyncio.run at a sync entry point:

from execnb.shell import *
from fastcore.utils import *
from fastcore.nbio import *

s = CaptureShell()
await s.run('1+1')

[{'data': {'text/plain': ['2']},
  'metadata': {},
  'output_type': 'execute_result',
  'execution_count': None}]

To execute a notebook and save it with outputs filled in, use CaptureShell.execute:

try:
    await s.execute('../tests/clean.ipynb', 'tmp.ipynb')
    print(read_nb('tmp.ipynb').cells[1].outputs)
finally: Path('tmp.ipynb').unlink()

[{'name': 'stdout', 'output_type': 'stream', 'text': '1\n'}, {'data': {'text/plain': '2'}, 'execution_count': None, 'metadata': {}, 'output_type': 'execute_result'}]

You can also execute notebooks from the command line with exec_nb:

!exec_nb --help

usage: exec_nb [-h] [--dest DEST] [--exc_stop] [--inject_code INJECT_CODE]
               [--inject_path INJECT_PATH] [--inject_idx INJECT_IDX] [--verbose]
               src

Execute notebook from `src` and save with outputs to `dest`

positional arguments:
  src                        Notebook path to read from

options:
  -h, --help                 show this help message and exit
  --dest DEST                Notebook path to write to (default: )
  --exc_stop                 Stop on exceptions? (default: False)
  --inject_code INJECT_CODE  Code to inject into a cell
  --inject_path INJECT_PATH  Path to file containing code to inject into a cell
  --inject_idx INJECT_IDX    Cell to replace with `inject_code` (default: 0)
  --verbose                  Show stdout/stderr during execution (default:
                             False)