iRe.py

"Export" tables and plots from Jupyter notebooks, for interactive viewing on the web.

iRe/www is the accompanying web app (demo: ire.runsascoded.com), which renders notebooks exported by iRe.py as interactive "reports."

• Installation
• Usage
• Testing
  • Local
    • macOS
    • Ubuntu
  • CI
  • Updating tests

Installation

pip install ire.py

Usage

Wrap the final expression in a notebook cell in ire.export:

from ire import export
import pandas as pd

df = pd.DataFrame({'n': list(range(1000)), 'square': [n**2 for n in range(1000)]})

# Previews DataFrame as usual, in Jupyter, but also
# saves the full DF to a file (Parquet, by default).
export(df)

Supported types:

• pandas.DataFrame (export as Parquet or SQLite)
• plotly.Figure

See also: examples/

• export-test.ipynb
• export-test-no-dvc.ipynb

Testing

Tests execute the notebooks in examples/ above, and verify the notebooks and exported files are unchanged (vs. HEAD).

Local

macOS

To run tests locally:

examples/run-tests.sh

# Or, run+validate a specific notebook:
cd examples
run-tests.sh export-test.ipynb

Ubuntu

The exported SQLite tables in export-test.ipynb and export-test-no-dvc.ipynb are not byte-for-byte identical across platforms:

db_hex() {
  git show "$1:examples/.ire/state_stats2.db" | hexdump -C
}
diff <(db_hex HEAD) <(db_hex ubuntu)
# 7c7
# < 00000060  00 2e 6e ba 0d 00 00 00  02 0c e4 00 0d 39 0c e4  |..n�......�..9.�|
# ---
# > 00000060  00 2e 63 01 0d 00 00 00  02 0c e4 00 0d 39 0c e4  |..c.......�..9.�|

The ubuntu branch points at a commit that patches them to the expected values. You'll want to cherry-pick it into your branch, to get tests passing on Ubuntu (as CI does, below).

CI

See .gitlab-ci.yml.

This cherry-picks the ubuntu branch, to patch tests' expected SQLite DBs, before running them.

Updating tests

docker/build.sh
docker run --rm -it -v $PWD/.git:/root/src/.git -v $PWD/.dvc/cache:/root/src/.dvc/cache ire.py

See build.sh, Dockerfile, entrypoint.sh.