make generated plot HTML reproducible

The HTML/js contains UUIDs generated by python's built-in uuid.uuid4. By nature, those are unique and totally random.

But if you generate documents as part of a pipeline (e.g. via https://dvc.org/ ) then it would be great to have them reproducible down to the byte.

import plotly.express as px

fig =px.scatter(x=range(10), y=range(10))
fig.write_html("file1.html")
fig.write_html("file2.html")

The files will be identical except from the UUIDs.

I "solved" this problem by monkey-patching the relevant function:

# monkey patch
import uuid
import random

uuid4_rnd = random.Random(0)
def uuid4_seeded():
    """Generate a random UUID using random.Random"""
    return uuid.UUID(bytes=uuid4_rnd.randbytes(16), version=4)

uuid.uuid4 = uuid4_seeded

fig =px.scatter(x=range(10), y=range(10))
fig.write_html("file.html")

re-executing this snippet will produce identical files.

It would be great to exchange one function (or set a seed for an internal version of uuid) in plotly instead of hijacking the python built-in.

Commencez par le point d’entrée fig.write_html et suivez les endroits où les UUID sont insérés dans le HTML/JavaScript généré. Comparez plusieurs écritures de la figure d’exemple et déterminez comment Plotly devrait exposer la génération déterministe des UUID ou l’initialisation d’une seed. C’est terminé lorsque des fichiers HTML identiques octet par octet peuvent être générés sans faire de monkey-patching de uuid.uuid4 de Python.