Add optional download requirements for dev and plotly express

[ndrezn]

Closes #2279

This PR:

• Adds the install shortcut plotly[express], explained in more detail below
• Renames the following files, to match conventions used in other Plotly repos:
  • requirements.txt to requires-install.txt
  • optional-requirements.txt to requires-optional.txt
• Adds an additional file requires-express.txt, which contains the dependencies required by Plotly Express which are not required by the base Plotly package (currently, this is just numpy)
• Adds additional detail to the error message displayed when a user tries to import Plotly Express without having numpy installed
• Does a very light round of cleanup on requires-optional.txt, removing outdated dependencies which are definitely not needed anymore

---

This PR adds the ability to install Plotly with:

pip install plotly[express]

which bundles numpy in the install, which is required by plotly.express,

This also revises the error message raised when using plotly.express if numpy is not installed, providing further instructions on how to install.

Open to ideas on better wording.

(plotly.py) ➜  plotly.py git:(better-pandas-warning) python
Python 3.12.3 (main, Apr  9 2024, 08:09:14) [Clang 15.0.0 (clang-1500.3.9.4)] on darwin
Type "help", "copyright", "credits" or "license" for more information.
>>> import plotly.express
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/Users/nathandrezner/Plotly/plotly.py/packages/python/plotly/plotly/express/__init__.py", line 10, in <module>
    raise ImportError(
ImportError: Plotly Express requires numpy to be installed. You can install numpy using pip with:

$ pip install numpy

Or install Plotly Express and its dependencies directly with:

$ pip install "plotly[express]"

You can also use Plotly Graph Objects to create a large number of charts without installing
numpy. See examples here: https://plotly.com/python/graph-objects/

[ndrezn]

We can plausibly drop the note about graph objects but basically I want some sort of explanation of why pandas isn't a dependency.

[emilykl]

We shouldn't recommend pip install plotly.express , I believe it only exists for backwards compatibility and it's confusing to have multiple install paths.

If we want to have a one-liner "install everything" command that should be a change made inside Plotly.py.

[ndrezn]

I was thinking of adding:

pip install plotly[express]

as an option which goes with usual pip patterns for optional dependencies. If we went that route I would expect that we drop support for pip install plotly.express entirely.

[gvwilson]

I assumed from the PR title that this was going to improve the install-time or runtime error messages for "you need Pandas" but there seem to be other changes as well? I trust you...

[ndrezn]

@gvwilson there are other changes following @emilykl 's comment. It might make sense to break this into two PRs to be honest...

[gvwilson]

👍 on breaking the PR - thank you

[emilykl]

IMO the naming should be plotly[pandas] rather than plotly[express] -- much more clear what it's actually doing.

I'm in favor of this change, but let's not merge it right away. This PR changes a number of lines in setup.py so it's not just an addition; it could feasibly break the install process for some users. Need to do really solid QA to make sure it runs smoothly.

[ndrezn]

@gvwilson on further thought I think it makes sense to keep these changes in one PR. The text in the warning message is directly related to the installation recommendations that are also added in this PR -- I've updated the title & description to make the primary change more clear.

[gvwilson]

👍

[ndrezn]

@emilykl is this PR good to go with the latest changes?

[gvwilson]

See also #4790

[marthacryan]

@ndrezn So is pip install plotly[pandas] exactly the same as pip install pandas?

[emilykl]

@marthacryan pip install plotly[pandas] would be roughly equivalent to pip install plotly && pip install pandas I believe.

[marthacryan]

Oh interesting. What is the benefit here over just using pip install plotly pandas?

[ndrezn]

The benefit is we recommend a specific version number for compatibility. I don't think we have issues with different versions of Pandas but this makes it easy to pin a version of Pandas for use with Plotly if we ever encounter issues.

In fact we're doing that here, where we install the latest pandas for new versions of Python but force older pandas for older Python. People in the original issue were having funky errors caused by this incompatibility: #2279 (comment)

[gvwilson]

conversation with @ndrezn he believes this should go in before #4790 (Narwhals support) so moving it to the 3.0 release cycle.

[marthacryan]

The benefit is we recommend a specific version number for compatibility. I don't think we have issues with different versions of Pandas but this makes it easy to pin a version of Pandas for use with Plotly if we ever encounter issues.

In fact we're doing that here, where we install the latest pandas for new versions of Python but force older pandas for older Python. People in the original issue were having funky errors caused by this incompatibility: #2279 (comment)

Oh ok that makes sense! Maybe for documenting the usage of that we could make that more clear?

[LiamConnors]

Is this step required in contributing.md? I think it's covered by the editable install of plotly further down.

[gvwilson]

@ndrezn @emilykl if you can spare time today, it would be great to get this one in - thanks

[emilykl]

good catch @LiamConnors , I'll remove this

(I verified that the setup instructions still work without this step, at least on Mac)

[LiamConnors]

Looks good to me. Could you add a changelog entry for it too.