simplify deprecation message for altair.themes?

[mattijn]

What is your suggestion?

In the deprecation message of altair.themes we see currently among other things:

altair/altair/__init__.py

Line 681 in c41e29e

" def custom_theme() -> alt.theme.ThemeConfig:\n"

    # Updated
    @alt.theme.register('theme_name', enable=True)
    def custom_theme() -> alt.theme.ThemeConfig:
        return {'height': 400, 'width': 700}

I'm wondering if we really need to add -> alt.theme.ThemeConfig when defining this function. Within a Jupyter Notebook I really rarely see this approach being utilised and in combination with the usage of a @ decorator to register the theme, also something I really rarely see people doing a Jupyter Notebook, this deprecation message feels a bit overwhelming to me.

Have you considered any alternative solutions?

@dangotbanned, if it's not too much trouble, I prefer making one change at a time as a user. Can we just do this?

# Updated
@alt.theme.register('theme_name', enable=True)
def custom_theme():
    return {'height': 400, 'width': 700}

[dangotbanned]

@mattijn I do understand the concern.

One thing I'd add though is that the annotation is what provides the static typing feedback.

So even if you don't understand typing, if you copy and paste the code - your IDE can let you know if something is wrong.

Original

import altair as alt

@alt.theme.register('theme_name', enable=True)
def custom_theme() -> alt.theme.ThemeConfig:
    return {'height': 400, 'width': 700}

Alternatives

If we just want to shorten it we could say:

from altair import theme

@theme.register('theme_name', enable=True)
def custom_theme() -> theme.ThemeConfig:
    return {'height': 400, 'width': 700}

or

from altair.theme import register, ThemeConfig

@register('theme_name', enable=True)
def custom_theme() -> ThemeConfig:
    return {'height': 400, 'width': 700}

Both of these would be a shorter message - while not limiting the benefits of the original

[mattijn]

I'm in JupyterLab and I don't get any feedback🤷‍♂️

[dangotbanned]

I'm in JupyterLab and I don't get any feedback🤷‍♂️

I think we'll need to figure out a good suggestion for that case as part of #3645

Non-typing users

• Class-syntax is required for completions
• Still likely an improvement over using a dict
• Point to resources on how they can get the full benefits
  • ipython/jupyter
    • Language servers/plugins/extensions
  • IDE
    • typing-typing-related-tools
    • I'd probably call out VSCode w/ pylance as a common case

For example https://github.com/jupyter-lsp/jupyterlab-lsp

Edit: Looks more complicated than adding an extension in VSCode

[mattijn]

Can we do this?

import altair as alt

@alt.theme.register('theme_name', enable=True)
def custom_theme():
    return alt.theme.ThemeConfig(
        {'height': 400, 'width': 700}
    )

Still a single import altair as alt import and also just a dictionary as input of the alt.theme.ThemeConfig, an object that you have when facing this message.

[dangotbanned]

#3694 (comment)

Yeah @mattijn that works for me

You can also use keyword args if you're using ThemeConfig directly.
Not sure if that would be preferred but thought I'd mention as an option

[mattijn]

Once JupyterLab and Jupyter Notebook (classic) have proper language service support, these types of changes will be easier to implement. However, we're not there yet. Thanks for helping to find an intermediate solution!