themes and default styling

[chriddyp]

update is great. it'd be fun if we wrote some helper functions that allowed folks to save skeleton themed figures to their file system and apply them. maybe even update the default styling!

I just wrote up a lil notebook here that demonstrates how this might work: http://nbviewer.ipython.org/gist/chriddyp/f4adabf92a3b517a01f5

would love your feedback @theengineear @etpinard @aneda @cldougl

[theengineear]

is there still any plan to try and integrate themes such that you include a theme in the json to be applied only in the frontend? that way it would be shared amongst all the api-wrapping libraries. thoughts?

[theengineear]

i do like your proposal though, it certainly gets the job done!

[aneda]

@chriddyp I love it! should we incorporate this in the doc? I think it would be great to do all the galleries this way

[chriddyp]

@theengineear - I really like being able to just update the figure locally, and I think it allows some more flexibility if you want to apply a theme, and then override some its properties. If it was just in the front-end, then the theme would always have precedence.

I think that we could incorporate the front-end themes by making them an endpoint where you could download the ones that you've made in the front-end (at some point).

it might be cool to add a download_theme(url) method that would download the content at the endpoint, json parse it, plotly validate it, and then save it. Then, we could keep a public repo of all of our favorite themes for people to use, and as @aneda mentioned we could include them in the docs so that it's easy for people to remake their gorgeous graphs!

this solution is also cool because it would work with more complicated layouts, like subplots etc.

in fact, maybe it should be rebranded "templates"

[theengineear]

Yup, I'm with you on the ends, but I worry about the means. Applying a theme ended up being a non-trivial task in the frontend. I'm ok with doing this in our libraries, but it does mean lot of duplication.

If we did port this to python, we could always stick it in our backend and then make an api request to have our backend apply the theme and return the result. That could then be used across all our libraries.

Again, if it's worth the effort and we have team members to do it, I'm all for it. However, there's going to be a lot of duplication and then maintainability costs.

[jackparmer]

I'd love to see this done in 2 stages -

First, ship Plotly with a few useful default themes like cufflinks has done and add their usage to the docs:

https://github.com/santosjorge/cufflinks/blob/master/cufflinks/plotlytools.py#L58

This should be super easy. I'd personally love to see ggplot and seaborn themes... the number of lines of styling currently needed to match these themes is crazy.

Second, figure out how to save and reimplement saved themes.

[jackparmer]

Here are the themes that are new in MPL 1.4:
https://youtu.be/PmDUx62cTTs?t=6m12s

[jonmmease]

Closing in favor of #1161. This functionality is going to be built on top of the Plotly.js template mechanism.