Add a cheatsheet

[cigrainger]

I'm excited about cheatsheets and something like this would beat "Ten minutes to Explorer", especially for those coming from dplyr or pandas who just need an easy reference.

dplyr: https://nyu-cdsc.github.io/learningr/assets/data-transformation.pdf
also dplyr: https://www.rstudio.com/wp-content/uploads/2015/02/data-wrangling-cheatsheet.pdf
pandas: https://pandas.pydata.org/Pandas_Cheat_Sheet.pdf

I started playing over the weekend but I think the diagrams are really powerful and got myself stuck trying to figure out how to replicate them in vega lite.

[billylanchantin]

I'd love to see cheatsheets too! Personally, I think it would've saved me some documentation hunting.

For example, early on I saw there was a DataFrame.filter/2 but I couldn't find a corresponding Series.filter/2. I eventually found Series.mask, but it wasn't obvious. Something like a task-oriented cheatsheet would've made that much easier to find.

I think the diagrams are really powerful and got myself stuck trying to figure out how to replicate them in vega lite.

I also feel like I've lost a good bit of time trying to get vegalite to output specific visualizations. It's great for a lot of cases. But when you start doing non-data-driven things like annotating your visualization, I found that it gets tricky.

Were you thinking of doing a .cheatmd? Or a .pdf?

[cigrainger]

Glad to hear it would be helpful! I was thinking of doing a .cheatmd so we can easily put it into the docs. I feel like it should be possible to get close with heatmaps based on specific categorical values. And I also agree that it's confusing we have Series.mask/2 instead of Series.filter/2.

@philss or @josevalim I'm sure there was a convo about this but I can't remember what the reasoning was for this anymore. #326 (comment)

[josevalim]

We changed the implementation and renamed at the same time but I am fine with reverting the name back to filter. :) It should be a quick change and we can add:

@deprecated "Use Explorer.Series.filter/2 instead"
def mask(s1, s2), do: filter(s1, s2) 

It will certainly be much easier to find.

[billylanchantin]

Oh I didn't mean to pick up a stray issue! I was just using it as an example.

We changed the implementation and renamed at the same time but I am fine with reverting the name back to filter.

It may be worth having both since mask/2 and filter/2 accept different datatypes: mask takes a boolean series while filter/2/filter_with/2 take a query/function. If you have the boolean series on hand, you'd want mask/2. But if you're finding that you need to build the boolean series e.g. with transform/2 only to pass it right into mask/2, filter/2 would be convenient.

[billylanchantin]

I feel like it should be possible to get close with heatmaps based on specific categorical values.

I agree! I was more worried about the arrows:

Though if you could embed the diagrams in a table, you could probably achieve a similar effect w/o the need for the arrow annotations.

[josevalim]

It may be worth having both since mask/2 and filter/2 accept different datatypes: mask takes a boolean series while filter/2/filter_with/2 take a query/function.

The issue is that doing it with a function is horribly expensive and should be generally avoided.

[billylanchantin]

Hey I wrote this a little after the earlier discussion. If it's not helpful just ignore me :)

https://vega.github.io/editor/#/gist/e0675e1408ba1944deb1a747f03a060d/spec.json

DPLYR	VegaLite

Note that it does appear to be possible to add margins to the rectangles:

https://vega.github.io/vega-lite/examples/rect_mosaic_labelled_with_offset.html

But my cursory reading of that example makes it seem a bit complex:

    {
      "calculate": "datum.y + (datum.rank_Cylinders - 1) * datum.distinct_Cylinders * 0.01 / 3",
      "as": "ny"
    },

[cigrainger]

Super helpful! Thank you @billylanchantin! I also don't think the margins are too important :).