data-visualization | configurable data visualization | Data Visualization library

For posterity and for anyone who also struggled in piecing together how to do stuff on the canvas, here is an implementation following what @rob mayoff said by drawing everything on the canvas.

bokeh ditched their high-level chart interface quite some time ago because it was too much to maintain. Instead they adopted a different package that provided a high level charting interface called Holoviews

Holoviews works with many different renderers, but bokeh is the default, so you can have Holoviews create your plot objects, and then use bokeh to tweak/fine tune them.

https://holoviews.org/gallery/demos/bokeh/boxplot_chart.html#demos-bokeh-gallery-boxplot-chart.

You may use ellipse.alpha argument that can find here.

To plot sentiment over time you need a date column and a sentiment column. Then you can count the sentiment by date with count(sentiment, date) and then you can plot date along the x axis, n up the y axis, and fill by sentiment.

If you want stacked bar charts, remove position = "dodge" from geom_col()

To combine legends for two scales you must do two things:

1. Use guide_legend() as the guide for both scales (already default for size).
2. Ensure all name, breaks, labels, limits, etc. parameters match between the scales.

To use discrete colours from the viridis package for a continuous scale you can use the binned colour/fill scale scale_{colour/fill}_viridis_b().

A way to solve this problem is to use the set_xlim function from Matplotlib and apply it to your subplots. The code then looks like this:

The parameter facecolor is responsible for the background color. You can remove it to have white background.

Use set_window_title to add a new title to the window.

Modified code looks like this:

You can use the coefficients from the lda model to generate a plane separating the discriminant volumes. Effectively, the plane is the set of points in the 3D space where the sum of the (x, y, z) co-ordinates multiplied by their respective coefficients from the model is equal to the model's threshold (i.e. the plane where the model can't discriminate one group from the other).

We can do this by creating a 10 x 10 grid of equally spaced values along the x and y axes and calculating the z value that gives us the threshold value based on the model:

You should have two functions in flask - with two different URLs.

First function with url / should send only template - and this URL without limit=100&offset=100 you should load in browser.

Second function with url ie. /get_data should send JSON data - and Vue.http.get() should use it with limit=100&offset=100 like /get_data?limit=100&offset=100

If you want to do it with single url then you should use some if/else to detect if page is loaded by browser and send HTML - or by Vue and send JSON. Maybe you could use some HTTP header with informat that you expecte response with JSON. Accept: application/json

Full working code with two functions