New settings for custom kwargs

[fdion]

Overview

• Added settings for render_mode and title_format. These can be passed to show() through settings.
• the settings have default values (defined in constants.py)
• they were added to SettingsStore
• Added derived identifiers for building title
• Added utility to create list of active filter expressions
• Added utility to remove units in parenthesis (for title - axis labels not affected)
• Adds generated title to kwargs, works on all charts
• In the process of testing, found and fixed:
  • wordcloud didn't pass title to imshow
  • histogram was missing y (comes into play with orientation / histfunc etc)
  • pie and scatter_matrix missing color
  • scatter_matrix missing symbol and size
• even with default title, it can be overwritten with custom kwargs
• it is also possible to add to default title by using {title} in the title custom kwarg (ie. Chart 6: {title}

Note, default title format is:

"{name}: {title_columns}{title_dimensions}{names}{title_y}{title_z}{over_by}{title_x} {selection}<br>"\
"<sub>{groupings}{filters} {title_trendline}</sub>"

Definition of the fields (also in utility.py comment):

Besides all dragger selections (x, y, z, color etc), and all custom kwargs, extra template variables are:

          date: current datetime
          filters: active filter expressions, comma delimited
          title_x: x minus units and with log scale if selected
          title_y: y minus units and with log scale if selected
          title_z: z minus units and with log scale if selected
          title_dimensions: dimensions list minus units
          title_columns: columns list minus units
          title_trendline: trendline description
          vs: when doing a title with x vs y, use {x}{vs}{y}
          over_by: when doing a title y over x, use {y}{ober_by}{x}. Preferred. for distributions, will use "by"
          name: dataframe name
          total: total number of observations
          subset: observations with active filters applied
          selection: ready to use string representing {subset} observations of {total}
          groupings: groupings tied to Legend and not on legend: marker_symbol, line_group, size etc

The title can also contain a few select html tags, like <i></i>, <b></b>, <sub></sub>, <br> etc (whatever plotly supports)

Testing

render_mode

Passed render_mode = 'auto', 'svg', 'webgl' and not passing it at all, worked as intended. svg was a few times slower than webgl/auto on a large set as expected, but at least that allows use of pandasgui without webgl.

Title

Histogram

filter applied:

testing orientation:

Scatter

size not in legend, but in subtitle:

Lowess trendlines (requires statsmodels to be installed and the use of the trendline custom kwargs set to lowess:

OLS trendlines (requires statsmodels to be installed and the use of the trendline custom kwargs set to ols:

Line

This will have to be revisited once line and bar have groupby as optional, title is not totally accurate as it is not reflecting masked observations and the average.

Using log scale on Y:

Regular scale:

Bar

See comment above on groupby for lines (for bar, in one case, average, in the other would be a count)

This is the current result with groupby:

With groupby disabled it would have looked something like this:

Box

Works fine with dark theme:

Title above has additional text, entered through custom kawrgs:

While on the subject of dark theme, trying a different plot type:

Violin

[edited] Violin has similar title to box plot

Scatter 3D

most basic 3d plot:

3d with log scale on Y, filter applied and using size:

Heatmap

Heatmap with X/Y:

Heatmap with X/Y/Z default aggregation (note, this is sum for heatmap):

Heatmap with X/Y/Z specified histfunc of max:

Contour

Contour with X/Y:

Contour with X/Y/Z default aggregation (note, this is count for contour):

Heatmap with X/Y/Z specified histfunc of max:

Pie

Standard pie chart:

Pie chart with custom kwarg:

Scatter matrix

Multiple dimensions (with units removed in title), no filters, origin / model year in legend (color, symbol), size in subtitle, all possible elements in play:

Wordcloud

With filter applied:

Closes

This closes #101

This closes #102

END OF LINE

[adamerose]

Nice work thanks