Here's what's new in Release 1.1 (9 October 2017):
Bug fixes
Fixed an issue with datetime64 values. Previously, in the
display()
function's table view, datetime64 values in Pandas appeared as Unix timestamps. These date values are now properly displayed (
#279
).
When importing PixieDust, it will no longer override an existing function named "display." In which case, users can still call PixieDust's display function using
pixiedust.display()
. If no pre-existing functions named "display" are found, then PixieDust will work as usual, allowing users to call display directly. (
#455
).
New behavior when calling the display function on an unsupported entity. Instead of telling users that "No PixieDust visualization was found" when attempting to call
display()
on a data object that's not a PySpark or Pandas DataFrame, PixieDust now falls back to the IPython display method (
#343
).
Mapbox user layers are fixed. Previously, PixieDust would break this feature because of changes to Python's geojson package ( #453 ).
Map visualizations no longer aggregate latitude-longitude data. Previously, PixieDust expected an aggregation function in this situation, making it difficult to simply explore points on a map ( #410 ).
Better handling of invalid API keys for map visualizations. Previously, if a maps API key was missing or invalid, PixieDust would display a black map. Now, it surfaces an error message to the user ( #367 ).
Enhancements
An alpha release of PixieApps web publishing is now live. Introduced in release 1.0.8 , PixieApps allow developers to run interactive UI elements, directly from notebook cells. Now with PixieApps' web publishing feature, the same interactive UI elements developed to run in-notebook can now be published as a web application at the click of a button. See the announcement blog for more details ( #450 ).
Mapbox is now the default option for visualizing geospatial data on a map. Previously, PixieDust would choose Google maps by default, which requires additional configuration ( #451 ).