Roadmap.rst

Roadmap

(As of 4/2019)

The Panel library is new, but it builds on at least a decade of work in previous and related Python projects. Thus there are many parts of Panel that are very solidly implemented, with features that have been tested and improved over years of active use by large communities of Bokeh and Param users. Other features are new to Panel itself or were added to support Panel, and will gradually become more solid as they get into widespread use.

The currently funded priority items for development in 2018-2019 include:

1. Polishing the new features, documentation, and examples: As issues, bugs, and usabilty gaps are reported on the issue tracker, we will address these as time permits.
2. GridSpec: Panels can currently be created using interact() or as nested pn.Row() or pn.Column() objects, potentially with pn.Spacer() objects to adjust spacing. A complementary approach like Matplotlib's GridSpec makes it simpler to make an arrangement into a regular grid by defining the total rows and columns from the start, then assigning individual panes to specifc ranges of rows and columns in the grid. An initial implementation is underway, but it will need more work before it becomes usable.
3. Widget improvements: There are many improvements needed to the detais of the provided widgets, such as making numeric values editable, improving the stepsize behavior, making layouts more compact, lining up widgets better, combining multiple related widgets per row, allowing widget groups to be laid out horizontally, and so on. These improvements will happen gradually, and mostly depend on changes to the underlying Bokeh library.
4. Static export: Panel widgets are currently only "live" when backed by a Python server, but in many cases the mapping from widget to displayed object follows a well-defined pattern, which can be captured in JavaScript code that runs with or without a Python process. Supporting static export to .html that can be sent in an email or posted on any website will require capturing these typical communication patterns, as well as providing a set of communication channels to ship with exported documents. Examples of the results of this general approach are on holoviews.org, but Panel will need its own separate implementation for static export.
5. Responsive sizing: Panel objects (panes and similar viewable items) currently accept a fixed height and width in pixels. In many cases, it's desirable to have at least some of the objects responsively adjust to the screen size available, so that they can make use of whatever screen or window area is available. Responsive sizing can be implemented once Bokeh's current layout-refactoring project has been completed.

Other features we'd love to see in Panel but which are not currently funded or scheduled include:

1. Other plotting libraries: Panel already supports a wide variety of libraries, including all the libraries currently in use by the authors or their collaborators. Most other libraries can trivially be supported as well, if anyone can provide an example of a plot already supported by Panel converted into this other library. As mentioned in issue 2, the ipywidgets-based libraries will be more complicated to support, but any library that produces PNG, SVG, HTML, or another basic type should be very straightforward to include.
2. Themes: By default, Panel apps use Bokeh's default theming, but other look-and-feel options can be provided by using other available Bokeh themes, making your own, or embedding into a Bokeh Jinja2 template. We'll add examples of such theming to the Panel website as we develop them.
3. BI tools: Panel is designed for displaying content that you already have developed in a Jupyter notebook, but it could also be used as a way for building business-intelligence-style dashboards that mix traditional Python plotting output with BI-style indicators like speedometers, single number displays (e.g. stock tickers), and so on. Adding a small number of such plots or widgets could help make it simpler to build BI-type dashboards where the content comes directly from Python.
4. Jupyter notebook extension: The Jupyter-Dashboards project provided a Jupyter extension allowing drag and drop layout for Jupyter notebook cells. This approach worked well in many cases, but the underlying server used for deployment is no longer maintained, and so that approach remains only a proof of concept. Panel, on the other hand, is fully deployable outside of the notebook as well as in it, and so it would be great to have a drag-and-drop interface either based on the Jupyter-Dashboards project or developed separately to provide similar capabilities. As for Jupyter Dashboards, the layout information could be stored as cell metadata in the notebook, providing hints for Panel to set up the dashboard layout when served separately. There are likely to be tricky issues with control flow that would need to be addressed when a notebook is used in a non-linear dashboard in this way.
5. Support for Traitlets: Panel currently supports declarative user interfaces for objects defined using the Param library, and similar support could be added for objects defined using the Traitlets library. Traitlets and Param offer similar functionality, so the benefit of adding Traitlets support would primarily be for users who have large bodies of code already written as traitlets.

If any of the functionality above is interesting to you (or you have ideas of your own!) and can offer help with implementation, please open an issue on this repository. And if you are lucky enough to be in a position to fund our developers to work on it, please contact sales@anaconda.com.