Toyplot

- Added a user guide section on embedding plots.
- Added a user guide example of datetime objects as tick labels.
- Make the Toyplot sourcecode fully PEP-8 conforming - thanks to Chris Morgan.
- Worked around problems with numpy.broadcast_arrays() in numpy 1.8.
- Removed LaTeX table formatting functionality that was replaced by table axes.
- Added a new backend to display figures in a standalone Qt window.
- Switched to the Python logger module for warnings / errors.
- Updated the public API for specifying scalar color palettes / maps, and deprecated separate color palette / map API parameters.
- Changed the way we encode opacities, for compatibility with Inkscape and Adobe Illustrator.
- Removed the obsolete toyplot.selenium backend.
- Treat hlines() and vlines() as annotation (so they don't affect the data domain), unless the caller specifies otherwise.
- Created new Qt backends to generate PDF and PNG figures.
- Figures can be resized consistently across all browsers, particularly Firefox and IE.
- Reorganized the backend documentation, and explicitly documented the distinction between backends and displays.
- Fixed a case where canvas resizing didn't handle explicit units correctly.
- Added a new section on interaction to the user guide.
- Allow figure creators to override the default filename when users export data from an interactive figure.
- Significant changes to our travis-ci.org test environment.
- toyplot.data.Table.matrix() didn't work in Python 3.
- Removed toyplot.data.Table.to_csv(), we want to discourage people from using Toyplot for data manipulation.
- Many objects didn't render properly in Jupyter notebooks with Python 3.
- Added parameters to disable the row and column labels in matrix visualizations.

- Unicode text wasn't handled correctly by text marks.
- Added an experimental matrix visualization using table axes.
- Added a "title" property for table cells.
- Fix inconsistencies in our use of alignment-baseline and text-anchor CSS properties.
- Added a new section to the user guide on the convenience API.
- Allow real-world units for canvas layouts, and tweak the parameter order for corner layouts.
- Expanded user guide documentation on canvas layouts.
- Added table axes regions for all four sides and corners, plus a property to access every cell in a region.
- Added automatic conversion from numpy NpzFile to toyplot.data.Table.
- Added experimental support for graph visualization.
- Allow toyplot.data.Table initialization from a sequence of 2-tuples.
- Cairo backends were ignoring -toyplot-anchor-shift.
- Cairo backends didn't handle all supported alignment-baseline values.
- Added matrix and table visualizations to the convenience API.
- Added accessors for shape, row count, and column count for table axes and regions.
- Added toyplot.locator.Integer, and a step parameter to control labelling for matrix visualizations.
- Always return a unicode string from toyplot.canvas.Canvas._repr_html_(), for compatibility with Jupyter / IPython notebooks running Python 3 kernels.
- Assign a sensible default filename for CSV downloads, for browsers that support it.
- Added a contributed Conda build recipe.
- Allow toyplot.data.Table to be initialized from a 2D numpy array.
- Rename the toyplot.axes.Table "title" parameter to "label" for consistency with the other axes.
- Added a new "Labels and Legends" section to the user guide.
- Added a new "Tick Locators" section to the user guide.
- Added experimental toyplot.data.contiguous() function to identify contiguous ranges in an array.
- Fix a problem with interactive Y coordinates when using a log scale that straddled the origin.

- Switched to https://travis-ci.org/sandialabs/toyplot for continuous integration testing.
- Switched to https://coveralls.io/r/sandialabs/toyplot to track test coverage.
- Added a custom CSS style `-toyplot-anchor-shift` for controlling horizontal text offsets.
- Added new documentation on color, text alignment, units, data tables, and table axes to the user guide.
- Callers can increase the number of table rows and columns when creating table axes from a data table.
- Overhauled Toyplot's handling of real-world units, allowing arbitrary units throughout the API, and made it explicit that the default canvas units are CSS pixels.
- Added axis visibility options to the convenience API.
- `toyplot.data.Table` can be converted to a `numpy` matrix.
- Positive angles yield counterclockwise rotation throughout the API, for consistency with trigonometry.
- Rendered text automatically expands a plot's domain to avoid clipping.
- Fixed a longstanding problem displaying mouse coordinates outside the data domain for a plot.
- Moved interaction-specific markup from the SVG backend to the HTML backend.
- When exporting data from a figure, only the caller-supplied data is exported.
- The API makes an explicit distinction between text used for "annotation" and text used for data.
- Many small fixes.

- Began continuous integration testing.
- Switched from ost.io to https://gitter.im/sandialabs/toyplot for support requests.
- Made the HTML backend the primary renderer.
- Improved logarithmic tick formatting and customization.
- Increased consistency between the fill() and plot() APIs.
- Simplified the way colors are inherited for line plots and scatter plots.
- Added basic functionality for reading and writing CSV files.
- Note: for pedagical purposes only - Toyplot is _not_ a data manipulation tool!
- Ongoing improvements to the table axes API:
- Added support for table titles.
- Added support for hiding table headers.
- Table headers can have multiple rows.
- Ensure that visible cells are rendered in a deterministic order.
- Create a default grid line between table header and body.
- Added support for user-configurable gaps between cells.

- Switched to toyplot.data.Table for all internal data storage.
- Reorganized the codebase into smaller, more focused modules.
- Added a new backend to produce WebM video.
- Data tables can be rendered to LaTeX.
- New table axes for rendering tables as data graphics.

- Introduced support for Python 3.
- Removed pure black from the default styling.
- Allow regression tests to run without optional dependencies.