Toyplot

- Allow fill marks to be used as annotation.
- Explicitly disable data export from annotation marks.
- Add an experimental `<axis>.domain.show` parameter to control whether the domain is displayed using axis spines.
- `toyplot.data.read_csv(convert=True)` will try to parse integer as well as floating-point types.
- Completely rewrote the table coordinates implementation.
- Table coordinates support advanced, numpy-style indexing for all rows, columns, cells, gridlines, and gaps.
- Added API to delete table coordinate rows and columns.
- Added API to insert table coordinate rows and columns.
- By default, all table cells are vertically and horizontally centered with a default font.
- Matrix visualizations no longer bold row and column indices by default.
- End-users can export CSV data from table coordinates and matrix visualizations.
- Added table-cell bar plots and line plots that use the data already contained in the table.

- Pandas data frame indices (including hierarchical indices) can optionally be included when converting to `toyplot.data.Table`.
- Fixed a Python 3 portability issue.
- Table coordinates didn't format NaN values properly when using a custom formatting string.
- The `arrow` module is only imported when needed.
- New documentation on grouping table rows.
- Documented platform-specific timezone naming issues.
- Improved documentation of the color factory objects in `toyplot.color`.
- Use consistent naming for numberline coordinates.
- Made it easier to iterate over `toyplot.data.Table` rows.
- Interactive mouse coordinates work correctly with numberlines and shared axes, and are only displayed by click / touch events.
- Position ticks relative to axes with a `location` property, and deprecate the tick labels `location` property.
- Fixed a problem rendering bars with a log scale and nonzero domain minimum.
- Removed the API to change text during animation.
- Significant cleanup and organization of HTML backend code and generated markup.
- Renamed the `toyplot.axes` module to `toyplot.coordinates` for consistency, clarity.
- Added `toyplot.canvas.Canvas.cartesian()` and deprecated `toyplot.canvas.Canvas.axes()`.
- Added `toyplot.locator.Uniform` and deprecated `toyplot.locator.Basic`.
- Added documentation links to external libraries, where practical.
- Added `text-shadow` to the list of valid CSS text attributes.
- Updated dependencies to require numpy >= 1.8.0, and eliminated code that inadvertently depended on numpy >= 1.9.
- Experimental support for displaying `PIL` and `scikit-image` images.
- Added a `style` property to `toyplot.canvas.Canvas`.
- Deprecated implicit conversion from palettes to colormaps for matrix visualization.
- Provide better error messages if a caller passes anything but a canvas to a rendering backend.
- Add support for multi-series marks in legends.
- Updated links to point to our new documentation domain, `http://toyplot.readthedocs.io`.
- Axis labels support the same `location` and `offset` parameterization as axis ticks / tick labels.

- Added more complex indexing / slicing options to toyplot.data.Table.
- Deprecated `toyplot.data.Table.rows()` and `toyplot.data.Table.columns()`.
- Removed support for custom markers.
- `-toyplot-anchor-shift` didn't work correctly with rotated text.
- Documented text alignment behavior for rotated text.
- Added `location` parameter for axis labels.
- Improved text alignment defaults for rotated and unrotated axis labels.
- Don't alter the axis domain if tick labels aren't visible.
- Change the default linear color map to a diverging blue-red palette.
- Pandas data frames with duplicate column names can be converted to `toyplot.data.Table`.
- Allow callers to suppress NaNs in table visualization cells.
- Render color arrays as swatches in Jupyter notebooks.
- Added `toyplot.color.brewer.palette()`, `toyplot.color.brewer.map()`, and `toyplot.color.diverging.map()`.
- Deprecated `toyplot.color.brewer()` and `toyplot.color.diverging()`.
- `toyplot.color.LinearMap` color stops can be explicitly positioned.
- Added `toyplot.color.linear.map()` with "Blackbody", "ExtendedBlackbody", "Kindlmann" and "ExtendedKindlmann" color maps.
- Deprecated implicit conversions from color palettes to color maps during color mapping.
- Split color-related documentation into separate "Color" and "Color Mapping" sections of the user guide.
- Improved debugging output when a regression test fails.
- Many code coverage improvements.

- Added rich text support, using a limited subset of HTML markup.
- Added a tick locator for displaying timestamp data with properly formatted times.
- Created a new, pure-Python PDF backend using ReportLab.
- Created a new PNG backend that renders by rasterizing PDFs with Ghostscript.
- Removed deprecated PDF and PNG backends.
- Added numberline axes, for displaying one-dimensional data.
- Refactored the scatterplot mark to support data with any number of dimensions.
- Added one-dimensional scatterplot support to numberlines.
- Completely redesigned the color scale implementation to use numberlines.
- Added API for easily adding color scales to axes and matrix visualizations.
- Provided both size and area parameters to specify marker sizes.
- Moved log scales to a dedicated section of the user guide.
- Optimized graph layout when every vertex already has a position.
- Removed the GraphViz graph layout strategy.
- Use consistent naming for matrix visualization parameters.
- toyplot.data.read_csv() can optionally convert string values to numeric values.
- Replaced toyplot.color.lighten() with toyplot.color.spread(), which is more flexible.
- Display toyplot color values as swatches in Jupyter notebooks.
- Expanded the color documentation in the user guide.
- Reduced regression test boilerplate code.
- Test coverage improvements.

- Documented installation for Anaconda and FreeBSD.
- Experimental support for graph visualization, with flexible layout algorithms, shared layouts and node "pinning".
- Allow cartesian axes to fill the available range while maintaining their aspect ratio.
- Axis ticks can extend above or below the axis spine.
- Positioning an axis spine positions its ticks and tick labels as well.
- Added support for shared axes / multiple axes, to display multiple overlapping domains in a single plot.
- Format specifiers are available for the Extended and Heckbert tick locators, courtesy of Johann du Toit.
- Began using pylint as a regular code quality check.
- Pandas data frames are automatically converted when creating data tables / table axes.
- Created a new default PDF backend using the ReportLab library.
- Switched to toyplot.qt.png as the default PNG backend.
- Provide better feedback when using the toyplot.pdf and toyplot.png meta backends.

- Removed deprecated colormap and palette parameters from the API.
- Allow simplified color mapping specifications.
- Improved test coverage.
- Fix a problem embedding embedding axes in tables using more than one merged cell.
- Add table cell width / height support for real-world units.
- Hide masked values in table axes.
- Reorganize the installation documentation.
- Add support for rotated text in table cells.
- Add top/bottom/left/right label support for matrix visualizations.
- Add new toyplot.locator.Null do-nothing tick locator.
- Add matrix visualization support for right / bottom ticks.
- Add custom locator support for matrix visualizations.
- Make matrix visualization color parameters consistent with the rest of the API.
- Add missing reference documentation for toyplot.projection module.
- Cleanup the toyplot.color.broadcast(...) API and implementation.
- Make the API for specifying color mapping consistent across all visualization types.
- Allow per-datum titles on line plots and scatterplots.
- Expand the color section in the user guide to cover color mapping.
- Add a new section on null data to the user guide.
- Eliminate nuisance warnings from numpy.
- Automatically validate source notebooks as part of the documentation build.