Writing A Kiva Backend

To write a Kiva backend you need to implement all the abstract methods of the AbstractGraphicsContext. Once this is done, instances of the new backend should be able to be used in most places that one of the current backends is. Most sophisticated 2D drawing libraries are very similar to Kiva in terms of the facilities they offer, and so in many cases this is just a matter of translating between the Kiva API and the underlying library.

In some cases, however, the underlying drawing capabilities are more primitive.

GraphicsContextBase

Kiva provides a core set of functionality in GraphicsContextBase that makes it easier to implement the full Kiva API on top of a more basic drawing API (for example one which can’t draw Bezier curves or arcs, or have a notion of graphics state). To use this class, you need to subclass and implement a number of device_ methods that actually perform the drawing operations. These include:

• device_update_fill_state

• device_update_line_state

• device_fill_points

• device_stroke_points

• device_draw_image

• device_show_text

• device_get_full_text_extent

• device_set_clipping_path

• device_destroy_clipping_path

The GraphicsContextBase class tracks all the required state internally, including saving and restoring state. However the device code needs to be able to handle applying the current affine transformations and styling when drawing polygons, text and images, either in the implementation code or in the underlying graphics library.