Apatite is being developed as part of the Natural Programming Project in the Human-Computer Interaction Institute at Carnegie Mellon University.
(If you like Apatite, please check out our other documentation tool, called Jadeite!)
| NEWS! Apatite won first place in the "Yahoo! Undergraduate Research Awards" competition at Carnegie Mellon University, May 6, 2009! See Dan Eisenberg with his winning poster. |
Apatite initially displays a single column containing the most popular items in five different categories: Packages, Classes, Methods, Actions, and Properties. The larger an item's font size is, the more commonly it is used.
Clicking on the [+] sign of a category will expand that section to show you a longer list. Clicking the [-] sign will bring you back to the original format.
The column can also be filtered by a set of keywords by typing them into the text box at the top. Font sizes are automatically adjusted to reflect the relative popularity of each item.
When you click on an item, a new column appears which only contains items that are associated with your selection. This column works in the same way as the first one, and items with larger font sizes are the most relevant.
As you traverse through the interface, new columns are organized to display the items it thinks are most relevant based on all of your previous selections.
Hovering over the ? button will display information about a particular item. Clicking on it opens up that item's full documentation in a new window.
-----------------
Jeffrey Stylos, Daniel S. Eisenberg, Brad A. Myers. "Apatite: Associative Browsing of APIs." Submitted for publication.
Abstract:
Learning how to use new APIs can be difficult, in part because they are large and can be difficult to explore effectively using existing tools. Current object-oriented API documentation is usually organized by class, however programmers sometimes have a desired action or method in mind without knowing which class it might belong to. Programmers using existing API documentation often struggle to determine which of an API’s many classes and methods are the important ones to pay attention to, spending considerable time looking at obscure and irrelevant parts of APIs as a result. To help with these problems, we designed Apatite, an API browsing tool that lets programmers browse by association—investigating which packages, classes, methods go with others. Apatite also adds browsing by actions and properties, and uses statistics about the popularities of items and how often they are used together to provide weighted views of the most relevant items. We used an iterative design process to help tailor the tool to programmers’ needs, and a preliminary user study suggests that it offers advantages over existing tools.
Download:
(pdf)
Copyright © 2009 - Carnegie Mellon University
