The Dojo Loader¶
Since program complexity tends to increase geometrically as program size increases linearly, the only hope of success when implementing a large program is to break it into several small programs. From the beginning, the Dojo Toolkit was designed for use when building large programs which is why Dojo was among the first libraries to include a formal module system.
The idea of a module system is fairly straight-forward:
- define an API to use when defining modules
- construct a loader that implements the API so that modules can be assembled into a complete program at run-time
Since each individual resource implies an HTTP transaction to download that resource, bandwidth and latency constraints can result in sluggish loading performance. Fortunately, this problem is easy to fix. Once the program is constructed, the modules can be analyzed by a program and aggregated so that only few resources are actually downloaded, with each of those resources containing several logical modules. Typically, this step occurs when moving from development to deployment.
In version 1.7.0, the Dojo loader and build system have been completely rewritten to support both the AMD API and full legacy backward compatibility. Several new supporting APIs have also been constructed, including the AMD plugin, has.js, dojo configuration, dojo loader micro-events, error reporting, and tracing APIs. Lastly, most of the dojo and dijit trees as well as large parts of the dojox tree have been converted to use the AMD API.
The tutorials and reference manuals listed below describe all of this machinery.