The AEOM provided a syntactic layer under which any application could publish its internal objects, allowing those objects to be manipulated in a standardized way. Unlike other similar-sounding concepts such as ToolTalk, there was a clear, orthogonal distinction between nouns and verbs; thus, instead of providing separate commands for "close document" and "close window", there was a single "close" verb which could take references to "document" or "window" objects, or any other object that the application published.
The objects that an application made available through its AEOM support were arranged in a hierarchy. At the top was the application itself, referenced via a null object descriptor. Other objects were referenced by (recursively) specifying their parent object, together with other information identifying it as a child of that parent, all collected in an AERecord. An iterator was provided by parents to enumerate their children, or children of a certain class, allowing applications to address a set of elements. The system was generally similar to the Document Object Model used in XML, although with some differences in access patterns.
Each object could have elements and properties; elements were other objects, which might be created or deleted, while properties could not be created or deleted but had values which might be interrogated or changed. For instance, the application might have one or more window elements representing windows showing the contents of currently-open documents. These windows might have properties such as their title, position and size.
An application could define custom verbs for operating on its objects. The AEOM also specified various standard verbs which (it was hoped) applications would implement in a consistent fashion, such as open, close, create element, delete, set data, and get data. Each verb was defined as an AppleEvent of a specific type and class, together with particular parameters of particular types that were expected to be present. For instance, the "get data" event was the standard means for obtaining the value of a property: it took essentially one parameter, which was an object descriptor identifying the property to be queried. The value of that property would be returned in the reply event. The "set data" event took two parameters, the object descriptor for the property to set and the new value for the property; the reply event was only expected to return a success status or failure error code.
Note that the entire AppleEvent architecture identifies things using four-byte OSType codes, studiously avoiding actual words or phrases in English (or any other language). Instead, the correspondence between internal AppleEvent codes and external natural-language descriptions is specified through the aete (AppleEvent Terminology Extension) resource -- the "extension" being to the standard terminology built into AppleScript itself. An application may provide multiple 'aete' resources for multiple languages, in keeping with the original multilingual design of AppleScript itself
For instance, consider the following AppleScript sequence controlling a fictional drawing application:
tell application "ScriptableDraw"
set background colour of window "New Drawing" to background colour of window "Old Drawing"
This actually involves the sending of two AppleEvents to the target application (and the receipt of their corresponding replies): first, a get-data event is sent to retrieve the background colour property of the window identified by the name "Old Drawing"; then a set-data event is sent to apply the value returned as the background colour property of the window named "New Drawing".
Since this sort of access pattern was typical, AppleScript made widespread use of the
tell statement, which switched the context to the named object in a fashion similar to the
with statement found in Visual Basic or Pascal. All commands after the
tell to the corresponding
end tell would be sent to the object named in the
tell, instead of the default object, which was the current application.
Object descriptors allowed the identification of objects in various ways. The most interesting one was using a where-clause (which translated into AppleScript terminology as a filter expression). For instance, the AppleScript 1.0 SDK shipped with the source code for an example application called the Scriptable Text Editor, which would respond to scripts such as:
tell application "Scriptable Text Editor"
tell window "Example Document"
set text style of every word whose length > 7 to bold
Even today, it is rare to find this kind of power in general-purpose scripting languages outside of SQL.
Adding support for the AEOM in the "classic" Mac OS was a difficult process. Application developers had to identify their objects and hand-write code to allow them to be addressed. This typically took the form of code for returning the "next" object of a particular type, allowing AppleScript to iterate over them. But since the OS did not contain an object model, this work was left entirely to the developers, many of whom did not implement it. Oddly, even Apple's own application framework, MacApp, did not offer such a model except for the GUI objects it knew about, once again making the developer do most of the work of scripting the objects representing the data itself. Largely for these reasons, AppleScript support was not very widespread.
Apple did attempt to address this problem with the introduction of various object "suites", which represented standard objects and verbs that were expected to be supported by different types of applications. For instance, all applications were expected to support the "core suite", and any application editing text was expected to support the "text suite". By selecting a suitable set of suites, the developer could at least reduce the workload of planning how to expose their objects. Yet because these objects were generally not part of the system itself (with the exception of the severely limited TextEdit editor), the actual implementation was left to the developer.
Applications developed in Cocoa, the system formerly known as OpenStep, offer a rich object runtime that can be queried from any other application. This makes implementation of the AEOM considerably easier, dramatically reducing the amount of code needed in the average application. Additionally the majority of Cocoa applications are constructed primarily from Cocoa-standard objects, all of which were upgraded to offer fairly extensive scriptability. This extends not only to GUI objects as under MacApp, but also to data objects inside them, including text, tables and various list objects. A text file is used to map the internal "object-like" names onto human-readable versions, and in most cases creating this is all that is needed to add fairly substantial scriptability to most programs.
So while Cocoa applications are not AOEM based, and often use subtly different objects than Apple's originally defined standard objects, Cocoa apps are generally much more scriptable than their "classic" counterparts -- in fact it is uncommon to find a Cocoa application that is not scriptable to some degree.