This chapter describes the basics of defining and compiling classes. It discusses the following topics:
When viewing this book online, use the preface
of this book to quickly find other topics.
The following shows a simple Caché class definition, with some typical elements:
Class Demo.MyClass Extends %RegisteredObject
Property Property1 As %String;
Property Property2 As %Numeric;
Method MyMethod() As %String
Note the following points:
Caché provides a large set of class definitions that your classes can use in the following general ways:
You can use Caché classes as superclasses for your classes.
You can use Caché classes as values of properties, values of arguments to methods, values returned by methods, and so on.
Some Caché classes simply provide specific APIs. You typically do not use these classes in either of the preceding ways. Instead you write code that calls methods of the API.
The most common choices for superclasses are as follows:
The most common choices for values of properties, values of arguments to methods, values returned by methods, and so on are as follows:
Later chapters of this book discuss these categories of classes.
The phrase object class
refers to any subclass of %RegisteredObject
. With an object class, you can create an instance of the class, specify properties of the instance, and invoke methods of the instance. A later chapter describes these tasks (and provides information that applies to all object classes).
The generic term object refers to an instance of an object class.
There are three general categories of object classes:
The following figure shows the inheritance relationship among these three classes. The boxes list some of the methods defined in the classes:
Collection classes and stream classes are object classes with specialized behavior.
The phrase data type class
refers to any class whose ClassType keyword equals datatype
or any subclass of such a class. These classes are not object classes (a data type class cannot define properties, and you cannot create an instance of the class). The purpose of a data type class (more accurately a data type generator class) is to be used as the type of a property of an object class.
A Caché class definition can include the following items, all known as class members
Parameters A parameter defines a constant value for use by this class. The value is set at compilation time, in most cases.
Methods Caché supports two types of methods: instance methods and class methods. An instance method
is invoked from a specific instance of a class and performs some action related to that instance; this type of method is useful only in object classes
. A class method
is a method that can be invoked whether or not an instance of its class is in memory; this type of method is called a static method
in other languages.
Properties A property contains data for an instance of the class. Properties are useful only in object classes
. The following subsection provides more information.
Class queries A class query defines an SQL query that can be used by the class and specifies a class to use as a container for the query. Often (but not necessarily), you define class queries in a persistent class, to perform queries on the stored data for that class. You can, however, define class queries in any class.
XData blocks An XData block is a named unit of data defined within the class, typically for use by a method in the class. These have many possible applications.
Projections A class projection provides a way to extend the behavior of the class compiler.
The projection mechanism is used by the Java and C++ projections; hence the origin of the term projection
Formally, there are two kinds of properties: attributes and relationships.
Attributes hold values. Attribute properties are usually referred to simply as properties
. Depending on the property definition, the value that it holds can be any of the following:
This section discusses basic class definitions in more detail. It discusses the following topics:
Typically, you use Studio
to define classes. You can also define classes programmatically using the Caché class definition classes or via an XML class definition file. If you define an SQL table using SQL DDL statements, the system creates a corresponding class definition.
When you define a class, one of your earliest design decisions is choosing the class (or classes) which to base your class. If there is only a single superclass, include Extends
followed by the superclass name, at the start of the class definition.
Class Demo.MyClass Extends Superclass
If there are multiple superclasses, specify them as a comma-separated list, enclosed in parentheses.
Class Demo.MyClass Extends (Superclass1, Superclass2, Superclass3)
It is not necessary to specify a superclass when you create a class. It is common to use %RegisteredObject
as the superclass even if the class does not represent any kind of object, because doing so gives your class access to many commonly used macros, but you can instead directly include the include files that contain them.
You can also create your own include files and include them in class definitions as needed.
To include an include file at the beginning of a class definition, use syntax of the following form. Note that you must omit the .inc
extension of the include file:
To include multiple include files at the beginning of a class definition, use syntax of the following form:
Include (MyMacros, YourMacros)
Note that this syntax does not have a leading pound sign (in contrast to the syntax required in a routine). Also, the Include
directive is not case-sensitive, so you could use INCLUDE
instead, for example. The include file name is case-sensitive.
In some cases, it is necessary to control details of the code generated by the class compiler. For one example, for a persistent class, you can specify an SQL table name, if you do not want to (or cannot) use the default table name. For another example, you can mark a class as final, so that subclasses of it cannot be created. The class definitions support a specific set of keywords for such purposes. If you need to specify class keywords, include them within square brackets after the superclass, as follows:
Class Demo.MyClass Extends Demo.MySuperclass [ Keyword1, Keyword2, ...]
A class parameter defines a constant value for all objects of a given class. To add a class parameter to a class definition, add an element like one of the following to the class:
Parameter PARAMNAME as Type;
Parameter PARAMNAME as Type = value;
Parameter PARAMNAME as Type [ Keywords ] = value;
To add a property to a class definition, add an element like one of the following to the class:
Property PropName as Classname;
Property PropName as Classname [ Keywords ] ;
Property PropName as Classname(PARAM1=value,PARAM2=value) [ Keywords ] ;
Property PropName as Classname(PARAM1=value,PARAM2=value) ;
Depending on the class used by the property, you might also be able to specify property parameters, as shown in the third and fourth variations.
Notice that the property parameters, if included, are enclosed in parentheses and precede any property keywords. Also notice that the property keywords, if included, are enclosed in square brackets.
You can define two kinds of methods in Caché classes: class methods and instance methods.
To add a class method to a class definition, add an element like the following to the class:
ClassMethod MethodName(arguments) as Classname [ Keywords]
is the name of the method and arguments
is a comma-separated list of arguments. Classname
is an optional class name that represents the type of value (if any) returned by this method. Omit the As Classname
part if the method does not return a value.
Method MethodName(arguments) as Classname [ Keywords]
Class and class members follow specific naming conventions. These are detailed in this section.
This section describes the rules for class and member names, such as maximum length, allowed characters, and so on. A full class name includes its package name, as described in the next section.
Every identifier must be unique within its context (that is, no two classes can have the same name). Caché has the following limits on package, class, and member names:
Each package name can have up to 189 unique characters.
Each class name can have up to 60 unique characters.
The combined length of the name of a property and of any indices on the property should be no longer than 180 characters.
The full name of each member (including the unqualified member name, the class name, the package name, and any separators) must be 220 characters or fewer.
Each name can include Unicode characters.
Identifiers preserve case: you must exactly match the case of a name; at the same time, two classes cannot have names that differ only in case. For example, the identifiers id1
are considered identical for purposes of uniqueness.
Identifiers must start with an alphabetic character, though they may contain numeric characters after the first position. Identifiers cannot contain spaces or punctuation characters with the exception of package names which may contain the .
character. On a Unicode system, identifiers may contain Unicode characters.
Certain identifiers start with the %
character; this identifies a system item. For example, many of the methods and packages provided with the Caché library start with the %
Member names can be delimited, which allows them to include characters that are otherwise not permitted. To create a delimited member name, use double quotes for the first and last characters of the name. For example:
Property "My Property" As %String;
Every class has a name that uniquely identifies it. A full class name consists of two parts: a package name and a class name: the class name follows the final .
character in the name. A class name must be unique within its package; a package name must be unique within a Caché namespace. For details on packages, see the chapter Packages.
Because persistent classes are automatically projected as SQL tables, a class definition must specify a table name that is not
an SQL reserved word
; if the name of a persistent class is an SQL reserved word, then the class definition must also specify a valid, non-reserved word value for its SQLTableName
Every class member (such as a property or method) must have a name that is unique within its class and with a maximum length of 180 characters. Further, a member of a persistent cannot use an SQL reserved word
as its identifier. It can define an alias, however, using the SQLName
keyword of that member (as appropriate).
InterSystems strongly recommends that you do not give two members the same name. This can have unexpected results.
A Caché class can inherit from already existing classes. If one class inherits from another, the inheriting class is known as a subclass
and the class or classes it is derived from are known as superclasses
The following shows an example class definition that uses two superclasses:
Class User.MySubclass Extends (%Library.Persistent, %Library.Populate)
The syntax shown here corresponds to the Super keyword, which is visible in the Studio Inspector and in class definitions exported as XML.
In addition to a class inheriting methods from its superclasses, the properties inherit additional methods from system property behavior classes and, in the case of a data type attribute, from the data type class.
For example, if there is a class defined called Person
Class MyApp.Person Extends %Library.Persistent
Property Name As %String;
Property DOB As %Date;
It is simple to derive a new class, Employee
, from it:
Class MyApp.Employee Extends Person
Property Salary As %Integer;
Property Department As %String;
This definition establishes the Employee
class as a subclass of the Person
class. In addition to its own class parameters, properties, and methods, the Employee
class includes all of these elements from the Person
You can use a subclass in any place in which you might use its superclass. For example, using the above defined Employee
classes, it is possible to open an Employee
object and refer to it as a Person
Set x = ##class(MyApp.Person).%OpenId(id)
We can also access Employee-specific attributes or methods:
Write x.Salary // displays the Salary property (only available in Employee instances)
The leftmost superclass that a subclass extends is known as its primary superclass
. A class inherits all the members of its primary superclass, including applicable class keywords
, properties, methods, queries, indices, class parameters, and the parameters and keywords of the inherited properties and inherited methods. Except for items marked as Final
, the subclass can override (but not delete) the characteristics of its inherited members.
See the next section for more details about multiple inheritance.
By means of multiple inheritance, a class can inherit its behavior and class type from more than one superclass. To establish multiple inheritance, list multiple superclasses within parentheses. The leftmost superclass is the primary superclass.
For example, if class X
inherits from classes A
, and C
, its definition includes:
Class X Extends (A, B, C)
The default inheritance order for the class compiler is from left to right, which means that differences in member definitions among superclasses are resolved in favor of the leftmost superclass (in this case, A
, and B
Specifically, for class X
, the values of the class parameter values, properties, and methods are inherited from class A
(the first superclass listed), then from class B
, and, finally, from class C
also inherits any class members from B
has not defined, and any class members from C
that neither A
has defined. If class B
has a class member with the same name as a member already inherited from A
, then X
uses the value from A
; similarly, if C
has a member with the same name as one inherited from either A
, the order of precedence is A
, then B
, then C
Because left-to-right inheritance is the default, there is no need to specify this; hence, the previous example class definition is equivalent to the following:
Class X Extends (A, B, C) [ Inheritance = left ]
To specify right-to-left inheritance among superclasses, use the Inheritance keyword with a value of right
Class X Extends (A, B, C) [ Inheritance = right ]
With right-to-left inheritance, if multiple superclasses have members with the same name, the superclass to the right takes precedence.
Even with right-to-left inheritance, the leftmost superclass (sometimes known as the first superclass) is still the primary superclass. This means that the subclass inherits only the class keyword values of its leftmost superclass there is no override for these.
For example, in the case of class X
inheriting from classes A
, and C
with right-to-left inheritance, if there is a conflict between a member inherited from class A
and one from class B
, the member from class B
overrides (replaces) the previously inherited member; likewise for the members of class C
in relation to those of classes A
. The class keywords for class X
come exclusively from class A
. (This is why extending classes A
in that order with left-to-right inheritance is not the same as extending classes B
in that order with right-to-left inheritance; the keywords are inherited from the leftmost superclass in either definition, which makes the two cases different.)
Before version 2010.1 of Caché, inheritance order was always right-to-left and could not be changed. Classes from an older instance that has upgraded will automatically continue to use right-to-left inheritance due to a class dictionary upgrade. Hence, existing code does not require any changes, even though new classes use left-to-right inheritance by default from 2010.1 onward.
The following example shows a class definition with some commonly used keywords:
/// This sample persistent class represents a person.
Class MyApp.Person Extends %Persistent [ SqlTableName = MyAppPerson ]
/// Define a unique index for the SSN property.
Index SSNKey On SSN [ Unique ];
/// Name of the person.
Property Name As %String [ Required ];
/// Person's Social Security number.
Property SSN As %String(PATTERN = "3N1""-""2N1""-""4N") [ Required ];
This example shows the following keywords:
For the class definition, the Extends keyword specifies the superclass (or superclasses) from which this class inherits.
Note that the Extends keyword has a different name when you view the class in other ways; see the next section.
For the class definition, the SqlTableName keyword determines the name of the associated table, if the default name is not to be used. This keyword is meaningful only for persistent classes, which are described later in this book.
For the index definition, the Unique keyword causes Caché to enforce uniqueness on the property on which the index is based (SSN
in this example).
For the two properties, the Required keyword causes Caché to require non-null values for the properties.
is not a keyword but instead is a property parameter; notice that PATTERN
is enclosed in parentheses, rather than square brackets.
Later chapters of this book discuss many additional keywords, but not all of them. Apart from keywords related to storage (which are not generally documented), you can find details on the keywords in the Caché Class Definition Reference
. The reference information demonstrates the syntax that applies when you view a class in the usual edit mode.
In many but not all cases, when you specify a keyword for a class definition or for a class member, you add an element of one of the following forms to the class or class member:
In the Studio Inspector, the compiler keywords and their values are presented differently. For example, consider the following class definition:
/// This sample persistent class represents a person.
/// <p>Maintenance note: This class is used by some of the bindings samples.
Class Sample.Person Extends (%Persistent, %Populate, %XML.Adaptor)
For this class, the Studio Inspector displays the following table of keywords:
Notice that both Name and Description are keywords. If you edit Description in the Inspector, Studio updates the comments in the class definition, and vice versa. Similarly, there is a keyword named Super, which specifies the superclasses of this class. If you edit that, Studio updates the Extends
part of the class definition.
The Studio Inspector has similar behavior when you display a class member. In that case, the Inspector window displays a table of all the member keywords and the values of those keywords for the currently selected member. (For a property, the Inspector window also lists the available property parameters and their current values.)
When you export a class definition to XML, the exported file looks like the following:
<Export generator="Cache" version="25" zv="Cache for Windows (x86-64) 2015.1 (Build 416U)" ts="2014-12-19 15:27:27">
This sample persistent class represents a person.
<p>Maintenance note: This class is used by some of the bindings samples.]]></Description>
Most of the XML elements in this file correspond to the compiler keywords.
The purpose of the Class Reference is to advertise, to other programmers, which parts of a class can be used, and how to use them. The following shows an example:
This reference information shows the definitions of class members, but not their actual implementations. For example, it shows method signatures but not their internal definitions. It includes links between elements so that you can rapidly follow the logic of the code; in some cases, this is quicker than using Studio. There is also a search option.
To create documentation to include in the Class Reference, create comments within the class definitions specifically comments that start with ///
. If you precede the class declaration with such comments, the comments are shown at the top of the page for the class. If you precede a given class member with such comments, the comments are shown after the generated information for that class member. Once you compile the class, you can view its generated class documentation the next time you open the Class Reference
documentation. If you add no Class Reference comments, items that you add to a class or package appear appropriately in the lists of class or package contents, but without any explanatory text.
You can extend any existing Class Reference comments from within Studio, either by editing the Description
field for a class in the Studio Inspector window, or by adding specially formatted lines to the class code. The syntax rules for Class Reference comments are strict:
All Class Reference comments that describe a class or class member must appear in a consecutive block immediately before the declaration of the item that they describe.
Each line in the block of comments must start with three slashes: ///
Note that, by default, the presentation combines the text of all the ///
lines and treats the result as single paragraph. You can insert HTML line breaks (<br>). Or you can use HTML formatting (such as <p> and </p>), as discussed in the subsection
The three slashes must begin at the first (left-most) position in the line.
No blank lines are allowed within Class Reference comments.
No blank lines are allowed between the last line of the Class Reference comments and the declaration for the item that they describe.
If you add Class Reference comments using the Description
field with a Studio wizard or in the Studio Inspector window, Studio handles these details for you (apart from the length restriction). If you add Class Reference comments directly into the code, Studio alerts you to some Class Reference syntax errors: for example, if you insert a blank line between the comments and the declaration, or if you use an insufficient number of slashes at the beginning of a line within a Class Reference text block. However, Studio does not alert you to any other types of bad syntax within Class Reference comments.
Class Reference comments allow plain text, plus any standard HTML element and a small number of specialized elements, as shown in the following code sample:
/// <p>Transforms <i>Star</i> order messages for <i>ChartScript</i>. <br/>
/// Developed Nov 2004 by <b>MT Engineering Team</b>. <br/>
/// See also <class>StarADTtoChartScript</class> and
/// <class>StarMRGtoChartScript</class> </p>
/// <p>Only Orders for these Departments pass: </p>
/// <p>As long as they are one of the following:</p>
/// <li>New Child Order</li>
/// <li>Child Order Status Change</li>
/// <li>Order Cancellation</li>
/// <p>Data Transformation sets "T" in MSH 11 for Test environment.</p>
Extends Ens.DataTransformDTL [ ProcedureBlock ]
// The data transformation class code goes here.
The previous example formats the Class Reference
entry for the class as follows:
You can use HTML tags within the comments in a class. With regard to the allowed HTML elements, adhere to as strict an HTML standard as you can, for example XHTML. This ensures that your comments can be interpreted by any browser. In addition to standard HTML, you can use the following tags: CLASS, METHOD, PROPERTY, PARAMETER, QUERY, and EXAMPLE. (As with standard HTML tags, the names of these tags are not case-sensitive.) The most commonly used tags are described here. See the documentation for %CSP.Documatic
for details of the others.
Use to tag class names. If the class exists, the contents are displayed as a link to the class' documentation. For example:
/// This uses the <CLASS>Sample.Person</CLASS> class.
Use to tag programming examples. This tag affects the appearance of the text. Note that each ///
line becomes a separate line in the example (in contrast to the usual case, where the lines are combined into a single paragraph). For example:
/// set o=..%New()
/// set o.MyProperty=42
/// set o.OtherProp="abc"
/// do o.WriteSummary()
Use to tag method names. If the method exists, the contents are displayed as a link to the method's documentation. For example:
/// This is identical to the <METHOD>Unique</METHOD> method.
Use to tag property names. If the property exists, the contents are displayed as a link to the property's documentation. For example:
/// This uses the value of the <PROPERTY>State</PROPERTY> property.
Here is a multi-line description using HTML markup:
/// The <METHOD>Factorial</METHOD> method returns the factorial
/// of the value specified by <VAR>x</VAR>.
Caché class definitions are compiled into application routines by the class compiler. Classes cannot be used in an application before they are compiled.
The class compiler differs from the compilers available with other programming languages, such as C++ or Java, in two significant ways: first, the results of compilation are placed into a shared repository (database), not a file system. Second, it automatically provides support for persistent classes.
Specifically, the class compiler does the following:
It generates a list of dependencies
classes that must be compiled first. Depending on the compile options used, any dependencies that have been modified since last being compiled will also be compiled.
It resolves inheritance
it determines which methods, properties, and other class members are inherited from superclasses. It stores this inheritance information into the class dictionary for later reference.
For persistent and serial classes, it determines the storage structure needed to store objects in the database and creates the necessary runtime information needed for the SQL representation of the class.
It creates one or more routines that contain the runtime code for the class. The class compiler groups methods according to language (ObjectScript and Basic) and generates separate routines, each containing methods of one language or the other.
If you specify the Keep Generated Source
option with the class compiler, you can view the source for the routines using the
command (from the
menu) within Studio.
It compiles all of the generated routines into executable code.
It creates a class descriptor. This is a special data structure (stored as a routine) that contains all the runtime dispatch information needed to support a class (names of properties, locations of methods, and so on).
There are several ways to invoke the class compiler:
From within Studio using the option in the
If you use SQL DDL statements to create a table, the class compiler is automatically invoked to compile the persistent class that corresponds to the table.
The signature of this method is as follows:
classmethod GetDependencies(ByRef class As %String, Output included As %String, qspec As %String) as %Status
is either a single class name (as in the example), a comma-separated list of class names, or a multidimensional array of class names. (If it is a multidimensional array, be sure to pass this argument by reference.) It can also include wildcards.
is a multidimensional array of the names of the classes that will be compiled when class
is a string of compiler flags and qualifiers. See the next subsection. If you omit this, the method considers the current compiler flags and qualifiers.
method also allows you to supply flags and qualifiers that affect the result. Their position in the argument list is described in the explanation of the Compile()
method. To view the applicable flags, execute the command:
This produces the following output:
b - Include sub classes.
c - Compile. Compile the class definition(s) after loading.
d - Display. This flag is set by default.
e - Delete extent.
h - Generate help.
i - Validate XML export format against schema on Load.
k - Keep source. When this flag is set, source code of
generated routines will be kept.
l - Lock classes while compiling. This flag is set by default.
p - Percent. Include classes with names of the form %*.
r - Recursive. Compile all the classes that are dependency predecessors.
s - Process system messages or application messages.
u - Update only. Skip compilation of classes that are already up-to-date.
y - Include classes that are related to the current class in the way that
they either reference to or are referenced by the current class in SQL usage.
These flags are deprecated a, f, g, o, q, v
Default flags for this namespace =dil
You may change the default flags with the SetFlags(flags,system) classmethod.
To view the full list of qualifiers, along with their description, type, and any associated values, execute the command:
Qualifier information displays in a format similar to one of the following:
Description: Validate imported XML files against the schema definition.
Default Value: 1
Description: Check system classes for up-to-dateness
Default Value: 0
Description: Skip classes or expanded classes that are up-to-date.
Enum List: none,all,expandedonly,0,1
Default Value: expandedonly
Present Value: all
Negated Value: none
If the compiler is called while an instance of the class being compiled is open, there is no error. The already open instance continues to use its existing code. If another instance is opened after compilation, it uses the newly compiled code.
You might want to make some of your classes deployed before you send them to customers; this process hides the source code.
For any class definitions that contain method definitions that you do not want customers to see, compile the classes and then use $SYSTEM.OBJ.MakeClassDeployed()
. For example:
When a class is in deployed mode, its method and trigger definitions have been removed. (Note that if the class is a data type class, its method definitions are retained because they may be needed at runtime by cached queries.)
You can open the class definition in Studio, but it is read-only.
You cannot export or compile a deployed class, but you can compile its subclasses (if they are not deployed).
There is no way to reverse or undo deployment of a class. You can, however, replace the class by importing the definition from a file, if you previously exported it. (This is useful if you accidentally put one of your classes into deployed mode prematurely.)
© 1997-2019 InterSystems Corporation, Cambridge, MA