CUBA Platform. Developer’s Manual

The base entity classes and interfaces are described in detail in this section.

When creating entity classes it is recommended to choose a base class according to the following rules:

This section describes all annotations of entity classes and attributes supported by the platform.

Annotations from the javax.persistence package are required for JPA, annotations from com.haulmont.* packages are designed for metadata management and other mechanisms of the platform.

In this manual, if an annotation is identified by a simple class name, it refers to a platform class, located in one of the com.haulmont.* packages.

The standard use of JPA for enum attributes involves an integer database field containing a value obtained from the ordinal() method. This approach may lead to the following issues with extending a system in production:

CUBA-style approach to solving these problems is to detach the value stored in the database from ordinal value of the enumeration. In order to do this, the field of the entity should be declared with the type, stored in the database (Integer or String), while the access methods (getter / setter) should be created with the actual enumeration type.

Example:

In this case, the enumeration class can look like this:

For correct reflection in metadata, the enumeration class must implement the EnumClass interface.

As the examples show, grade attribute corresponds to the Integer type value stored in the database, which is specified by the id field of CustomerGrade enumeration, namely 10, 20 or 30. At the same time, the application code and metadata framework use CustomerGrade enum through access methods, which perform the actual conversion.

A call to getGrade() method will simply return null, if the value in the database does not correspond to any of the enumeration values. In order to add a new value, for example, HIGHER, between HIGH and PREMIUM, it is sufficient to add new enumeration value with id = 15, which ensures that sorting by Customer.grade field remains correct.

Enumeration values can be associated with localized names that will be displayed in the user interface of the application.

CUBA platform supports soft deletion mode, when the records are not deleted from the database, but instead, marked in a special way, so that they become inaccessible for common use. Later, these records can be either completely removed from the database using some kind of scheduled procedure or restored.

Soft deletion mechanism is transparent for an application developer, the only requirement is for entity class to implement SoftDelete interface. The platform will adjust data operations automatically.

Soft deletion mode offers the following benefits:

The negative impact of soft deletion is increase in database size and likely need for additional cleanup procedures.

Let us consider the basic metadata interfaces.

The main source for metadata structure generation are annotated entity classes.

Entity class will be present in the metadata in the following cases:

All entities inside same root package are put into the same MetaModel instance, which is given the name of this package. Entities within the same MetaModel can contain arbitrary references to each other. References between entities from different meta-models can be created in the order of declaration of metadata.xml files in cuba.metadataConfig property.

Entity attribute will be present in metadata if:

Metaclass and metaproperty parameters are determined on the base of the listed annotations parameters as well as field types and class methods. Besides, if an attribute does not have write access method (setter), it becomes immutable (read only).

Datatype interface describes a valid data type for the entity attribute if it is not a reference. Each Datatype implementation corresponds to a single Java class.

All of the instances are registered in repository – Datatypes class, which performs loading and initializing of Datatype implementation classes in the following way:

Datatype instance can be obtained in two ways:

Datatypes are associated with entity attributes during application start according to the following rules:

Basic methods of Datatype interfaces:

Datatype determines two sets of methods for formatting and parsing: considering and not considering locale. Conversion considering locale is applied everywhere in user interface, ignoring locale – in system mechanisms, for example, serialization in REST API.

Parsing formats ignoring locale are specified in the above mentioned datatypes.xml file.

The parsing formats considering locale are provided in the main localized messages pack, in the strings containing the following keys:

All the listed formats are specified in the main localized message pack of CUBA base projects by default, and can be overridden in the similar files of the application project.

Entity meta-annotations are a set of key/value pairs providing additional information about entities.

Meta-annotations are accessed using meta-class getAnnotations() method.

The sources of meta-annotations are:

A view can be created in two possible ways:

Let us consider in details the declarative way for creation and working with views.

ViewRepository is a Spring bean, accessible to all application blocks. The reference to ViewRepository can be obtained using injection or through the Metadata infrastructure interface. ViewRepository.getView() methods are used to retrieve view instances from the repository. deployViews() methods from AbstractViewRepository basic implementation are used to deploy XML descriptors to the repository.

Two views named _local and _minimal are available in the views repository for each entity by default:

The detailed structure of view XML descriptors is explained here.

The example below shows a view descriptor for the Order entity which provides loading of all local attributes, associated Customer and the Items collection:

The recommended way of grouping and deployment of view descriptors is as follows:

To create a managed bean, add the @org.springframework.stereotype.Component annotation to the Java class. For example:

It is recommended to assign a unique name to the bean in the {project_name}_{class_name} form and to define it in the NAME constant.

The managed bean class should be placed inside the package tree with the root specified in the context:component-scan element of the spring.xml file. In this case, the spring.xml file contains the element:

which means that the search for annotated beans for this application block will be performed starting with the com.sample.sales package.

Managed beans can be created on any tier, because the Spring Framework container is used in all standard blocks of the application.

A reference to the bean can be obtained through injection or through the AppBeans class. As an example of using the bean, let us look at the implementation of the OrderService bean that delegates the execution to the OrderWorker bean:

In this example, the service starts a transaction, merges the detached entity obtained from the client level into the persistent context, and passes the control to the OrderWorker bean, which contains the main business logic.

The following example shows how to create a JMX bean.

All JMX beans of a project are declared in one MBeanExporter instance in the map/entry elements of the beans property. The key is JMX ObjectName, the value – the bean’s name specified in the @Component annotation. ObjectName begins with the name of the web application, because several web applications, which export the same JMX interfaces, can be deployed into one application server instance (i.e., into one JVM).

This section describes some of the JMX beans available in the platform.

The interface helps to obtain references to configuration interfaces.

Examples:

Messages interface provides methods to get localized message strings.

Let us consider interface methods in detail.

Metadata interface provides access to metadata session and view repository.

Interface methods:

Resources interface maintains resources loading according to the following rules:

In practice, explicit identification of URL or classpath: prefix is rarely used, so resources are usually downloaded either from the configuration folder or from classpath. The resource in the configuration folder overrides the classpath resource with the same name.

Resources methods:

Scripting interface is used to compile and load Java and Groovy classes dynamically (i.e. at runtime) as well as to execute Groovy scripts and expressions.

Scripting methods:

Cache of the compiled classes can be cleaned at runtime using CachingFacadeMBean JMX bean.

See also ScriptingManagerMBean.

This interface provides authorization – checking user access rights to different objects in the system. Most of the interface methods delegate to the corresponding methods of current UserSession object, but before this they search for an original meta-class of the entity, which is important for projects with extensions. Besides methods duplicating UserSession functionality, this interface contains isEntityAttrReadPermitted() and isEntityAttrUpdatePermitted() methods that check attribute path availability with respect to availability of all attributes and entities included in the path.

The Security interface is recommended to use everywhere instead of direct calling of the UserSession.isXYXPermitted() methods.

See more in User Authentication.

TimeSource interface provides the current time. Using new Date() and similar methods in the application code is not recommended.

Examples:

The interface is used to obtain current user session object. See more in User Authentication.

The interface is used to obtain UUID values, including those used for entity identifiers. Using UUID.randomUUID() in the application code is not recommended.

To call from a static context, you can use the UuidProvider class, which also has an additional fromString() method that works faster than the standard UUID.fromString() method.

DataManager interface provides CRUD functionality on both middle and client tiers. It is a universal tool for loading entity graphs from the database and saving changes applied to detached entity instances.

DataManager always starts a new transaction and commits it on operation completion, thus returning entities in the detached state.

DataManager always applies security restrictions when invoked from a client, and ignores them by default when invoked from middleware code. If you want to check security when using DataManager in your middleware code, obtain a wrapper via DataManager.secure() method and call its methods. Alternatively, you can set the cuba.dataManagerChecksSecurityOnMiddleware application property to turn on security check for the whole application.

DataManager methods are listed below:

While loading data, DataManager may also implement additional functionality described below.

Properties that determine configuration and deployment parameters are specified in special property files named according to the *app.properties pattern. Each application block contains a set of such files, including files from the platform base projects. The set of property files is defined as follows:

For example, the set of property files of the Middleware block is specified in the web/WEB-INF/web.xml file of the core module and looks as follows:

The classpath: prefix means that the corresponding file can be found in the Java classpath, while file: prefix means that it should be loaded from the file system. Java system properties can be used: in this example, catalina.home is the Tomcat installation path.

An order in which files are declared is important because the values, specified in each subsequent file override the values of the properties with the same name, specified in the preceding files. This allows you to override platform properties in the application.

The last file in the above set is local.app.properties. It can be used to override application properties upon deployment. If the file does not exist, it is silently ignored. You can create this file on the application server and define all properties specific to the environment in it. As a result, the settings will be separated from the application, and you will be able to update the application without fear of losing the specific configuration information. The Using Tomcat in Production section contains an example of using the local.app.properties file.

For Desktop Client, JVM command line arguments serve as an equivalent of local.app.properties. In this block, the properties loader treats all the arguments containing "=" sign as a key/value pair and uses them to replace corresponding application properties specified in app.properties files.

Application properties that represent runtime parameters are stored in the SYS_CONFIG database table.

Such properties have the following distinctive features:

It is important to mention, that access to properties stored in the database on the client side leads to Middleware requests. This is less efficient than retrieving properties from local *app.properties files. To reduce the number of requests, the client caches properties for the lifetime of configuration interface implementation instance. Thus, if you need to access the properties of a configuration interface from some UI screen for several times, it is recommended to get the reference to this interface upon screen initialization and save it to a screen controller field for further access.

The configuration interfaces mechanism enables working with application properties using Java interface methods, providing the following benefits:

Example of reading the transaction timeout value in the Middleware block:

If injection is impossible, the configuration interface reference can be obtained via the Configuration infrastructure interface:

A message pack is a set of properties files with the names in format of messages{_XX}.properties located in a single Java package. XX suffix indicates the language of the messages in this file and corresponds to the language code in Locale.getLanguage(). It is also possible to use other Locale attributes, for example, country. In this case the pack file will look like messages{_XX_YY}.properties. One of the files in the pack can have no language suffix – it is the default file. The name of the message pack corresponds to the name of the Java package, which contains the pack files.

Let us consider the following example:

This pack consists of 3 files – one for Russian, one for French and a default file. The name of the pack is com.abc.sales.gui.customer.

Message files contain key/value pairs, where the key is the message identifier referenced by the application code, and the value is the message itself in the language of the file. The rules for matching pairs are similar to those of java.util.Properties properties files with the following specifics:

Messages are retrieved from the packs using Messages interface methods according to the following rules:

Each standard application block should have its own main message pack. For the client tier blocks the main message pack contains main menu entries and common UI elements names (for example, names of OK and Cancel buttons). The main pack also determines Datatype transformation formats for all application blocks, including Middleware.

cuba.mainMessagePack application property is used to specify the main message pack. The property value can be either a single pack or list of packs separated by spaces. For example:

In this case the messages in the second pack of the list will override those from the first pack. Thus, the messages defined in the base projects packs can be overridden in the application project.

To display localized names of the entities and attributes in UI, create special message packs in the Java packages containing the entities. Use the following format in message files:

The example of default English localization of com.abc.sales.entity.Customer entity – /com/abc/sales/entity/messages.properties file:

Such message packs are usually used implicitly by the framework, for example, by Table and FieldGroup visual components. Besides, you can obtain the names of the entities and attributes using the following methods:

To localize the enumeration names and values, add messages with the following keys to the message pack located in the Java package of the enumeration class:

For example, for enum

default English localization file /com/abc/sales/messages.properties should contain the following lines:

Localized enum values are automatically used by different visual components such as LookupField. You can obtain localized enum value programmatically: use getMessage() method of the Messages interface and simply pass the enum instance to it.

User session is the main element of access control mechanism of CUBA applications. It is represented by the UserSession object, which is associated with the currently authenticated user and contains information about user rights. The UserSession object can be obtained in any application block using the UserSessionSource infrastructure interface.

The UserSession object is created on Middleware during LoginService.login() method execution after the user is authenticated using a name and a password. The object is then cached in the Middleware block and returned to the client tier. When running in cluster, the session object is replicated to all cluster members. The client tier also stores the session object after receiving it, associating it with the active user in one way or another (for example, storing it in HTTP session). Further on, all Middleware invocations on behalf of this user are accompanied by passing the session identifier (of UUID type). This process does not need any special support in the application code, as the session identifier is passed automatically, regardless of the signature of invoked methods. Processing of client invocations in the Middleware starts from retrieving session from the cache using the obtained identifier. Then the session is associated with the request execution thread. The session object is deleted from the cache when the LoginService.logout() method is called or when the timeout defined by cuba.userSessionExpirationTimeoutSec application property expires.

Thus the session identifier created when the user logs into the system is used for user authentication during each Middleware invocation.

The UserSession object also contains methods for current user authorization – validation of the rights to system objects: isScreenPermitted(), isEntityOpPermitted(), isEntityAttrPermitted(), isSpecificPermitted(). However, it is recommended to use the Security infrastructure interface for programmatic authorization.

The UserSession object can contain named attributes of arbitrary serializable type. The attributes are set by setAttribute() method and returned by getAttribute() method. The latter is also able to return the following session parameters, as if they were attributes:

The session attributes are replicated in the Middleware cluster, same as the other user session data.

Standard user login process:

Password hashing algorithm is implemented by the EncryptionModule type bean and is specified in cuba.passwordEncryptionModule application property. SHA-1 is used by default.

The platform has a mechanism for the protection against password brute force cracking. The protection is enabled by the cuba.bruteForceProtection.enabled application property on Middleware. If the protection is enabled then the combination of user login and IP address is blocked for a time interval in case of multiple unsuccessful login attempts. A maximum number of login attempts for the combination of user login and IP address is defined by the cuba.bruteForceProtection.maxLoginAttemptsNumber application property (default value is 5). Blocking interval in seconds is defined by the cuba.bruteForceProtection.blockIntervalSec application property (default value is 60).

It is possible that the user password (actually, password hash) is not stored in the database, but is verified by external means, for example, by means of integration with LDAP. In this case the authentication is in fact performed by the client block, while the Middleware "trusts" the client by creating the session based on user login only, without the password, using LoginService.loginTrusted() method. This method requires satisfying the following conditions:

Login to the system is also required for scheduled automatic processes as well as for connecting to the Middleware beans using JMX interface. Formally, these actions are considered administrative and they do not require authentication as long as no entities are changed in the database. When an entity is persisted to the database, the process requires login of the user who is making the change so that the login of the user responsible for the changes is stored.

An additional benefit from login to the system for an automatic process or for JMX call is that the server log output is displayed with the current user login if the user session is set to the execution thread. This simplifies searching messages created by specific process during log parsing.

System access for the processes within Middleware is done using LoginWorker.loginSystem() call passing the login (without password) of the user on whose behalf the process will be executed. As result, UserSession object will be created and cached in the corresponding Middleware block but it will not be replicated in the cluster.

See more about processes authentication inside Middleware in System Authentication.

SecurityContext class instance stores information about the user session for the current execution thread. It is created and passed to AppContext.setSecurityContext() method in the following moments:

In the first two cases, SecurityContext is removed from the execution thread when the request execution is finished.

If you create a new execution thread from the application code, pass the current SecurityContext instance to it as in the example below:

The following rules should be followed when creating your own exception classes:

The platform contains a special unchecked exception class SilentException. It can be used to interrupt execution without showing any messages to the user or writing them to the log. SilentException is declared in the global module, and therefore is accessible both in Middleware and client blocks.

If an exception is thrown on Middleware as a result of handling a client request, the execution terminates and the exception object is returned to the client. The object usually includes the chain of underlying exceptions. This chain can contain classes which are inaccessible for the client tier (for example, JDBC driver exceptions). For this reason, instead of sending this chain to the client we send its representation inside a specially created RemoteException object.

The information about the causing exceptions is stored as a list of RemoteException.Cause objects. Each Cause object always contains an exception class name and its message. Moreover, if the exception class is "supported by client", Cause stores the exception object as well. This enables passing information to the client in the exception fields.

Exception class should be annotated by @SupportedByClient if its objects should be passed to the client tier as Java objects. For example:

Thus, when an exception is thrown on Middleware and it is not annotated by @SupportedByClient the calling client code will receive RemoteException containing original exception information in a string form. If the source exception is annotated by @SupportedByClient, the caller will receive it directly. This enables handling the exceptions declared by Middleware services in the application code in the traditional way – using try/catch blocks.

Bear in mind that if you need the exception supported by client to be passed on the client as an object, it should not contain any unsupported exceptions in its getCause() chain. Therefore, if you create an exception instance on Middleware and want to pass it to the client, specify cause parameter only if you are sure that it contains the exceptions known to the client.

ServiceInterceptor class is a service interceptor which packs the exception objects before passing them to the client tier. Besides, it performs exceptions logging. All information about the exception including full stack trace is logged by default. If it is not desirable, add @Logging annotation to the exception class and specify the logging level:

For example:

Unhandled exceptions in Web Client and Desktop Client blocks thrown on the client tier or passed from Middleware, are passed to the special handlers mechanism. This mechanism is implemented in the gui module and available for both blocks.

The handler should be a managed bean implementing the GenericExceptionHandler interface, handle processing in its handle() method and return true, or immediately return false, if this handler is not able to handle the passed exception. This behaviour enables creating a "chain of responsibility" for handlers.

It is recommended to inherit your handlers from the AbstractGenericExceptionHandler base class, which is able to disassemble the exceptions chain (including ones packed inside RemoteException) and handle specific exception types. Exception types supported by this handler are defined by passing strings array to the base constructor from the handler constructor. Each string of the array should contain one full class name of the handled exception, for example:

If the exception class is not accessible on the client side, specify its name with the string literal:

In the case of using AbstractGenericExceptionHandler as a base class, the processing logic is located in doHandle() method and looks as follows:

If the name of the exception class is insufficient to make a decision whether this handler can be applied to the exception, canHandle() method should be defined. This method accepts also the text of the exception. If the handler is applicable for this exception, the method must return true. For example:

In the application project, you can use any DBMS supported by the ORM framework (EclipseLink). Follow the steps below:

In addition to cuba.dbmsType application property, there is an optional cuba.dbmsVersion property. It affects the choice of interface implementations for DbmsFeatures, SequenceSupport, DbTypeConverter, and the search for database init and update scripts.

The name of the implementation class of the integration interface is constructed as follows: TypeVersionName. Here, Type is the value of the cuba.dbmsType property (capitalized), Version is the value of cuba.dbmsVersion, and Name is the interface name. The package of the class must correspond to that of the interface. If a class with the same name is not available, an attempt is made to find a class with the name without version: TypeName. If such class does not exist either, an exception is thrown.

For example, the com.haulmont.cuba.core.sys.persistence.Mssql2012SequenceSupport class is defined in the platform. This class will take effect if the following properties are specified in the project:

The search for database init and update scripts prioritizes the type-version directory over the type directory. This means that the scripts in the type-version directory replace the scripts with the same name in the type directory. The type-version directory can also contain some scripts with unique names; they will be added to the common set of scripts for execution, too. Script sorting is performed by path, starting with the first subdirectory of the type or type-version directory, i.e. regardless of the directory where the script is located (versioned or not).

For example, the init script for Microsoft SQL Server versions below and above 2012 should look as follows:

Create and update SQL scripts are text files with a set of DDL and DML commands separated by the "^" character. The "^" character is used, so that the ";" separator can be applied as part of complex commands; for example, when creating functions or triggers. The script execution mechanism splits the input file into separate commands using the "^" separator and executes each command in a separate transaction. This means that, if necessary, it is possible to group several single statements (e.g., insert), separated by semicolons and ensure that they execute in a single transaction.

An example of the update SQL script:

Groovy update scripts have the following structure:

The execution mechanism passes the following variables to the Groovy scripts:

An example of the Groovy update script: