This article discusses the configuration and data required to define schema elements in a Building Block.
Schema definitions are contained within a simple directory structure with a specific layout including table definitions, SQL scripts for various stages of the update process, views, functions, stored procedures, etc.
- Schema.xml - A required file that defines the table structure to install in the Learn database.
- datatemplates - a directory containing delimited files used as seed data for tables defined in schema.xml
- functions, stored-procedures, triggers, views - user-defined functions, stored procedures, triggers, and views, respectively
- pre_update_sql - Scripts that are run before DDL updates
- post_schema_update_sql - Scripts that are run after DDL updates
- post_update_sql - Scripts that are run after post_schema_update_sql
Note that for all objects defined from files (functions, triggers, stored procedures, and views), there are a handful of constraints on the file's structure -
- Only a single object may be defined per file
- The file name must be the same as the object name (with the exception of the SQL and db-type extension)
- On Oracle, create statements do not need to use the or replace clause, as the installation code automatically drops all objects, based on the file name
Each directory containing raw SQL objects (view definitions, pre- and post-update scripts, etc.) must include a manifest.txt file to specify object creation or script execution order in order to ensure compilation dependencies are met. The manifest is simply a list of newline-separated script names (without the database-specific suffix).
The main file, schema.xml is in the root of the directory, and is a very straightforward representation of the application's table structure. Keep in mind it is built to support the set of data definitions that Blackboard supports in its database implementation, so there is not necessarily fine-grained control over the creation of things like indexes.
The root element is the "schema" - it is a simple container, with no processed attributes. As with all elements, <schema> may be commented via attribute or element. Except for comment, the only other element that appears under schema is table
The basic table definition is very simple - a single attribute, "name". Tables may contain the following elements:
- column. One or more columns may be defined.
- index. Any number of indexes may be defined.
- primary-key. Only a single primary key may be defined for the table.
- foreign-key. Any number of foreign keys may be defined.
|Table names MUST be prefixed with the vendor-id from the bb-manifest.xml|
Each table must contain one or more columns to define the data that can be stored.
|name||Required. The column name|
|data-type||Required. The type to use for the column; the schema file uses a generic syntax based on SQL Server, that gets translated to a database-specific type.|
|nullable||Flags whether the column can contain SQL NULL values. Default is 'true'.|
|default||The value to insert in the database if none is supplied from the application tier. If the type is a string, this attribute is a literal AND MUST INCLUDE SINGLE QUOTES.|
|identity||Flags whether the column should be treated as an identity column. Default is 'false'. On SQL Server, the column will get created as an auto-increment identity column. On Oracle, a sequence will be created.|
The following table shows the data types that may be referenced in the column definitions, and correspond to data types in use by the Blackboard application. Some of the types, such as "int" are used as synonyms for an underlying type commonly used in the Learn application, and don't support the same specifiers that may be used when creating a similar column using raw DDL. The conventions already in use by the Blackboard database affect many of these specifications - for example, because of difference in the database characterset and the NLS characterset, text fields must be explicitly designed for single- vs. multi-byte data.
|nvarchar||Used for text that may contain multi-byte characters. All user-entered data should use this type. Additionally, a length specifier is required (e.g., nvarchar(50)), just as if the type were being directly defined with DDL.|
|varchar||Used for text that will only contain ASCII or ISO-8859-1 characters. This is only recommended for fields that will store known inputs. As with nvarchar, a length specifier is required.|
|int||An integer value, mapped to numeric(38) in Oracle and 'int' on SQL Server|
|char||A fixed length, non-internationalized string. A length specifier must be provided.|
The following represents a sample of data type conventions used by Blackboard-defined schema.
- Mapping boolean values. Data type char(1), with a constraint limiting the values to Y/N, named with a _ind (for "indicator") suffix.
- Primary keys. Data type int, name with _pk1 suffix.
Constraints on the columns values are defined with the value-constraint element. A single, required name attribute must be provided to identify the constraint. Each accepted value is defined with a child accepted-value element, with a single, mandatory value attribute. Unlike default values defined on columns, constraint values do not need the quote literals.
|name||Required. The name of the constraint that will be created in the database.|
Primary keys must contain one columnref subelement that includes a single attribute, name, that references the column name to include in the primary key. In SQL Server, the key is mapped as an identity field with an auto-incremented value. On Oracle, a sequence is automatically created with the table name plus _seq suffix.
|Keep that in mind when creating table names, as there is an Oracle limit of 30 characters for object names. So, in practice, names defined in the schema XML must be shorter than 26 characters (because the automatically applied suffixes will extend the object name)|
Indexes may be defined via the index element.
|name||An identifier for the index; Blackboard naming conventions typically include the table name with a suffix of 'if', 'ak', or 'ie', followed by a number.|
|unique||True/false flag indicating whether a uniqueness constraint should be applied|
References to data stored in other tables are defined via the foreign-key element.
|name||Required. An identifier for the referential integrity constraint created by the key. The Blackboard naming convention is table name, followed by 'fk', followed by a numeric suffix to distinguish multiple constraints. E.g., announcements_fk2.|
|reference-table||Required. Indicates which table is referenced by columnref.|
|on-delete||Indicates what action should be taken if the referenced column is deleted. Valid values are setnull, which means the column should be set to null, and delete, which means the row should be deleted. If setnull is specified, the column specified in columnref must have nullable="true".|
|Although on-delete is not a required attribute, it is very important to consider how a core Blackboard table is being referenced. Failure to specify an appropriate on-delete action could result in core functionality breaking (e.g., it could cause deletion of a core object to fail).|
Comments may be included on any element, via a comment element or attribute. The main factor in which kind of comment is used is simply source readability. Longer comments should be created via elements. In either case, HTML must be escaped, as there is no defined namespace. While no validation is used in processing schema XML and HTML tags would not cause a failure, escaping/encoding is considered a best practice in order to avoid potential markup conflicts ("table" for example).
The Blackboard platform supports both SQL Server and Oracle. While the schema definition framework transparently converts the generic XML syntax to the correct flavor, the same is not true for SQL scripts. That means developers should provide a version for the database versions supported using a simple extension to indicate the target database: db-oracle for Oracle-specific scripts and db-mssql for SQL Server-specific scripts.