JDBC Translator

The JDBC translator bridges between SQL semantic and data type difference between Teiid and a target RDBMS. Teiid has a range of specific translators that target the most popular open source and proprietary databases.

jdbc-ansi

Declares support for most SQL constructs supported by Teiid, except for row limit/offset and EXCEPT/INTERCECT. Translates source SQL into ANSI compliant syntax. This translator should be used when another more specific type is not available.

jdbc-simple

Same as jdbc-ansi, except disables support for function, UNION, and aggregate pushdown.

access

For use with Microsoft Access 2003 or later. If used in a dynamic vdb the importer defaults to importKeys=false and excludeTables=.*[.]MSys.* to avoid issues with the metadata provided by the JDBC ODBC bridge. You may need to adjust these values if you use a different JDBC driver.

db2

For use with DB2 8 or later and DB2 for i 5.4 or later.

DB2 specific execution properties:

• DB2ForI- indicates that the the DB2 instance is DB2 for i. Defaults to false.
• supportsCommonTableExpressions- indicates that the DB2 instance supports Common Table Expressions. Defaults to true. Some older versions, or instances running in a conversion mode, of DB2 lack full common table expression support and may need support disabled.

derby

For use with Derby 10.1 or later.

excel-odbc

For use with Excel 2003 or later via the JDBC-ODBC bridge.

greenplum

For use with the Greenplum database.

h2

For use with H2 version 1.1 or later.

hana

For use with SAP Hana.

Apache HBase Translator

The Apache HBase Translator is exposes querying functionality to HBase Tables. Apache Phoenix is a JDBC SQL interface for HBase - see Phoenix Data Sources that is required for this translator as it pushes down commands into Phoenix SQL.

The DatabaseTimezone property should not be used with this translator.

The HBase Translator doesn't support Joins. Phoenix uses the HBase Table Row ID as the Primary Key, which map to . This Translator is developed with Phoenix 4.3+ for HBase 0.98.1+.

hive

For use with Hive v.10 and SparkSQL v1.0 and later.

Hive has limited support for data types. It is does not have native support for time/xml or LOBs. These limitations are reflected in the translator capabilities. A Teiid view can use these types, however the transformation would need to specify the necessary conversions. Note that in those situations, the evaluations will be done in Teiid engine.

Hive only supports EQUI join, so using any other joins types on its source tables will result in inefficient queries.

To write criteria based on partitioned columns, modeled them on source table, but do not include them in selection columns.

The Hive importer does not have concept of catalog or source schema, nor does it import keys, procedures, indexes, etc.

Hive specific importer properties:

• trimColumnNames- For Hive 0.11.0 and later the the DESCRIBE command metadata is inappropriately returned with padding, set to true to strip trim white space from column names. Defaults to false.

*useDatabaseMetaData- For Hive 0.13.0 and later the normal JDBC DatabaseMetaData facilities are sufficient to perform an import. Set to true to use the normal import logic with the option to import index information disabled. Defaults to false. When true, trimColumnNames has no effect.

If false the typical JDBC DatabaseMetaData calls are not used so not all of the common JDBC importer properties are applicable to Hive. You may still use excludeTables regardless.

"Database Name" When the database name used in the Hive is different than "default", the metadata retrieval and execution of queries does not work as expected in Teiid, as Hive JDBC driver seems to be implicitly connecting (tested with < 0.12) to "default" database, thus ignoring the database name mentioned on connection URL. This can workaround in the Teiid in JBoss AS environment by setting the following in data source configuration. <new-connection-sql>use {database-name}</new-connection-sql> This is fixed in > 0.13 version Hive Driver. See https://issues.apache.org/jira/browse/HIVE-4256

hsql

For use with HSQLDB 1.7 or later.

impala

For use with Cloudera Impala 1.2.1 or later.

Impala has limited support for data types. It is does not have native support for time/date/xml or LOBs. These limitations are reflected in the translator capabilities. A Teiid view can use these types, however the transformation would need to specify the necessary conversions. Note that in those situations, the evaluations will be done in Teiid engine.

Impala only supports EQUI join, so using any other joins types on its source tables will result in inefficient queries.

To write criteria based on partitioned columns, modeled them on source table, but do not include them in selection columns.

Impala Hive importer does not have concept of catalog or source schema, nor does it import keys, procedures, indexes, etc.

Impala specific importer properties:
*useDatabaseMetaData- Set to true to use the normal import logic with the option to import index information disabled. Defaults to false.

If false the typical JDBC DatabaseMetaData calls are not used so not all of the common JDBC importer properties are applicable to Impala. You may still use excludeTables regardless.

Some versions of Impala requires the use of a LIMIT when performing an ORDER BY. If no default is configured in Impala, then an exception can occur when a Teiid query with an ORDER BY but no LIMIT is issued. You should set an Impala wide default, or configure the connection pool to use a new connection sql string to issue a SET DEFAULT_ORDER_BY_LIMIT statement. See the Cloudera docs for more on limit options - such as controlling what happens when the limit is exceeded.

ingres

For use with Ingres 2006 or later.

ingres93

For use with Ingres 9.3 or later.

intersystems-cache

For use with Intersystems Cache Object database (only relational aspect of it)

informix

For use with any Informix version.

metamatrix

For use with MetaMatrix 5.5.0 or later.

modeshape

For use with Modeshape 2.2.1 or later. The PATH, NAME, LOCALNODENAME, DEPTH, and SCORE functions should be accessed as pseudo-columns, e.g. "nt:base"."jcr:path". Teiid UFDs (prefixed by JCR_) are available for CONTIANS, ISCHILDNODE, ISDESCENDENT, ISSAMENODE, REFERENCE - see the JCRFunctions.xmi. If a selector name is needed in a JCR function, you should use the pseudo-column "jcr:path", e.g. JCR_ISCHILDNODE(foo.jcr_path, 'x/y') would become ISCHILDNODE(foo, 'x/y') in the ModeShape query. An additional pseudo-column "mode:properties" should be imported by setting the ModeShape JDBC connection property teiidsupport=true. The column "mode:properties" should be used by the JCR_REFERENCE and other functions that expect a .* selector name, e.g. JCR_REFERENCE(nt_base.jcr_properties) would become REFERENCE("nt:base".*) in the ModeShape query.

mysql mysql5

For use with MySQL version 4.x and 5 or later respectively. Also compatible with backwards compatible MySQL derivatives, including MariaDB.

The MySQL Translators expect the database or session to be using ANSI mode. If the database is not using ANSI mode, an initialization query should be used on the pool to set ANSI mode:

set SESSION sql_mode = 'ANSI'

If you may deal with null timestamp values, then set the connection property zeroDateTimeBehavior=convertToNull. Otherwise you'll get conversion errors in Teiid that '0000-00-00 00:00:00' cannot be converted to a timestamp.

netezza

For use with any Netezza version.

The current Netezza vendor supplied JDBC driver performs poorly with single transactional updates. As is generally the case when possible use batched updates.

Netezza specific execution properties:

• SqlExtensionsInstalled- indicates that SQL Extensions including support fo REGEXP_LIKE are installed. Defaults to false.

oracle

For use with Oracle 9i or later.

Sequences may be used with the Oracle translator. A sequence may be modeled as a table with a name in source of DUAL and columns with the name in source set to{{<sequence name>.[nextval|currval].}}

CREATE FOREIGN TABLE seq (nextval integer OPTIONS (NAMEINSOURCE 'seq.nextval'), currval integer options (NAMEINSOURCE 'seq.currval') ) OPTIONS (NAMEINSOURCE 'DUAL')

With Teiid 8.5 it's no longer necessary to rely on a table representation and Oracle specific handling for sequences. See DDL Metadata for representing currval and nextval as source functions.

You can also use a sequence as the default value for insert columns by setting the column to autoincrement and the name in source to <element name>:SEQUENCE=<sequence name>.<sequence value>.

A rownum column can also added to any Oracle physical table to support the rownum pseudo-column. A rownum column should have a name in source of rownum. These rownum columns do not have the same semantics as the Oracle rownum construct so care must be taken in their usage.

Oracle specific importer properties:

• useGeometryType- Use the Teiid Geomety type when importing columns with a source type of SDO_GEOMETRY. Defaults to false.
• useIntegralTypes- Use integral types rather than decimal when the scale is 0. Defaults to false.

Oracle specific execution properties:

• OracleSuppliedDriver- indicates that the Oracle supplied driver (typically prefixed by ojdbc) is being used. Defaults to true. Set to false when using DataDirect or other Oracle JDBC drivers.

Geo Spatial function support

Oracle translator supports geo spatial functions. The supported functions are:

Relate = sdo_relate

CREATE FOREIGN FUNCTION sdo_relate (arg1 string,  arg2 string,  arg3 string) RETURNS string;
CREATE FOREIGN FUNCTION sdo_relate (arg1 Object,  arg2 Object,  arg3 string) RETURNS string;
CREATE FOREIGN FUNCTION sdo_relate (arg1 string,  arg2 Object,  arg3 string) RETURNS string;
CREATE FOREIGN FUNCTION sdo_relate (arg1 Object,  arg2 string,  arg3 string) RETURNS string;

Nearest_Neighbor = sdo_nn

CREATE FOREIGN FUNCTION sdo_nn (arg1 string,  arg2 Object,  arg3 string,  arg4 integer) RETURNS string;
CREATE FOREIGN FUNCTION sdo_nn (arg1 Object,  arg2 Object,  arg3 string,  arg4 integer) RETURNS string;
CREATE FOREIGN FUNCTION sdo_nn (arg1 Object,  arg2 string,  arg3 string,  arg4 integer) RETURNS string;

Within_Distance = sdo_within_distance

CREATE FOREIGN FUNCTION sdo_within_distance (arg1 Object,  arg2 Object,  arg3 string) RETURNS string;
CREATE FOREIGN FUNCTION sdo_within_distance (arg1 string,  arg2 Object,  arg3 string) RETURNS string;
CREATE FOREIGN FUNCTION sdo_within_distance (arg1 Object,  arg2 string,  arg3 string) RETURNS string;

Nearest_Neigher_Distance = sdo_nn_distance

CREATE FOREIGN FUNCTION sdo_nn_distance (arg integer) RETURNS integer;

Filter = sdo_filter

CREATE FOREIGN FUNCTION sdo_filter (arg1 Object,  arg2 string,  arg3 string) RETURNS string;
CREATE FOREIGN FUNCTION sdo_filter (arg1 Object,  arg2 Object,  arg3 string) RETURNS string;
CREATE FOREIGN FUNCTION sdo_filter (arg1 string,  arg2 object,  arg3 string) RETURNS string;

postgresql

For use with 8.0 or later clients and 7.1 or later server.

PostgreSQL specific execution properties:

• PostGisVersion- indicate the PostGIS version in use. Defaults to 0 meaning PostGIS is not installed. Will be set automatically if the database version is not set.
• _ProjSupported- boolean indicating if Proj is support for PostGis. Will be set automatically if the database version is not set.

prestodb

The PrestoDB translator, known by the type name prestodb, exposes querying functionality to PrestoDB Data Sources. In data integration respect, PrestoDB has very similar capabilities of Teiid, however it goes beyond in terms of distributed query execution with multiple worker nodes. Teiid's execution model is limited to single execution node and focuses more on pushing the query down to sources. Currently Teiid has much more complete query support and many enterprise features.

The PrestoDB translator supports only SELECT statements with a restrictive set of capabilities. This translator is developed with 0.85 version of PrestoDB and capabilities are designed for this version. With new versions of PrestoDB Teiid will adjust the capabilities of this translator. Since PrestoDB exposes a relational model, the usage of this is no different than any RDBMS source like Oracle, DB2 etc. For configuring the PrestoDB consult the PrestoDB documentation.

sqlserver

For use with SQL Server 2000 or later. A SQL Server JDBC driver version 2.0 or later (or compatible e.g. JTDS 1.2 or later) should be used.  The SQL Server DatabaseVersion property may be set to 2000, 2005, 2008, or 2012, but otherwise expects a standard version number - e.g. "10.0".

SQL Server specific execution properties:

• JtdsDriver- indicates that the open source JTDS driver is being used. Defaults to false.

sybase

For use with Sybase version 12.5 or later.

If used in a dynamic vdb and no import properties are specified (not recommended, see import properties below), then exceptions can be thrown retrieving system table information. You should specify a schemaPattern or use excludeTables to exclude system tables if this occurs.

If the name in source metadata contains quoted identifiers (such as required by reserved words or words containing characters that would not otherwise be allowed) and you are using a jconnect Sybase driver, you must first configure the connection pool to enable quoted_identifier:

jdbc:sybase:Tds:host.at.some.domain:5000/db_name?SQLINITSTRING=set quoted_identifier on

If you are a jconnect Sybase driver and will target the source for dependent joins, you should allow the translator to send more values by setting the JCONNECT_VERSION. Otherwise you will get exceptions with statements that have more than 481 bind values:

jdbc:sybase:Tds:host.at.some.domain:5000/db_name?SQLINITSTRING=set quoted_identifier on&JCONNECT_VERSION=6

Sybase specific execution properties:

• JtdsDriver- indicates that the open source JTDS driver is being used. Defaults to false.

sybaseiq

For use with Sybase IQ version 15.1 or later.

teiid

For use with Teiid 6.0 or later.

teradata

For use with Teradata V2R5.1 or later.

With Teradata driver version 15 date, time, and timestamp values by default will be adjusted for the Teiid server timezone. To remove this adjustment, set the translator DatabaseTimezone property to GMT or whatever the Teradata server defaults to.

vertica

For use with Vertica 6 or later.

Actian Vector in Hadoop

For use Actian Vector in Hadoop, download the JDBC driver at http://esd.actian.com/platform. Note the port number in connection URL is "AH7" which maps to 16967.

Execution Properties - shared by all JDBC Translators

Importer Properties - shared by all JDBC Translators

When specifying the importer property, it must be prefixed with "importer.". Example: importer.tableTypes

[1] JavaDoc for DatabaseMetaData
[2] The fully qualified name for exclusion is based upon the settings of the translator and the particulars of the database. All of the applicable name parts used by the translator settings (see useQualifiedName and useCatalogName) including catalog, schema, table will be combined as catalogName.schemaName.tableName with no quoting. For example Oracle does not report a catalog, so the name used with default settings for comparison would be just schemaName.tableName.

Warning The default import settings will crawl all available metadata. This import process is time consuming and full metadata import is not needed in most situations. Most commonly you'll want to limit the import by at least schemaPattern and tableTypes.

Example importer settings to only import tables and views from my-schema. See also VDB Definition

<model ...

  <property name="importer.tableTypes" value="TABLE,VIEW"/>
  <property name="importer.schemaPattern" value="my-schema"/>
  ...
</model>

Usage

Usage of a JDBC source is straight-forward. Using Teiid SQL, the source may be queried as if the tables and procedures were local to the Teiid system.

Native Queries

Physical tables, functions, and procedures may optionally have native queries associated with them.  No validation of the native query is performed, it is simply used in a straight-forward manner to generate the source SQL.  For a physical table setting the teiid_rel:native-query extension metadata will execute the native query as an inline view in the source query.  This feature should only be used against sources that support inline views.  The native query is used as is and is not treated as a parameterized string. For example on a physical table y with nameInSource="x" and teiid_rel:native-query="select c from g", the Teiid source query"SELECT c FROM y" would generate the SQL query "SELECT c FROM (select c from g) as x".  Note that the column names in the native query must match the nameInSource of the physical table columns for the resulting SQL to be valid.

For physical procedures you may also set the teiid_rel:native-query extension metadata to a desired query string with the added ability to positionally reference IN parameters - see Parameterizable Native Queries.  The teiid_rel:non-prepared extension metadata property may be set to false to turn off parameter binding.  Note this option should be used with caution as inbound may allow for SQL injection attacks if not properly validated.  The native query does not need to call a stored procedure.  Any SQL that returns a result set positionally matching the result set expected by the physical stored procedure metadata will work.  For example on a stored procedure x with teiid_rel:native-query="select c from g where c1 = $1 and c2 = '$$1'", the Teiid source query "CALL x(?)" would generate the SQL query "select c from g where c1 = ? and c2 = '$1'".  Note that ? in this example will be replaced with the actual value bound to parameter 1.

Direct Query Procedure

This feature is turned off by default because of the security risk this exposes to execute any command against the source. To enable this feature, override the execution property called SupportsDirectQueryProcedure to true.

By default the name of the procedure that executes the queries directly is native. Override the execution property DirectQueryProcedureName to change it to another name.

The JDBC translator provides a procedure to execute any ad-hoc SQL query directly against the source without Teiid parsing or resolving. Since the metadata of this procedure's results are not known to Teiid, they are returned as an object array. ARRAYTABLE can be used construct tabular output for consumption by client applications.

SELECT x.* FROM (call jdbc_source.native('select * from g1')) w,
 ARRAYTABLE(w.tuple COLUMNS "e1" integer , "e2" string) AS x

SELECT x.* FROM (call jdbc_source.native('insert into g1 (e1,e2) values (?, ?)', 112, 'foo')) w,
 ARRAYTABLE(w.tuple COLUMNS "update_count" integer) AS x

SELECT x.* FROM (call jdbc_source.native('update g1 set e2=? where e1 = ?','blah', 112)) w,
 ARRAYTABLE(w.tuple COLUMNS "update_count" integer) AS x

SELECT x.* FROM (call jdbc_source.native('delete from g1 where e1 = ?', 112)) w,
 ARRAYTABLE(w.tuple COLUMNS "update_count" integer) AS x

JCA Resource Adapter

The resource adapter for this translator provided through data source in JBoss AS, Refer to Admin Guide for "JDBC Data Sources" configuration section.