ODBC Data Wrapper Reference

ODBC HeavyConnect and import is currently a beta feature.

This topic provides setup and configuration details and information about supported data types when using ODBC HeavyConnect or import. This reference also describes ODBC data wrapper assumptions and supported mappings of HeavyDB column types to ODBC column types, and provides RDMS-specific notes.

ODBC Driver Installation

To HeavyConnect to or import from an external RDMS using the ODBC data wrapper, make sure that an appropriate ODBC driver is installed for the RDMS you want to access. RDMS vendors websites usually provide ODBC driver download and configuration instructions, and in certain cases, ODBC drivers are managed by third-party vendors that provide these instructions.

ODBC Configuration Files

Two key configuration files need to be set up when using ODBC HeavyConnect or import: odbcinst.ini and odbc.ini.

The odbcinst.ini file specifies a list of installed drivers, along with required configuration for those drivers (the path to the driver binary). The following example is an odbcinst.ini configuration file for an instance with installed drivers for PostGreSQL, Redshift, and Snowflake:

[ODBC Drivers]
PostgreSQL=Installed
Redshift=Installed
Snowflake=Installed

[PostgreSQL]
Description=PostgreSQL ODBC driver
Driver=/usr/lib/x86_64-linux-gnu/odbc/psqlodbca.so

[Redshift]
Description=Redshift ODBC driver
Driver=/opt/amazon/redshiftodbc/lib/64/libamazonredshiftodbc64.so

[Snowflake]
Description=Snowflake ODBC Driver
Driver=/usr/lib/snowflake/odbc/lib/libSnowflake.so

Use an intuitive name for the drivers, such as the official name of the RDMS, so that users can easily know the driver names to use in connection strings.

The odbc.ini file contains configuration for database-specific connections. Each connection configuration is identified by a data source name (DSN). The following example is an odbc.ini file with configuration for the default postgres database on a local instance of PostgreSQL running on port 5432:

[ODBC Data Sources]
default_postgres_db=default_postgres_db

[default_postgres_db]
Description=Local default PostgreSQL database
Driver=PostgreSQL
Database=postgres
Servername=localhost
Port=5432

Locate odbcinst.ini and odbc.ini in the /etc/ directory, or in the home directory for cases where the server process is started under a specific user account.

The PostgreSQL unicode ODBC driver is currently not supported.

Setting Up ODBC Connectors for Heavy Immerse

In certain cases, an odbcinst.ini configuration file that omits the [ODBC Drivers] section is automatically generated when ODBC drivers (such as the PostgreSQL driver) are installed. This section of the configuration file is required for the ODBC-related connectors to appear in Heavy Immerse.

In addition, the driver names specified in odbcinst.ini must match the RDMS name for connectors to appear in Immerse. Specifically, the driver name should be PostgreSQL for PostgreSQL, Redshift for AWS Redshift, and Snowflake for Snowflake.

Working with HEAVY.AI Docker Images

For docker deployments of HEAVY.AI, ODBC driver installations can be managed in a Dockerfile that uses the HEAVY.AI Docker image as a base. For example, you can create a docker image with installed PostGreSQL, Redshift, and Snowflake ODBC drivers using a Dockerfile with the following content.

Always consult the license and terms and conditions of the ODBC drivers you install. The examples here are provided for illustrative purposes only.

FROM heavyai/heavyai-ee-cuda

# Install PostGreSQL ODBC driver.
# The Snowflake ODBC driver depends on unixodbc.
RUN apt-get update && apt-get install -y odbc-postgresql unixodbc

# Install Redshift ODBC driver.
RUN wget https://s3.amazonaws.com/redshift-downloads/drivers/odbc/1.4.52.1000/AmazonRedshiftODBC-64-bit-1.4.52.1000-1.x86_64.deb
RUN dpkg -i ./AmazonRedshiftODBC-64-bit-1.4.52.1000-1.x86_64.deb
RUN rm ./AmazonRedshiftODBC-64-bit-1.4.52.1000-1.x86_64.deb

# Install Snowflake ODBC driver.
RUN wget https://sfc-repo.snowflakecomputing.com/odbc/linux/latest/snowflake-odbc-2.25.2.x86_64.deb
RUN dpkg -i ./snowflake-odbc-2.25.2.x86_64.deb
RUN rm ./snowflake-odbc-2.25.2.x86_64.deb

You can then build a new image with installed ODBC drivers using the Dockerfile.

The odbcinst.ini and odbc.ini configuration files should be bound to /etc/odbcinst.ini and /etc/odbc.ini, respectively, when running the container.

ODBC Data Wrapper Assumptions

  • The remote database management system supports the MIN, MAX, and LENGTH SQL functions. These functions are used internally for optimizations and validations.

  • The combination of the query specified in the foreign table’s SQL_SELECT option and the SQL_ORDER_BY option is guaranteed to return a result set in the same order every time the query is executed.

  • Queries for geo columns are always returned as WKT strings. This implies that either geo columns are stored as text columns containing WKT strings, geo columns are cast to text (for example, using a function like ST_AsText) in the SQL_SELECT query, or the remote database management system, by default, projects geo columns as WKT strings.

HeavyDB to ODBC Data Type Mapping

Numeric and Boolean Types

HeavyDB (Down) \ ODBC (Right)SQL_BIGINTSQL_INTEGERSQL_SMALLINTSQL_TINYINTSQL_DOUBLESQL_FLOATSQL_REALSQL_DECIMALSQL_NUMERICSQL_BIT

BIGINT

Yes (unsigned is not coercible)

Yes (if unsigned)

No

No

No

No

No

Coercible

No

No

BIGINT ENCODING FIXED (32) / INTEGER

Coercible

Yes

Yes (if unsigned)

No

No

No

No

Coercible

No

No

BIGINT ENCODING FIXED (16) / INTEGER ENCODING FIXED (16) / SMALLINT

Coercible

Coercible

Yes

Yes (if unsigned)

No

No

No

Coercible

No

No

BIGINT ENCODING FIXED (8) / INTEGER ENCODING FIXED (8) / SMALLINT ENCODING FIXED (8) / TINYINT

Coercible

Coercible

Coercible

Yes

No

No

No

Coercible

No

No

BOOLEAN

No

No

No

No

No

No

No

No

No

Yes

DECIMAL (Precision, Scale)

No

No

No

No

No

No

No

Yes (if precision and scale are within OmniSci’s range)

Yes (if precision and scale are within OmniSci’s range)

No

DOUBLE

No

No

No

No

Yes

Coercible

No

No

No

No

FLOAT

No

No

No

No

Coercible

Coercible

Yes

No

No

No

*[1] PostgreSQL requires setting the option BoolsAsChar=false to represent boolean columns as SQL_BIT.

Date and Time Types

HeavyDB (Down) \ ODBC (Right)SQL_TYPE_DATESQL_TYPE_TIMESTAMPSQL_TYPE_TIMESQL_TYPE_UTCDATETIMESQL_TYPE_UTCTIME

DATE / DATE ENCODING DAYS (32)

Coercible

No

No

No

No

DATE ENCODING DAYS (16)

Coercible

No

No

No

No

TIME

No

No

Yes

No

No

TIME ENCODING FIXED (32)

No

No

Yes

No

No

TIMESTAMP (0)

No

Coercible

No

No

No

TIMESTAMP (3)

No

Coercible

No

No

No

TIMESTAMP (6)

No

Coercible

No

No

No

TIMESTAMP (9)

No

Coercible

No

No

No

TIMESTAMP ENCODING FIXED (32)

No

Coercible

No

No

No

Geo and String Types

HeavyDB (Down) \ ODBC (Right)SQL_CHARSQL_VARCHARSQL_LONGVARCHARSQL_WCHARSQL_WVARCHARSQL_WLONGVARCHAR

TEXT ENCODING DICT

Yes

Yes

Yes

Yes

Yes

Yes

TEXT ENCODING (16)

Yes

Yes

Yes

Yes

Yes

Yes

TEXT ENCODING (8)

Yes

Yes

Yes

Yes

Yes

Yes

TEXT ENCODING NONE

Yes

Yes

Yes

Yes

Yes

Yes

LINESTRING

No

Yes

Yes

No

No

No

MULTIPOLYGON

No

Yes

Yes

No

No

No

POINT

No

Yes

Yes

No

No

No

POLYGON

No

Yes

Yes

No

No

No

Array Type

Array data types currently are not supported.

See https://docs.microsoft.com/en-us/sql/odbc/reference/appendixes/sql-data-types for more details about ODBC data types, and consult the specific RDMS ODBC driver documentation for details about how the RDMS data types map to the ODBC data types.

RDMS-Specific Notes

PostgreSQL

  • The official PostgreSQL ODBC driver can be downloaded from https://odbc.postgresql.org/

  • Accessing BOOLEAN data in PostgreSQL requires the BoolsAsChar=false option to be added to the connection string or DSN configuration.

Google BigQuery

  • The BigQuery ODBC connector uses the OAuth 2.0 protocol for authentication; steps are provided by the official installation guide.

    • In all authentication methods, the Principal member (e.g. Service Account, User Account) must have at least a roles/bigquery.dataViewer permission grant on the GCP project.

    • Using a Service Account is the recommended method for authentication. Steps are provided in "Using a Google Service Account" in the installation guide.

    • If you are using a User Account for authentication, some steps may be missing from the installation guide in the "Using a Google User Account" section. Include the following steps:

Last updated