PostgreSQL, HAWQ Interpreter for Apache Zeppelin

Name Class Description
%psql.sql PostgreSqlInterpreter Provides SQL environment for PostgreSQL, HAWQ and Greenplum


This interpreter seamlessly supports the following SQL data processing engines:

This Video Tutorial illustrates some of the features provided by the Postgresql Interpreter.

Create Interpreter

By default Zeppelin creates one PSQL instance. You can remove it or create new instances.

Multiple PSQL instances can be created, each configured to the same or different backend databases. But over time a Notebook can have only one PSQL interpreter instance bound. That means you cannot connect to different databases in the same Notebook. This is a known Zeppelin limitation.

To create new PSQL instance open the Interpreter section and click the +Create button. Pick a Name of your choice and from the Interpreter drop-down select psql. Then follow the configuration instructions and Save the new instance.

Note: The Name of the instance is used only to distinct the instances while binding them to the Notebook. The Name is irrelevant inside the Notebook. In the Notebook you must use %psql.sql tag.

Bind to Notebook

In the Notebook click on the settings icon in the top right corner. The select/deselect the interpreters to be bound with the Notebook.


You can modify the configuration of the PSQL from the Interpreter section. The PSQL interpreter expenses the following properties:

Property Name Description Default Value
postgresql.url JDBC URL to connect to jdbc:postgresql://localhost:5432
postgresql.user JDBC user name gpadmin
postgresql.password JDBC password JDBC driver name. In this version the driver name is fixed and should not be changed org.postgresql.Driver
postgresql.max.result Max number of SQL result to display to prevent the browser overload 1000

How to use

Tip: Use (CTRL + .) for SQL auto-completion.

DDL and SQL commands

Start the paragraphs with the full %psql.sql prefix tag! The short notation: %psql would still be able run the queries but the syntax highlighting and the auto-completions will be disabled.

You can use the standard CREATE / DROP / INSERT commands to create or modify the data model:

drop table if exists mytable;
create table mytable (i int);
insert into mytable select generate_series(1, 100);

Then in a separate paragraph run the query.

select * from mytable;

Note: You can have multiple queries in the same paragraph but only the result from the first is displayed. [1], [2].

For example, this will execute both queries but only the count result will be displayed. If you revert the order of the queries the mytable content will be shown instead.

select count(*) from mytable;
select * from mytable;

PSQL command line tools

Use the Shell Interpreter (%sh) to access the command line PSQL interactively:

psql -h phd3.localdomain -U gpadmin -p 5432 <<EOF

This will produce output like this:

        Name        |  Owner  
 hawq_toolkit       | gpadmin
 information_schema | gpadmin
 madlib             | gpadmin
 pg_catalog         | gpadmin
 pg_toast           | gpadmin
 public             | gpadmin
 retail_demo        | gpadmin

Apply Zeppelin Dynamic Forms

You can leverage Zeppelin Dynamic Form inside your queries. You can use both the text input and select form parametrization features

SELECT ${group_by}, count(*) as count
FROM retail_demo.order_lineitems_pxf
GROUP BY ${group_by=product_id,product_id|product_name|customer_id|store_id}
ORDER BY count ${order=DESC,DESC|ASC}
LIMIT ${limit=10};

Example HAWQ PXF/HDFS Tables

Create HAWQ external table that read data from tab-separated-value data in HDFS.

CREATE EXTERNAL TABLE retail_demo.payment_methods_pxf (
  payment_method_id smallint,
  payment_method_code character varying(20)
) LOCATION ('pxf://${NAME_NODE_HOST}:50070/retail_demo/payment_methods.tsv.gz?profile=HdfsTextSimple') FORMAT 'TEXT' (DELIMITER = E'\t');

And retrieve content

select * from retail_demo.payment_methods_pxf


The PSQL Interpreter provides a basic auto-completion functionality. On (Ctrl+.) it list the most relevant suggestions in a pop-up window. In addition to the SQL keyword the interpreter provides suggestions for the Schema, Table, Column names as well.