Using parameterized Dataflow SQL queries

This page explains how to run parameterized Dataflow SQL queries.

Dataflow SQL supports named and positional query parameters. Specify query parameters with the following syntax:

Named parameters with the @ character followed by an identifier, such as @param_name
Positional parameters with the ? character

You can use either named or positional parameters in a query but not both.

Running parameterized queries

To run a parameterized query, define the data type and value of the query parameters. The type and value of the query parameters must match the type of the database objects that the parameters represent.

You can define query parameters using the Google Cloud console or Google Cloud CLI:

Console

To define query parameters, use the Dataflow SQL UI:

Go to the Dataflow SQL UI.

Go to the Dataflow SQL UI
Enter the Dataflow SQL query into the query editor.

For example, the following query selects taxi rides that cost at least the specified price:
```
SELECT *
FROM pubsub.topic.`pubsub-public-data`.`taxirides-realtime`
WHERE
 ride_status = "dropoff"
 AND meter_reading >= @price_min
```
Click Create Cloud Dataflow job to open a panel of job options.
In the SQL query parameters section of the panel, fill in the Name, Type, and Value fields. The name of the parameter in the example is price_min and the type is FLOAT64.

If the parameter is positional, omit the Name field.
(Optional) If the query contains more parameters, click Add parameter.

The Google Cloud console does not support ARRAY or STRUCT parameters. To specify ARRAY or STRUCT parameters, use the Google Cloud CLI.

gcloud

To define query parameters, use the gcloud dataflow sql query command and --parameter flags.

You can repeat the --parameter flag for each query parameter. Set each --parameter flag to name:type:value.

The following command runs a parameterized query on a Pub/Sub stream of taxi rides:

gcloud dataflow sql query \
  --job-name=job-name \
  --region=region \
  --bigquery-dataset=destination-dataset \
  --bigquery-table=destination-table \
  --parameter=status:STRING:dropoff \
  --parameter=price_min:FLOAT64:5.5 \
'SELECT *
FROM pubsub.topic.`pubsub-public-data`.`taxirides-realtime`
WHERE
  ride_status = @status
  AND meter_reading >= @price_min'

If the parameter is positional, set --parameter flags to :type:value.

Using arrays in parameterized queries

To use an array type in a query parameter, set the type to ARRAY<data-type>. Construct the value as a comma-separated list of elements enclosed in square brackets, such as ["pickup", "enroute", "dropoff"].

Console

The Dataflow SQL UI does not support ARRAY parameters in Dataflow SQL queries.

gcloud

To define an array parameter, set the --parameter flag of the gcloud dataflow sql query command to name:ARRAY<data-type>:value

The following command runs query with a named array parameter on a Pub/Sub stream of taxi rides:

gcloud dataflow sql query \
  --job-name=job-name
  --region=region
  --bigquery-dataset=destination-dataset \
  --bigquery-table=destination-table \
  --parameter='status:ARRAY<STRING>:["pickup", "enroute", "dropoff"]' \
'SELECT *
FROM pubsub.topic.`pubsub-public-data`.`taxirides-realtime`
WHERE
  ride_status IN UNNEST(@status)'

For more information, see the Dataflow SQL reference for ARRAY types.

Using timestamps in parameterized queries

To use a timestamp in a query parameter, set the type to STRING but construct the value in the canonical timestamp format.

Console

To define timestamps parameters, use the Dataflow SQL UI:

Go to the Dataflow SQL UI.

Go to the Dataflow SQL UI
Enter the Dataflow SQL query into the query editor.

For example, the following query selects taxi rides that occurred after the specified date:
```
SELECT *
FROM pubsub.topic.`pubsub-public-data`.`taxirides-realtime`
WHERE
  event_timestamp >= TIMESTAMP (@date_min)
```
Click Create Cloud Dataflow job to open a panel of job options.
In the SQL query parameters section, click Add parameter.
In the Type field, select STRING.
In the Value field, enter a timestamp.

gcloud

To define an timestamp parameter, set the --parameter flag of the gcloud dataflow sql query command to name:STRING:value

The following command runs query with a named timestamp parameter on a Pub/Sub stream of taxi rides:

gcloud dataflow sql query \
  --job-name=job-name \
  --region=region \
  --bigquery-dataset=destination-dataset \
  --bigquery-table=destination-table \
  --parameter='date_min:STRING:2020-01-01 00:00:00.000 UTC' \
'SELECT *
FROM pubsub.topic.`pubsub-public-data`.`taxirides-realtime`
WHERE
  event_timestamp >= TIMESTAMP (@date_min)'

For more information, see the Dataflow SQL reference for the TIMESTAMP types.

Using structs in parameterized queries

To use a struct in a query parameter, set the type to STRUCT<field-name data-type, ...>. Construct the value as a comma-separated list of key-value pairs.

For example, STRUCT<date_min TIMESTAMP, status STRING> defines a struct with a field named date_min of type TIMESTAMP and a field named status of type STRING.

Console

The Dataflow SQL UI does not support STRUCT parameters in Dataflow SQL queries.

gcloud

To define an array parameter, set the --parameter flag of the gcloud dataflow sql query command to name:STRUCT<field-name data-type, ...>:value

The following command runs query with a named struct parameter on a Pub/Sub stream of taxi rides:

gcloud dataflow sql query \
  --job-name=job-name \
  --region=region \
  --bigquery-dataset=destination-dataset \
  --bigquery-table=destination-table \
  --parameter='rides:STRUCT<riders_min INT64, status STRING>:
  {"riders_min": 2, "status": "dropoff"}' \
'SELECT *
FROM pubsub.topic.`pubsub-public-data`.`taxirides-realtime`
WHERE
  passenger_count >= @rides.riders_min
  AND ride_status = @rides.status'

For more information, see the Dataflow SQL reference for the STRUCT types.