[CT-502] [Bug] dbt seed returning '0 parameter markers' error

[crystalro0]

Is there an existing issue for this?

• I have searched the existing issues

Current Behavior

User is running dbt seed as a step in a job run in their deployment environment and getting the following error:
The SQL contains 0 parameter markers, but x parameters were supplied

Possibly related to this code snippet: https://github.com/dbt-labs/dbt-spark/blob/main/dbt/adapters/spark/connections.py#L284

Expected Behavior

Should be processing seed by converting the %s to ? before query execution.

Steps To Reproduce

1. Set up environment with Databricks and ODBC.
2. Create job with a dbt seed step.
3. Run job.
4. Error should occur.

Relevant log output

06:04:48    ('The SQL contains 0 parameter markers, but 32144 parameters were supplied', 'HY000')

Environment

- dbt: 1.0.4

What database are you using dbt with?

other (mention it in "Additional Context")

Additional Context

Databricks on AWS with SQL endpoint connection.
I believe this issue is related to this.

[crystalro0]

I have the internal ZD ticket number and some customer logs, if needed. Please let me know.

[jtcohen6]

@crystalro0 Thanks for opening! I'm going to transfer this issue to the dbt-spark repository, since I believe that's where the relevant code is located, and where any code changes would need to happen.

In v1.0, we added another spot where we'll return a different parameter marker depending on the connection method:

dbt-spark/dbt/include/spark/macros/materializations/seed.sql

Lines 1 to 3 in fbf9abe

	{% macro spark__get_binding_char() %}
	{{ return('?' if target.method == 'odbc' else '%s') }}
	{% endmacro %}

That means dbt's own debug-level logs should include ?, rather than %s; we don't need to trust that they'll be converted appropriately at the last minute via sqlparams.

I'll reach out to you for details about the ZD ticket and logs.

[jtcohen6]

Quick updates:

• I haven't been able to reproduce this, using the same seed CSV file (thanks @crystalro0!). When I run dbt seed with it in dbt Cloud (IDE), using an ODBC connection, it works just fine.
• Because spark__get_binding_char doesn't do quite the thing I thought it does, there's no easy way to tell just from the logs that the (in)correct binding character is being used. We should fix that more generally, hence Add method to _connection_keys #336

[github-actions]

This issue has been marked as Stale because it has been open for 180 days with no activity. If you would like the issue to remain open, please remove the stale label or comment on the issue, or it will be closed in 7 days.