[Bug]: dbt.hash macro expands with unsupported 'TEXT' datatype

[maxfirman]

Is there an existing issue for this?

• I have searched the existing issues

Current Behavior

The builtin dbt hash macro expands to include a cast to TEXT which is not supported in Dremio. i.e

dbt.hash('foo')

resolves to

md5(cast(foo as TEXT))

Expected Behavior

The macro should cast to a supported data type i.e:

md5(cast(foo as VARCHAR))

Steps To Reproduce

Create a macro to log the output of dbt.hash

{%- macro log_hash(field) -%}
  {{ log(dbt.hash('foo'), True) }}
{% endmacro %}

execute:

dbt run-operation log_hash --args '{"field": "a"}'

output:

10:29:23  Running with dbt=1.4.5
10:29:24  md5(cast(foo as TEXT))

Environment

- OS: Ubuntu 22
- dbt-dremio: 1.4.5
- Dremio Software: 24.0.2
- Dremio Cloud: No

Relevant log output

No response

[maxfirman]

You can see where the hash macro is implemented here

I think the fix would be for the Dremio adapter to override the TYPE_LABELS dictionary in the Column class as shown here in the BigQuery adapter.

[Conq1]

Until a fix is in place you can create you own macro as a workaround. (you probably already know this):

{% macro dremio__hash(field) -%}
    md5(cast({{ field }} as varchar))
{%- endmacro %}

I really love how easy it is to make these adapter specific macros. :)

[maxfirman]

Thanks @Conq1. I'm able to work around the problem by implementing dremio__hash as you describe. I agree its very useful to be able to do this!