Using variables inside the dbt_utils.date_spine macro

I’m trying to create a date spine that where the first and last timestamps are taken from another model.

For example, if I have a table that shows the times that someone logs onto their computer, I want to create a date spine that will take the earliest logon time and the latest logon time. I can use dbt_utils.date_spine if I enter set dates for the beginning and end of the spine. I have tried a few different methods of setting the start and end timestamps using variables, such as select min(logon_time) from activity_table but they don’t work.

Does dbt_utils.date_spine have to take set dates as arguments or is there a way to use variables?

For example, this works:
{{ dbt_utils.date_spine( datepart="second", start_date="2022-01-01", end_date="2022-01-01", ) }}

This does not:

{{ dbt_utils.date_spine( datepart="second", start_date=select min(logon_time) from activity_table, end_date=select max(logon_time) from activity_table, ) }}

Hey @dschupak hopefully you figured this out by now. I ran into the same issue and came up with a solution that works, but I don’t feel 100% great about it. Perhaps one of the experts here can chime in!

{% set get_first_sub_date %}
    select min(subscription_start_date)::string from {{ ref('stg_salesforce_sbqq_subscription') }}
{% endset %}
{%- set first_sub_date = dbt_utils.get_single_value(get_first_sub_date, default="'2017-01-01'") -%}

{% set get_last_sub_date %}
    select max(subscription_end_date)::string from {{ ref('stg_salesforce_sbqq_subscription') }}
{% endset %}
{%- set last_sub_date = dbt_utils.get_single_value(get_last_sub_date, default="'2026-12-31'") -%}

{{ dbt_utils.date_spine(
                        datepart="day",
                        start_date="'" + first_sub_date + "'",
                        end_date="'" + last_sub_date + "'"
                        )
}}

What particularly didn’t feel great was the parts where I had to cast the result to a ::string, and then later when using it in date_spine having to manually quote it.

I’m curious if anyone found a different way to solve this problem without the casting/manually quoting! I have been running into this problem a lot when combined data about campaigns in salesforce against ad campaigns where they constantly need to be left joined to a date spine.

This is close! If you have a look at how the date_spine macro is implemented, you’ll see that the first step is to run a datediff of start_date and end_date. This means that if you pass something in that is valid SQL, it will evaluate the valid SQL.

In this case, you just need to wrap it in brackets:

{{ dbt_utils.date_spine(
  datepart="second",
  start_date="(select min(logon_time) from ref('activity_table'))",
  end_date="(select max(logon_time) from ref('activity_table'))",    
) }}`

If you’re using a ref, you’ll need to quote them as I’ve done above, alternating single and double quotes.

1 Like

@ohitsmekatie in your case, you might prefer to pregenerate a date_spine table with all days between, say, 2000 and 2050 (whatever dates make sense).

That’s my normal recommendation, but it isn’t very viable when you’re doing second-by-second spining like the original post was. Salesforce ad campaigns are probably at a more tractable granularity though!

In BigQuery this did NOT work

{{ dbt_utils.date_spine(
  datepart="day",
  start_date="cast('2019-01-01' as date)",
  end_date="(select max(date) from ref('dim_date'))" 
) }}

Database Error in rpc request (from remote system.sql) Table-valued function not found: ref at [7:37]

But this DOES

{{ dbt_utils.date_spine(
  datepart="day",
  start_date="cast('2019-01-01' as date)",
  end_date="(select max(date) from " + ref('dim_date') | string + ")"
) }}

Maybe it helps someone in the future

1 Like

It is really nice macro although it is limited. I was thinking to use this parameter as :
datepart=“15 minutes”.
Also macro does not have flag working days or is_holiday like in calendar dimension, that is why calendar table can still be used as CTE expression for some cases better.