How to use jinja template with mix of params and airflow inbuilts - snowflake-cloud-data-platform

Question

I am trying to create a SQL template for Snowflake where I am trying to load a S3 file using SnowflakeOperator and s3 file is provided as xcom variable from upstream task.
Here is an example template for SQL
create or replace temp table {{ params.raw_target_table }}_tmp
as
select *
from '#{{ params.stage_name }}/{{ params.get_s3_file }}'
file_format => '{{ params.file_format }}'
;
params.get_s3_file is set to use ti like {{{{ti.xcom_pull(task_ids="foo", key="file_uploaded_to_s3")}}}}
I understand that in the template if used directly, it will work if it is not coming from params, but I want it to be configurable so I can use it with multiple dags/tasks.
Ideally I want this to work
create or replace temp table {{ params.raw_target_table }}_tmp
as
select *
from '#{{ params.stage_name }}/{{ti.xcom_pull(task_ids="{{params.previous_task}}", key="file_uploaded_to_s3")}}'
file_format => '{{ params.file_format }}'. --note the nested structure
;
So it resolves params.previous_task first and then gets the xcom values. Not sure how to instruct it do it.

Answer 1

When you use {{ <some code> }} jinja execute the code during the runtime, so this code is just hard python code (not template) executed during the runtime.
{{ti.xcom_pull(task_ids="{{params.previous_task}}", key="file_uploaded_to_s3")}} will try to pull the xcom with key file_uploaded_to_s3 from the task {{params.previous_task}} which doesn't exist. Instead of providing a string as task_ids, you can provide params.previous_task and jinja will replace it by the value of previous_task from the params dict:
create or replace temp table {{ params.raw_target_table }}_tmp
as
select *
from '#{{ params.stage_name }}/{{ti.xcom_pull(task_ids=params.previous_task, key="file_uploaded_to_s3")}}'
file_format => '{{ params.file_format }}'. --note the nested structure
;