Python model - more customization options

未关闭
  • 哪里不对？
#1,104 2 条评论 0 个 reaction 已指派 0 人 在 GitHub 查看

维护者通常 6 天内回复

还没有人认领这个 Issue。

评估

难度
5/5
预计耗时
一周以上
新手友好度
35/100
Issue 类型
功能
描述清晰度
需要澄清
活跃度
冷清
技术栈
python, sql
领域
data-engineering

调研方向

首先阅读 dbt/include/databricks/macros/adapters/python.sql，尤其是 databricks__py_write_table 和 py_get_writer_options，然后比较此处所示的覆盖宏和模型配置。在实现并验证约定的行为之前，先定义对写入选项和 ref 处理的受支持自定义方式，以及相应的配置和文档。

由索引模型根据 Issue 内容生成。

描述

enhancement Stale
Describe the feature

The current Python model macro is quite restrictive:

Would it be possible to offer more customization (e.g. including the overwriteSchema as option into py_get_writer_options() ?
Would it also be possible to allow more customization on the way refs are treated?

Describe alternatives you've considered

The alternative I went for now is overriding this macro. I don't mind, but I'm wondering if there could be a better way?

Additional context

Here is the macro I'm using (yes, i essentially just cut everything out, and changed spark.table to str):

{% macro databricks__py_write_table(compiled_code, target_relation) %}

{{ compiled_code }}

dbt = dbtObj(str)
df = model(dbt, spark)

{% endmacro %}

And here is my model:


version: 2

models:
  - name: test_model    
    config:
      test_var: "{{ var('test_var') }}"
      materialized: "table"
      submission_method: "serverless_cluster"
      partition_by:
        - test_var
    columns:
      - name: test_var        
        data_type: int
      - name: some_data        
        data_type: int

from datetime import datetime, timedelta


def model(dbt, session):
    test_var = dbt.config.get("test_var", None)
    partition_by = dbt.config.get('partition_by', [])
    
    query = f"""
    select
        '{test_var}' as test_var,
       123 as some_data
    from {dbt.ref('other_sql_model')}
    where test_var =  '{test_var}'
    """

    df = session.sql(query)
    target_table = f"{dbt.this.database}.{dbt.this.schema}.{dbt.this.identifier}"
    print(f"Target table: {target_table}")
    (
        df.write
        .mode("overwrite")
        .format("delta")
        .option("partitionOverwriteMode", "dynamic")
        .partitionBy(*partition_by)
        .saveAsTable(target_table)
    )
    return session.table(target_table)
Who will this benefit?

I think this customization (and associated documentation) would benefit teams that consider migrating to dbt, but having a lot of legacy python pipelines. Wrapping those pipelines with dbt with such simplified macro will allow "lift and shift" kind of move, and then gradual evolution of the pipelines to become more dbt-friendly.

Are you interested in contributing this feature?

I am happy to contribute, but I'd like to hear some critique and arguments against this. Overall, my primary interest is to enable usage of dbt also for the cases where logic is objectively too complex to be migrated to SQL in any reasonable time.

主要语言
Python
星标
391
派生
231
平均合并
9 天 7 小时
30 天内合并 PR
29

环境准备

从这里开始

  1. 先读完整个 Issue，再读项目的贡献指南。
  2. 在 Issue 下留言说明你要接手 —— 这能避免两个人做同样的事。
  3. Fork 仓库，在一个分支上完成修改。
  4. 提交 Pull Request，并在描述里引用这个 Issue 编号。

databricks/dbt-databricks 的其他 Issue

查看 databricks/dbt-databricks 的全部 Issue

相似的 Issue

更多 Python Issue

把新 issue 发到你的邮箱

精选适合新手参与的 GitHub issue 摘要。