Importing python function with spark.read.jdbc in to Repos

[mortenhaga]

Hi all!

Before we used Databricks Repos we used the run magic to run various utility python functions from one notebook inside other notebooks, fex like reading from a jdbc connections. We now plan to switch to repos to utilize the fantastic CI/CD possibilities that gives us. But we meet some challenges with this.

First, as I have understood, you cant use the run magic to run notebooks. Only arbritary files are allowed to be used. Then we re-wrote from notebook to .py files and sucessfully imported the functions but a spark related error occured.

This is the overview of my setup:

The python file with the function resides in the folder "utils" and the notebook I want to make use of it is inside the folder "landing".

The function looks like this:

class qybele_db_connection:
  
  #this class creates a secure connection to the seas database
  jdbcHostname = "x"
  jdbcPort = 3306
  jdbcDatabase = "x"
  jdbcUrl = "jdbc:mysql://{0}:{1}/{2}".format(jdbcHostname, jdbcPort, jdbcDatabase)
  #always use the databricks secrets manager to store password
  jdbcpassword = 'x'
  connectionProps = { "user": 'x', "password": jdbcpassword }
  
  @staticmethod
  def read(spark_session,query=str):
    
 
    """
    this static method uses the above variables in a spark.read.jdbc function to create the connection.
    Note that the function takes in a string, which is a query and passes it to the "table" method in the spark function.
    """
    
    try:
      
      
      print('Executing query')
      df=spark.read.jdbc(url=qybele_db_connection.jdbcUrl, table=query, properties=qybele_db_connection.connectionProps)
      
    except Exception as e:
      raise Exception("Issue with reading from seas database")
 
    return df

This is how I import it to the notebook:

 

• Do I really have to start with the whole sparksession thing when including "spark.read.jdbc" in python files for this kind of workflow?
• How can I stick to the run magic like before; ie running the notebook, instead of importing the python function?
• How does this effect the results when you use this notebook from repos in jobs/workflows?
• How does databricks secrets act when using repos like this?

[Anonymous · Accepted solution]

If you have a notebook in a repo, that notebook can include a %run. I attached a screenshot where you can see it in my notebook. You can tell it's a repo bc of the published button on the top.

[mortenhaga]

Thats...odd. I was sure I had tried that, but now it works somehow. I guess it has to be that now I did it with double quotation marks. Thanks anyway! Works like a charm.