LookML Updater

Data is higher quality if it is defined. Moreover, definitions should be consistent wherever the data are in our systems—we should never get out of sync as this causes confusion. Thus, our goal is to have a single source of truth of definitions and have those definitions propagate into other systems. This is one part of the propagation system: propagate from master source to LookML.

This code provides the ability to examine the dimensions, dimension_groups, and measures of LookML, examine whether they contain correct descriptions, compared to some master source.

If they are missing or are incorrect or otherwise out of sync, this code will update or inject the correct description into the LookML. This can then be committed and pushed to a branch for review in a pull request. This code just does the LookML updating.

Here is a possible, example workflow:

Configuration

You will need to specify the source of the definitions:

{
    "definitions": {
        "type": "CsvDefinitionsProvider",
        "filename": "definitions.csv"
    }
}

or

{
    "definitions": {
        "type": "MySQLReader",
        "query": "select * from test",
        "username": "myusername",
        "password": "mysecret",
        "port": 3306,
        "host": "127.0.0.1",
        "database": "mydb"
    }
}

How it works

• The code parses a LookML file into a JSON object using lkml parser.

• This format allows easy interrogation of dimensions and measures to see whether they contain a description.

• We use a master source of dimension and measure definitions. Currently, can come from a local CSV (using CsvDefinitionsProvider) or relational databases (MySQLReader, PostgreReader, and SqliteReader) including a BiqQuery table (using BqDefinitionsProvider) which, for instsnce, could be a sync from a data catalog tool.

• If a dimension doesn’t contain the description, it is injected in the LookML file. If a description exists but it does not match the master definition, it is updated

• the updated LookML file is written out. The goal is not to modify other lines of the file (e.g., by changing whitespace or formatting) so that a git diff highlights just the description line modifications, thus making a pull request as easy as possible.

To run

To run,

python run_updater.py --config config/config.json --infile my_amazing_view.lkml --outfile modified.lkml

(You can set outfile to be same as infile so that it modifies the existing file.)

Example

Given an input example.lkml LookML file

view: dim_geography {
 sql_table_name: `BQDW.DimGeography` ;;

  dimension: city_code {
    type: string
    description: "this
    is
    an
    exsiting
    multiline
    description"
    sql: ${TABLE}.CityCode ;;
  }

  measure: count {
    type: count
    drill_fields: [detail*]
  }
}`

and set of required definitions:

file,type,name,definition
example.lkml,dimension,city_code,"This
is
the
correct
description"
example.lkml,measure,count,this is the count description

Running

python run.py --config example_config.json --infile example.lkml --outfile example_outfile.lkml

(where example_config.json points to that CSV above) results in example_outfile.lkml

view: dim_geography {
 sql_table_name: `BQDW.DimGeography` ;;

  dimension: city_code {
    type: string
    description: "This
is
the
correct
description"	# programmatically added by LookML modifier
    sql: ${TABLE}.CityCode ;;
  }

  measure: count {
    description: "this is the count description"	# programmatically added by LookML modifier
    type: count
    drill_fields: [detail*]
  }
}

Fully automated script

full_auto_updater.sh is a bash script that creates a full automation:

• Checks out LookML code

• Runs updater

• if and only if any changes made:

  • commits

  • pushes to remote

  • creates a pull request.

As the comments in the script show, to run this script, you will need to set a few variables:

• ACCESS_TOKEN for github (or GITHUB_USER and GITHUB_PASSWORD)

• REMOTE URL for gitrepo to update

• definitions source

Current limitations, assumptions, and TODOs

• code assumes that the opening curly brace for a dimension or measure is on the same line, ie. dimension: myname { not dimension: myname \n\n\n{