Extract. Transform. Read.

A newsletter from Pipeline​

Hi past, present or future data professional!

It’s been a busy fall; I currently have 14 tasks in various states of development. Right now my JIRA board looks like I just won bingo—twice.

Unfortunately when you climb the tech ladder things only get busier which means you’re going to burn out unless you take steps toward proactivity.

For me this means learning which tasks I don’t need to (and really shouldn’t) do manually. And before you think I’m going to be like that developer who put his job on auto-pilot for 5 years, my prize for achieving this automation isn’t a week of Netflix binging–it’s more work.

If you’re overwhelmed by the idea of automation, I suggest you start by implementing 4 simple, small-scale automations.

Auto-fill column names in SQL queries

I work on queries with as many as 150 columns. Once, I had a task where I needed to replace a SELECT * with the explicit column name. Instead of wasting 30 minutes of dev time, I grabbed the columns from the INFORMATION_SCHEMA and iterated through them like the code snippet below.

from google.cloud import bigquery

import pandas as pd

query = “ SELECT column_name AS name FROM `project.dataset.INFORMATION_SCHEMA`.COLUMNS “

bq_client = bigquery.Client()

df = bq_client.query(query).to_dataframe()

for d in df[“name”]:

print(f”{d},”)

Never write another schema

Creating schemas is my least favorite part of data engineering. Unfortunately, they are incredibly important and can lead to nasty errors if incorrectly defined or, worse, set to auto detect.

Luckily, if you’re creating a schema based on an existing table, you can use the same INFORMATION_SCHEMA table to select the column names and types, which I explain here.

Backfill multiple CSV files

Like schema design, backfills are a pain that consume an inordinate amount of development time. Remember those 14 tasks I mentioned? At least 3 are backfills.

The worst kind of backfill is when you have to load data from a single file like a CSV. Fortunately, if you already have your files saved in a shared location like cloud storage, you can code an iterative process to download, transform and upload the final data.

Pro tip: Name your file with a date string to make it easy to identify and fill gaps programmatically.

Schedule a recurring refresh for your API credentials

As a junior engineer one of my quarterly chores was to manually refresh API credentials whenever our team calendar alert said a particular service’s creds would be expiring.

Instead, my solution and advice to you and your team is to determine the “life span” of your creds and create a function (or functions) that will perform the following steps:

• Validate the existing credentials
• Determine how many days until expiration
• If days until expiration is 0 or 1 then generate new creds (nearly every API I’ve worked with supports programmatic credential generation)
• Update your creds file or secret manager object (if using GCP Secret Manager)
• Repeat check daily

Instead of leading to laziness, automation encourages multitasking. If you implement any of the above solutions just be sure to test your output because the last thing anyone wants is a rogue autopilot.

To optimize your time, here are this week’s links as plain text.

• ​From Pool Boy To Senior Data Engineer In 2.5 Years​
• ​How To Determine Automation Costs​
• ​How To Edit BigQuery Columns Without Dropping​
• ​Backfill Your SQL Tables Without Breakage Before Anyone Finds Out You Were Wrong​

Questions? zach@pipelinetode.com

Thanks for ingesting,

-Zach Quinn