Issue 2: Kubeflow, Dracula and more

1. Blog: Accelerating ETL on KubeFlow with RAPIDS

Last week a blog post I wrote about running RAPIDS on Kubeflow was published on the NVIDIA Developer Blog.

This post was particularly exciting because it marks the culmination of months of engineering work. At the start of 2022 I set out to ensure that RAPIDS integrates seamlessly with KubeFlow and as a result ended up completely overhauling how Dask deployments work on Kubernetes.

It has taken many hours of engineering effort to even make this post possible, so a huge thank you to everyone both within RAPIDS and also in the Dask community who came together to make this happen.

2. My blog now uses the Dracula theme for syntax highlighting

I continually tweak how my blog looks, and my latest change was switching the syntax highlighter theme to Dracula.

I use Dracula day-to-day in VS Code and Iterm2 for terminal colours and syntax highlighting. I want my blog to be a representation of myself (the teal colour used throughout is the same as the wall in my office) so I thought why not use the same code theme.

# https://github.com/dracula/visual-studio-code/blob/master/src/dracula.yml
$schema: vscode://schemas/color-theme
name: Dracula
semanticClass: theme.dracula
semanticHighlighting: true
dracula:
  base:
    - &BG        '#282A36'
    - &FG        '#F8F8F2'
    - &SELECTION '#44475A'
    - &COMMENT   '#6272A4'
    - &CYAN      '#8BE9FD'
    - &GREEN     '#50FA7B'
    - &ORANGE    '#FFB86C'
    - &PINK      '#FF79C6'
    - &PURPLE    '#BD93F9'
    - &RED       '#FF5555'
    - &YELLOW    '#F1FA8C'

3. Code snippet: Expanding environment variables in arbitrary Python strings

Sometimes I want to allow a user to configure a template for a string and then populate it from various sources. We may want to generate a uuid or substitute in environment variables.

I often do this by leveraging Python’s str.format(). While f-strings are my preferred way of templating strings these days the older format style is great for creating a template string in one place and populating it in another.

We can then populate user provided strings with a whole array of useful things like a generated uuid or even the whole set of environment variables. If a kwarg passed to format isn’t used in the template it will quietly ignore it.

import os
import uuid

# Imagine the user configured this string somewhere
user_string = "Hello {USER} your uuid is {uuid}"

# Now let's populate the user string with all sorts of useful optional variables
filled_string = user_string.format(uuid=str(uuid.uuid4())[:10], **os.environ)

print(filled_string)  # for me prints "Hello jacob your uuid is a27fcecd-1"

4. Podcast: Fake Doctors, Real Friends

Most of the podcasts I listen to are either work related, informational or news. But if I need to unplug my brain a little I always reach for the Scrubs rewatch Podcast “Fake Doctors, Real Friends”.

I was a huge fan of Scrubs when it came out and it’s awesome to see the cast and creators going on to make even more brilliant TV recently like Ted Lasso. So it’s really nice to listen to the cast chat about their time on the show and relive one of my favorite watches while I’m doing the dishes.

5. Self-hosted: Tandoor

I have a server in my home office that I use to self-host many applications. I try to avoid relying on free apps and services that may disappear one day or sell my data to pay their server bills.

This week I’ve been especially enjoying Tandoor which is a recipe manager and meal planner. Often when I’m cooking or baking I’ll read a few similar recipes for what I want to make and then cobble together a hybrid recipe that is within my skillset, meets my family’s taste/dietary preferences and sometimes has my own twist on it. So I use Tandoor to import recipes from the web and then make modifications for my own reference.

I also like being able to replace the photos with my own once I’ve finished making something.

That’s all folks! I hope you had a good week and I’ll see you next Friday.