Python Bytes
Python Bytes
Jul 22, 2020
#191 Live from the Manning Python Conference
53 min

Special guest: Ines Montani

Michael #1: VS Code Device Simulator

  • Want to experiment with MicroPython?
  • Teaching a course with little IoT devices?
    • Circuit Playground Express
    • BBC micro:bit
    • Adafruit CLUE with a screen
  • Get a free VS code extension that adds a high fidelity simulator
  • Easily create the starter code (
  • Interact with all the sensors (buttons, motion sensors, acceleration detection, device shake detection, etc.)
  • Deploy and debug on a real device when ready

  • Had the team over on Talk Python.

Brian #2: pytest 6.0.0rc1

  • New features
    • You can put configuration in pyproject.toml
    • Inline type annotations. Most user facing API and internal code.
    • New flags - --no-header - --no-summary - --strict-config : error on unknown config key - --code-highlight : turn on/off code highlighting in terminal
    • Recursive comparison for dataclass and attrs
  • Tons of fixes
  • Improved documentation
  • There’s a list of breaking changes and deprications. But really, nothing in the list seems like a big deal to me.
  • Plugin authors, including myself, should go test this.
    • Already found one problem. pytest-check: stop on fail works fine, but failing tests marked with xfail show up as xpass. Gonna have to look into that. And might have to recruit Anthony to help out again.
  • To try it: pip install pytest==6.0.0rc1
  • I’m currently running through the pytest book to make sure it all still works with pytest 6. So far, so good.
    • The one hiccup I’ve found so far, TinyDB had a breaking change with 4.0, so you need to pip install tinydb==3.15.2 to get the tasks project to run right. I should have pinned that in the original
    • However, all of the pytest stuff is still valid.
  • Guido just tweeted: “Yay type annotations in pytest!”

Ines #3: TextAttack

  • Python framework for adversarial attacks and data augmentation for natural language processing
  • What are adversarial attacks? You might have seen examples like these:
    • image classifier predicting a cat even if the image is complete noise
    • people at protests wearing shirts and masks with certain patterns to trick facial recognition
    • Google Translate hallucinating bible texts if you feed it nonsense or repetitive syllables
  • What does it mean to "understand" a model?
    • How does it behave in different situations, with unexpected data?
    • We can't just inspect the weights – that's not how neural networks work
    • To understand a model, we need to run it and find behaviours we don't like
  • TextAttack lets you run various different “attacks” from the current academic literature
  • It also lets you create more robust training data using data augmentation, for example, replacing words with synonyms, swapping characters, etc.

Michael #4: What is the core of the Python programming language?

  • By Brett Cannon, core developer
  • Brett and I discussed Python implementation for WebAssembly before
  • Get Python into the browser, but with the fact that both iOS and Android support running JavaScript as part of an app it would also get Python on to mobile.
  • We have lived with CPython for so long that I suspect most of us simply think that "Python == CPython".
  • PyPy tries to be so compatible that they will implement implementation details of CPython.
  • Basically most implementations of Python strive to pass CPython's test suite and to be as compatible with CPython as possible.
  • Python’s dynamic nature makes it hard to do outside of an interpreter
  • That has led Brett to contemplate the question of what exactly is Python?
  • How much would one have to implement to compile Python directly to WebAssembly and still be considered a Python implementation?
  • Does Python need a REPL?
  • Could you live without locals()?
  • How much compatibility is necessary to be useful? The answer dictates how hard it is to implement Python and how compatible it would be with preexisting software.
  • [Brett] has no answers
    • It might make sense to develop a compiler that translates Python code directly to WebAssembly and sacrifice some compatibility for performance.
    • It might make sense to develop an interpreter that targets WebAssembly's design but maintains a lot of compatibility with preexisting code.
    • It might make sense to simply support RustPython in their WebAssembly endeavours.
    • Maybe Pyodide will get us there.
  • Michael’s thoughts:
  • How about a Python standard language spec? A standard-library “standard???!?” spec. It’s possible - .NET did it.
  • What would be build if we could build it with web assembly?
  • Interesting options open up, say with NodeJS like capabilities, front-end frameworks
  • This could be MUCH bigger if we got browser makes to support alternative runtimes through WebAssembly

Brian #5: Getting started with Pathlib

  • Chris May
  • Blog post: Stop working so hard on paths. Get started with pathlib!
  • PDF “field guide”: Getting started with Pathlib
  • Really great introduction to Pathlib
  • Some of the info
    • This file as a path object: Path(__file__)
    • Parent directory: Path(__file__).parent
    • Absolute path: Path(__file__).parent.resolve()
    • Two levels up: Path(__file__).resolve(strict=True).parents[1] See pdf for explanation.
    • Current working dir: Path.cwd()
    • Path building with /
    • Working with files and folders
    • Using glob
    • Finding parts of paths and file names.
  • Any time spent learning Pathlib is worth it.
  • If I can do it in Pathlib, I do. It makes my code more readable.

Ines #6: Data Version Control (DVC)

  • We're currently working on v3.0 of spaCy and one of the big features is going to be a completely new way to train your custom models, manage end-to-end training workflows and make your experiments reproducible
  • It will also integrate with a tool called DVC (short for Data Version Control), which we've started using internally
  • DVC is an open-source tool for version control, specifically for machine learning and data
  • Machine learning = code + data. You can check your code into a Git repo, but you can't really check in your datasets and model weights. So it's very difficult to keep track of changes.
  • You can think of DVC as “Git for data” and the command line usage is actually pretty similar – for example, you run dvc init to initialize a repo and dvc add to start tracking assets
  • DVC lets you track any assets by adding meta files to your repository. So everything, including your data, is versioned, and you can always go back to the commit with the best accuracy
  • It also builds a dependency graph based on the inputs and outputs of each step, so you only have to re-run a step if things changed
    • for example, you might have a preprocessing step that converts your data and then a step that trains your model. If the data hasn't changed, you don't have to re-run the preprocessing step.
  • They recently released a new tool called CML (short for Continuous Machine Learning), which we haven't tried yet.
    • CI for Machine Learning
    • Previews look pretty cool: you can submit a PR with some changes and a GitHub action will run your experiment and auto-comment on the PR with the results, changes in accuracy and some graphs (similar to tools like Code Coverage etc.)






Fast API Job Experience

Sebastián Ramírez - @tiangolo

  • I saw a job post the other day.
  • It required 4+ years of experience in FastAPI.
  • I couldn't apply as I only have 1.5+ years of experience since I created that thing.
  • Maybe it's time to re-evaluate that "years of experience = skill level".

Defragged Zebra

Kubernetes Podcast from Google
Kubernetes Podcast from Google
Adam Glick and Craig Box
KubeCon NA 2020, with Stephen Augustus
Join us for all the news from KubeCon NA 2020, and a conversation with conference co-chair Stephen Augustus. Stephen is a Senior Open Source Engineer on the VMware Tanzu team, a chair of Kubernetes’ SIG Release, and a leader in many other parts of the project, past and present. Do you have something cool to share? Some questions? Let us know: * web: * mail: * twitter: @kubernetespod Chatter of the week * The kākāpō wins Bird of the Year * We’re off for 2 weeks. See you on December 8! News of the week * Cisco acquires Banzai Cloud * CNCF announces Cloud Native Survey 2020 results * Red Hat: New edge features, industrial AI/ML blueprint and AWS launch * CNCF End User Tech Radar for storage * New End User benefits * Envoy Mobile joins the CNCF * New sandbox projects * cert-manager * cdk8s * Kyverno * OpenKruise * Pravega * SchemaHero * Tinkerbell * k8ssandra from Datastax * Episode 98 with Sam Ramji * k0s from Mirantis * announces Gloo Mesh Enterprise and rebrands products * Episode 55, with Idit Levine * Pinniped * Shipa launches Ketch * Kinvolk launches Headlamp * The SPIFFE book “Solving The Bottom Turtle” * Episode 45, with Andrew Jessup * Anthos Developer Sandbox * GKE ingress features * Ambassador Labs takes in $18m and launches v1.9 * Tanzu SQL: Postgres on Kubernetes * Lightning round: * Accurics extends Terrascan * AWS adds containers to Lightsail * Arrikto takes $10m in funding * Brobridge releases Gravity * CircleCI runner is GA * Cloud66 for agencies and multiple database support * Cloudflare Origin CA cert-manager plugin * Cloudical Vanillastack * Cloudify version 5.1 * Codefresh launches GitOps 2.0 features * Commvault backup-as-a-service * Diamanti Spektra 3.1 and customer portal * Dynatrace PurePath 4 * Elastisys Compliant Kubernetes * The Fairwinds Kubernetes Maturity Model * Garden takes “seed” funding * Gremlin adds soundproofing * Humio Operator * Instana adds observability tools on Kubernetes * Intuit runs TurboTax on Kubernetes * Kioxia announces a new storage offering * Kubecost adds features for monitoring outside a cluster * KubeMQ adds automatic network creation * Kubermatic updates KubeOne to v1.1 * Kubernative SINA * Kublr 1.19 * Lablup announced 20.09 RC * Magalix launches KubeAdvisor 2.0 * Mayadata launches Kubera Propel and Kubera Chaos * Mirantis adds extensions to Lens * Puppet Labs adds Relay to Puppet Enterprise * Reblaze announces Curiefense to add WAF to Envoy * Replicates wants to help you Troubleshoot * Styra adds new editions to DAS * Sysdig introduces Kubernetes-native network security (ZTNSK) and partners with IBM Cloud * TrilioVault for Kubernetes v2.0 * Zerto for Kubernetes * Google Open Source Live Kubernetes Links from the interview * KubeCon NA 2020 * Episode 117, with Constance Caramanolis * CNCF Twitch * SIG Friday: ping Stephen for the current link * Slack * CNCF Slack * Kubernetes Slack * Hallway Track * Kubernetes Podcast chat * CoreOS * CoreOS Tectonic * CoreOS acquired by Red Hat * Tectonic on Azure * SIG Azure * SIG Release * SIG PM (retired) * Kubernetes Enhancement Process * Receipts process KEP * Sidecar containers - KEP closed! * Production readiness review * Episode 10, with Josh Berkus and Tim Pepper * Release managers * Black Lives Matter announcement banner * Better announcements * Kubernetes Naming working group * Inclusive Naming project * Dan Kohn memorial * Stephen Augustus on Twitter and on the web
53 min
Streaming Audio: A Confluent podcast about Apache Kafka
Streaming Audio: A Confluent podcast about Apache Kafka
Confluent, original creators of Apache Kafka®
Multi-Tenancy in Apache Kafka ft. Anna Pozvner
Multi-tenancy has been quite the topic within the Apache Kafka® community. Anna Povzner, an engineer on the Confluent team, spends most of her time working on multi-tenancy in Kafka in Confluent Cloud. Anna kicks off the conversation with Tim Berglund (Senior Director of Developer Experience, Confluent) by explaining what multi-tenancy is, why it is worthy to be desired, and advantages over single-tenant architecture. By putting more applications and use cases on the same Kafka cluster instead of having a separate Kafka cluster for each individual application and use case, multi-tenancy helps minimize the costs of physical machines and also maintenance. She then switches gears to discuss quotas in Kafka. Quotas are essentially limits—you must set quotas for every tenant (or set up defaults) in Kafka. Anna says it’s always best to start with bandwidth quotas because they’re better understood. Stick around until the end as Anna gives us a sneak peek on what’s ahead for multi-tenant Kafka, including KIP-612, the addition of the connection rate quota, which will help protect brokers. EPISODE LINKS * Sharing is Caring: Toward Creating Self-Tuning Multi-Tenant Kafka (Anna Povzner, Confluent) Join the Confluent Community Slack * Learn more with Kafka tutorials, resources, and guides at Confluent Developer * Live demo: Kafka streaming in 10 minutes on Confluent Cloud * Use *60PDCAST* to get an additional $60 of free Confluent Cloud usage (details)
44 min
Software Defined Talk
Software Defined Talk
Software Defined Talk LLC
Episode 272: This time we’re doing it in green
This week we discuss IBM buying Instana, highlights from Kubecon and the rise of Substack. Plus, Coté updates us on his quest to live the iPad lifestyle. The Rundown IBM To Acquire Instana For Undisclosed Terms ( Instana Crunch Base Overview ( KubeCon Review Envoy Mobile Joins the CNCF ( Open Source Web Engine Servo to be Hosted at Linux Foundation - The Linux Foundation ( Relevant to your Interests The Evolution of Cloud ( 2020 State of DevOps Report | presented by Puppet, & CircleCi ( Amazon's Inferentia AI Chip Is Ready For Prime Time, Now Powers the Alexa Service ( FinOps Foundation launches Kubernetes Whitepaper with CNCF, adds vendor members, Densify, SoftwareOne, Virtasant and more... ( Google Pay reimagined: pay, save, manage expenses and more ( New Zoom feature can alert room owners of possible Zoombombing disruptions ( Fast Facts Your Pa$$word doesn't matter ( 11 facts about real world container use ( Half of all containers are now managed by cloud provider and third-party registries ( Apple Apple to lower commissions for small businesses on App Store ( Safely open apps on your Mac ( Yeah, Apple’s M1 MacBook Pro is powerful, but it’s the battery life that will blow you away ( Mac users couldn’t launch apps this afternoon after Apple verification server issue ( Nonsense Welcome to Slow Boring ( Zoom Is Temporarily Removing Its 40 Minute Limit on Video Calls for Thanksgiving Day ( Sponsors strongDM — Manage and audit remote access to infrastructure. Start your free 14-day trial today at: ( Teleport provides consolidated access to all computing resources such as servers, Kubernetes clusters or internal applications across all environments. Watch a demo, download the free version, or sign up for cloud at ( Listener Feedback Little Snitch (http://Little Snitch 4) from Jordy Mac users couldn’t launch apps this afternoon after Apple verification server issue ( Conferences Mykel Alvis All Day DevOps Talk ( SDT news & hype Join us in Slack ( Send your postal address to ( and we will send you free laptop stickers! Follow us on Twitch (, Twitter (, Instagram ( and LinkedIn ( Brandon built the Quick Concall iPhone App ( and he wants you to buy it for $0.99. Use the code SDT to get $20 off Coté’s book, ( Digital WTF (, so $5 total. Become a sponsor of Software Defined Talk (! Recommendations Matt: Rewatching Veep ( Brandon: The Undoing ( on HBO ( and Serverless Data APIs ( from the Cloudcast Coté: Dithering ( podcast ( ( (; Ted Laso. ( Photo Credit from Red Hat Container Coloring Book ( Photo Credit (
1 hr 5 min
Azure DevOps Podcast
Azure DevOps Podcast
Jeffrey Palermo
Scott Nichols on the State of Azure - Part 2 - Episode 116
This episode is part 2 of the interview with Scott Nichols! Be sure to tune in to part 1 first before joining in on this episode’s conversation. Scott Nichols is a Sr. Cloud Solutions Architect, Scott works for the commercial enterprise division serving the west region. He is also the leader of the .NET and the Azure user groups in Boise, Idaho. He started his career in the IT profession in 1993 as a mainframe and web developer. Since then, he has served as a Lead Software Engineer/Solution Architect, a Software Development Engineering Manager, a Sr. Cloud Software Solution Architect, a Sr. Enterprise Solution Architect, and of course, most recently, a Sr. Cloud Solutions Architect at Microsoft since 2019. In this second part, the interview transitions from discussing the state of Azure and the Cloud industry into talking about how customers are modernizing their existing applications and infrastructure for Azure. Scott shares about infrastructure as code tools he sees as having the most traction right now, his recommendations for those looking to get their application into Azure, and what he sees as being the most successful pathways for his customers utilizing Azure infrastructure. Topics of Discussion: [:38] Be sure to visit AzureDevOps.Show for past episodes and show notes. [1:01] About The Azure DevOps Podcast and Jeffrey’s offer to speak at virtual user groups. [1:11] Clear Measure is hiring! Be sure to check out the link in the show notes. [1:33] Jeffrey welcomes Scott back to the podcast for part 2 and shares about this episode’s conversation. [2:06] The infrastructure as code tools Scott sees as having the most traction. [4:35] Does Scott have any customers that have used Azure Bicep yet? [4:56] How Scott thinks about Azure infrastructure and what he sees as being most successful with his customers. [7:42] For certain tweaks in the infrastructure, where does an ARM template work? [9:45] What is the mechanism in Azure that would know that the Powershell script has not been run yet or has already been run so that it doesn’t run it again? [11:07] A word from Azure DevOps Podcast’s sponsor: Clear Measure. [11:38] For people looking to get their application into Azure, what would Scott recommend their next steps be? [20:37] Jeffrey thanks Scott for joining the podcast and Scott offers some parting words of advice for developers. Mentioned in this Episode: Azure DevOps Clear Measure (Sponsor) .NET DevOps for Azure: A Developer's Guide to DevOps Architecture the Right Way, by Jeffrey Palermo — Available on Amazon! — Click here to download the .NET DevOps for Azure ebook! Jeffrey Palermo’s Youtube Jeffrey Palermo’s Twitter — Follow to stay informed about future events! The Azure DevOps Podcast’s Twitter: @AzureDevOpsShow Scott Nichols’ LinkedIn Scott Nichols’ Twitter @TheScottNichols Azure Architecture Center - Microsoft Amazon Web Services (AWS) Azure Architectures - Microsoft Docs Google Cloud (GCP) Microsoft Cloud Adoption Framework (CAF) Azure Bicep on GitHub .NET Conference 2020 Python Terraform Azure Resource Manager (ARM) CICD PowerShell HashiCorp Want to Learn More? Visit AzureDevOps.Show for show notes and additional episodes.
23 min
The Cloudcast
The Cloudcast
Cloudcast Media
eBPF & Cilium Cloud-native Networking
Dan Wendlandt (@danwendlandt, CEO/Co-Founder @Isovalent talks about the evolution of cloud networking, eBPF and Cilium for programmable infrastructure, and blurring the lines between networking, security and service-mesh.  *SHOW: *476 *SHOW SPONSOR LINKS:* * CloudAcademy -Build hands-on technical skills. Get measurable results.  * Get 50% of the monthly price of CloudAcademy by using code CLOUDCAST * Datadog Security Monitoring Homepage - Modern Monitoring and Analytics * Try Datadog yourself by starting a free, 14-day trial today. Listeners of this podcast will also receive a free Datadog T-shirt. * Learn more about Fauna: * Try FaunaDB for Free: *CLOUD NEWS OF THE WEEK *- *PodCTL Podcast is Back (Enterprise Kubernetes) *- *SHOW NOTES:* * Isovalent (homepage) * Cilium (homepage) * What is eBPF?  *Topic 1 *- Welcome to the show. We’ve been following your work for a while (Nicira, OpenShift networking, etc), but tell our audience a little bit about your background. *Topic 2* - A few years ago I wrote an article that said, “if you’re in networking, the #1 skill you should learn is Linux”. Why has there been so much shift from “traditional networking” to so many new capabilities being implemented in software, and specifically Linux? *Topic 3 *- Help us understand these two new concepts - eBPF and Cilium. It’s new packet filtering, it’s container networking, it’s multi-cluster networking, it can help with observability - lots going on here.  *Topic 4 *- What are some of the gaps in today’s networking/filtering/observability stacks that can improve with eBPF/Cilium?*   * *Topic 5 *- We’ve seen quite a few companies evolve from expertise in an open-source project to commercial offerings. What lessons have you learned from other companies that shape how Isovalent will both go-to-market and also engage with ecosystem partners?* * *Topic 6 *- What are some of the common use-cases or applications you see that highlight the value of the Isovalent stack?  *FEEDBACK?* * Email: show at thecloudcast dot net * Twitter: @thecloudcastnet
36 min
More episodes
Clear search
Close search
Google apps
Main menu