Episode #302 The Blue Shirt Episode

Watch the reside stream:

Concerning the present

Sponsored by Microsoft for Startups Founders Hub.

Brian #1: Can Amazon’s CodeWhisperer write higher Python than you?

• Brian Tarbox
• “Regardless of the clickbait-y title, whether or not CW’s code is best or worse than mine is on the margins and probably not necessary. What is important is that it has the potential to avoid wasting me a ton of time and psychological house to deal with enhancing, refactoring and testing. It’s making me a greater programmer by taking up among the undifferentiated heavy lifting.”
• Some first rate code technology, beginning with Amazon API examples.
• The generated dataclass methodology was neat, however actually, the remark “immediate” in all probability took as a lot time to write down because the code would have.
• The generated take a look at case is workable, however I’d not think about {that a} good take a look at.
  • Maybe don’t lump collectively building, attribute entry, and assessments for all strategies in a single take a look at operate.
  • That mentioned, I’ve seen approach worse take a look at strategies in my profession. So, first rate start line.
• Associated and value listening to: Changelog #506: Steady Diffusion breaks the web w/ Simon Willison
  • Principally an episode about AI generated artwork.
  • There’s a little bit of a tie in to AI code technology, the ethics round it, and ensuring you stroll up the worth chain.
• I’m planning on enjoying with GitHub CoPilot.
  • I’ve been reluctant previously, however Simon’s interview is compelling to mix skilled engineering ability with AI code technology to presumably enhance productiveness. Simon does warn towards potential abuse by Junior devs and the “simply consider the code” downside that we additionally see with “copy from StackOverflow” conditions.

Michael #2: Apache Superset

• Apache Superset is a contemporary information exploration and visualization platform
• An intuitive interface for visualizing datasets and crafting interactive dashboards
• A wide selection of lovely visualizations to showcase your information
• Code-free visualization builder to extract and current datasets
• A world-class SQL IDE for getting ready information for visualization, together with a wealthy metadata browser
• A light-weight semantic layer which empowers information analysts to rapidly outline customized dimensions and metrics
• Out-of-the-box assist for many SQL-speaking databases
• Seamless, in-memory asynchronous caching and queries
• An extensible safety mannequin that permits configuration of very intricate guidelines on who can entry which product options and datasets.
• Integration with main authentication backends (database, OpenID, LDAP, OAuth, REMOTE_USER, and so on)
• The flexibility so as to add customized visualization plugins
• An API for programmatic customization

Brian #3: Recipes from Python SQLite docs

• Redowan Delowar
• Increasing on sqlite3 Python docs with extra examples, together with
  • Executing particular person and batch statements
  • Making use of user-defined callbacks: scalar and mixture
    • scalar instance exhibits utilizing a sha256 operate to hash passwords as their inserted into the database
  • Enabling tracebacks when callbacks elevate an error
  • Reworking varieties between SQLite and Python
  • Implementing authorization management
  • … way more …
• That is nice for not solely studying SQLite, but in addition, since these sorts of subjects exist in different databases, studying about databases.
• AND an important instance of studying a subsystem by creating little code snippets to test your understanding of one thing.
  • One mod I’d do in observe is to write down these examples as pytest features, as a result of I can then run them individually whereas holding a bunch in the identical file. 🙂

Michael #4: -ffast-math and oblique modifications

• Brendan Dolan-Gavitt downloaded 4 TB of Python packages containing native x86-64 libraries and see what number of of them use -ffast-math, doubtlessly altering floating level habits in any program unfortunate sufficient to load them!
• Python packages constructed with an appealing-sounding however harmful compiler choice, -ffast-math, may find yourself inflicting any program that makes use of them to compute incorrect numerical outcomes.
• When -ffast-math is enabled, the compiler will hyperlink in a constructor that units the FTZ/DAZ flags at any time when the library is loaded — even on shared libraries, which implies that any utility that masses that library could have its floating level habits modified for the entire course of.
• A complete of 2,514 packages ultimately rely upon a bundle that makes use of -ffast-math.
• Due to extremely related nature of the trendy software program provide chain, although a mere 49 packages have been truly constructed with -ffast-math, 1000’s of different packages, with a complete of at the least 9.7 million downloads over the previous 30 days, are affected.

Extras

Brian:

• Eager about changelogs
• Specializing in serving to groups with particular objectives
  • Engaged on an experiment in consulting with some lead engineers earlier than the coaching to altering the content material of a pytest course so the examples higher match what the group will want.
    • Sharing packages by way of inner system, as that’s often completely different than the PyPI course of.
    • Altering the database and API instance of the TalkPython pytest course content material to match a groups exterior assets and duty scope.
  • It takes additional time and thought, however in the long run may improve engagement and pleasure about testing and maintaining on Python’s evolving frequent practices.

Michael:

Joke: