Brought to you by Michael and Brian - take a Talk Python course or get Brian's pytest book

#322: Python Packages, Let Me Count The Ways

Published Tue, Feb 7, 2023, recorded Tue, Feb 7, 2023
Watch this episode on YouTube
Play on YouTube
Watch the live stream replay

About the show

Sponsored by Microsoft for Startups Founders Hub.

Connect with the hosts

Join us on YouTube at pythonbytes.fm/stream/live to be part of the audience. Usually Tuesdays at 11am PT. Older video versions available there too.

Brian #1: Packaging Python Projects

  • Tutorial from PyPA
  • This is a really good starting point to understand how to share Python code through packaging.
  • Includes discussion of
    • directory layout
    • creating package files, LICENSE, pyproject.toml, README.md, tests and src dir
    • how to fill out build-system section of pyproject.toml
      • using either hatchling, setuptools, flit, or pdm as backends
    • metadata
    • using build to generate wheels and tarballs
    • uploading with twine
  • However
    • For small-ish pure Python projects, I still prefer flit
      • flit init creates pyproject.toml and LICENSE
        • will probably still need to hand tweak pyproject.toml
      • flit build replaces build
      • flit publish replaces twine
    • The process can be confusing, even for seasoned professionals.
  • Further discussion later in the show

Michael #2: untangle xml

  • Convert XML to Python objects
  • Children can be accessed with parent.child, attributes with element['attribute'].
  • Call the parse() method with a filename, an URL or an XML string.
  • Given this XML:
        [HTML_REMOVED]
        [HTML_REMOVED]
          [HTML_REMOVED]
        [HTML_REMOVED]
    

Access the document:

    obj.root.child['name'] # u'child1'

Calvin #3: Mypy 1.0 Released

  • Mypy is a static type checker for Python, basically a Python linter on steroids
  • Started in 2012 and developed by a team at Dropbox lead by https://github.com/JukkaL
  • What’s New?
    • New Release Numbering Scheme
      • not using symver
      • Significant backward incompatible changes will be announced in the blog post for the previous feature release
      • feature flags will allow users to upgrade and turn on the new behavior
    • Mypy 1.0 is 40% faster than 0.991 against the Dropbox internal codebase
      • 20 optimizations included in this release
    • Mypy now warns about errors used before definition or possibly undefined variables
      • for example if a variable is used outside of a block of code that may not execute
    • Mypy now supports the new Self type introduced in PEP 673 and Python 3.11
    • Support ParamSpec in Type Aliases
    • Also, ParamSpec and Generic Self types are no loner experimental
    • Lots of Miscellaneous New Features
    • Fixes to crashes
    • Support for compiling Python match statements introduced in Python 3.10

Brian #4: Thoughts on the Python packaging ecosystem

  • Pradyun Gedam
  • Some great background on the internal tension around packaging.
  • Brian’s note: in the meantime
    • people are struggling to share Python code
    • the “best practice” answer seems to shift regularly
    • this might be healthy to arrive at better tooling in the long term, but in the short term, it’s hurting us.
  • From the article:
    • The Python packaging ecosystem unintentionally became the type of competitive space that it is today.
    • The community needs to make an explicit decision if it should continue operating under the model that led to status quo.
    • Pick from N different tools that do N different things is a good model.
    • Pick from N ~equivalent choices is a really bad user experience.
    • Picking a default doesn’t make other approaches illegal.
    • Communication about the Python packaging ecosystem is fragmented, and we should improve that.
  • Pradyun: “Many of the users who write Python code are not primarily full-time software engineers or “developers”.”
  • from Thea: “The reason there are so many tools for managing Python dependencies is because Python is not a monoculture and different folks need different things.”
  • opening up the build backend through pyproject.toml-based builds was good
  • but the fracturing of multiple “workflow” tools seems bad.
  • “I am certain that it is not possible to create a single “workflow” tool for Python software. What we have today, an ecosystem of tooling where each makes different design choices and technical trade-offs, is a part of why Python is as widespread as it is today. This flexibility and availability of choice is, however, both a blessing and a curse.”
  • On building a default workflow tool around pip
    • interesting idea
  • There’s tension between “we need a default workflow tool” and “unix philosophy: many focused tools that can work together”.

Michael #5: Top PyPI Packages

  • A monthly dump of the 5,000 most-downloaded packages from PyPI.
  • Also, a full copy of PyPI info too: github.com/orf/pypi-data

Calvin #6: SQLAlchemy 2.0 Released

  • #57 on the Top PyPI Packages 😸
  • Will be giving a SQLAlchemy tutorial at Python Web Conf
  • What’s New?
    • Significant API change from 1.4
    • You’ll want to follow the migration guide and see also the what’s new in 2.0 guide
    • Fully takes advantage of Python 3 features such as dataclasses, enums and inline annotations
    • Typing support in Core and ORM, but still should be considered beta
      • all SQLAlchemy stubs packages must be uninstalled all SQLAlchemy stubs packages must be uninstalled for typing to work
      • Mypy Plugin is considered deprecated now
    • Major speed increase in the all new fully ORM-integrated bulk INSERTs
      • sorry if you are on MySQL, they don’t support INSERT RETURNING yet
      • but MariaDB does support this
    • All new bulk optimized schema reflection architecture
      • Currently enabled for PostgreSQL and Oracle
      • 250% perf increase for Postgres
      • 900% per increase for Oracle
    • Native extensions ported to Cython
      • C extensions have been replaced by Cython
      • Benchmarks as fast or sometimes faster than the previous C extensions
      • Removes some risk of memory or stability issues introduced by C
    • SQLAlchemy is now pep-517 enabled and has a pyproject.toml at the root
      • means that local source building with pip can auto install the Cython dependancy

Extras

Brian:

  • Nothing to share yet, but I’m building a new alternative Python build backend.
    • which if course will be followed with a new workflow tool that follows “my workflow”.

Michael:

Calvin:


Want to go deeper? Check our projects