bk99.de entertain the web since 1997

bk99.de
HN · PythonCurated

Python 3.0 deliberately breaks with the past

Summary

The Python Software Foundation releases Python 3.0 and cleans up the language through deliberately incompatible changes. Python 3.0 was released on 3 December 2008. The release bore the code name Python 3000.

Ideas

  • print becomes a regular function.
  • Text and binary data get clearly separate types.
  • Integer division returns a floating-point result by default.
  • Iterators replace many lists that were previously built in full.
  • Exceptions and classes follow a unified syntax.
  • An automatic tool supports source code migration.

Insights

  • Long-term simplification can justify a controlled break in compatibility.
  • Migration tools reduce costs but do not replace complete tests.
  • Clear data types prevent entire classes of silent encoding bugs.
  • Parallel major versions prolong the transition for large ecosystems.

Facts

  • Python 2 and Python 3 were not fully source compatible.
  • The 2to3 tool automated many syntactic adjustments.

Recommendations

  • Migrate tests first and automate repeatable conversions.
  • Explicitly separate text, bytes and file encodings.
  • Check library support before switching production applications.

References

Read the original article

Search the Web Archive

HN · 2015

Python 3.5 adds async and await

Python 3.5 introduces native coroutine syntax, type annotations and matrix multiplication as important language and library extensions. Python 3.5 was released in September 2015. PEP 492 defined async and await.

HN · 2008

Python 2.6 prepares the transition to Python 3

Python 2.6 combines new language features with warnings and aids for the upcoming migration to Python 3. Python 2.6 was released on 1 October 2008. The version contained numerous preparations for Python 3.

HN · 2007

A spelling corrector in a few lines of Python

Peter Norvig builds a small probabilistic spelling corrector and uses it to explain the language model, error model, candidate search and honest evaluation. The corpus contains 1,115,504 words and 32,192 distinct entries. The development set reached 75 percent correct results.