C[omp]ute

From: andrew cooke <andrew@...>

Date: Tue, 10 May 2011 08:38:28 -0400

Over the weekend (and during breakfast this morning!) I've made some major
changes to the pytyp paper.  It is, I hope much more clearer about what I am
trying to do.

In particular, since version 0.15, is has a "summary" that is a non-technical
review of the work:

  ABCs [#]_ and type annotations are recent additions to the Python language.
  ABCs let programmers describe the kind of data they have in their programs.
  Type annotations associate these descriptions with particular variables.  So
  together these let programmers add type–related metadata to their code, and
  then use that information, through functions like ``isinstance()`` and
  ``issubclass()``, when the program runs.

  This paper describes work to provide more detailed descriptions: instead of
  saying "this is a list" you can say "this is a list of integers", for
  example.  It also describes ways that this information can be used: to check
  that the description matches the data, perhaps, or to indicate that
  different functions are called depending on the function arguments.

  To do all this in a clear, consistent way, I use ideas from "type systems"
  in other languages.  But that does not mean that I am adding such a system
  to Python; I am simply using the other languages to help decide how best to
  describe the data.  The end result stays true to the idea that Python is
  "dynamically typed".

  .. [#] Abstract Base Classes

As ever, the paper is at http://www.acooke.org/pytyp.pdf (and is still
incomplete - although I hope the underlying implementation is now final).

Andrew

PS Test of uni­code non­breaking space.