py-type.slide

Better type at Python
7 Nov 2015

Rick Mak
Oursky
rickmak@oursky.com
@rickmak
CC81327D

* Python 3.5 Release


https://docs.python.org/3.5/whatsnew/3.5.html
Released on September 13, 2015, 2 months ago

*New*library*modules*introduced*
 `typing`: PEP 484 – Type Hints

The syntax introduced long ago, Dec-2006
 PEP 3107 - Function Annotations

* Your view on type

- Anyone used typing?
- Anyone used function annotations?
- Anyone thinks using it will have performance boost or punishment?

* Get it working

- Python 3.5
    import typing
- Python 3.2-3.4
    pip install typing

* Some code

.code py-type/nohint.py

* Same code with type hints

.code py-type/hint.py

* More complex type hints

Dictionary

.code py-type/complex.py /1 OMIT/,/2 OMIT/

* More complex type hints

Sequence and List

.code py-type/complex.py /2 OMIT/,/3 OMIT/

* Forward references

.code py-type/complex.py /3 OMIT/,/4 OMIT/

* What are type hints good for?

- Make IDE works better
- Make developement of static analysis tools easier
- Improve readability, for tools and for human
- Serve as Documentation

* Make IDE works better

* Let see how PyCharm works

.image py-type/pycharm.png

* Make development of static analysis tools easier

* mypy

[[https://github.com/JukkaL/mypy][JukkaL/mypy]]

    pip install mypy-lang

Analyse a file without running it

    (venv)$ mypy callee.py
    callee.py, line 7: Argument 1 to "hello" has incompatible type "int"; expected "str"

* Serve as (additional) documentation

* Docstring for sphinx to gen autodoc

.code py-type/docstring.py

* Same information using func type annotation

.code py-type/doctype.py


* Python2 with pyi stub

An *.pyi file that only have interface

.code py-type/py2.pyi

Also useful for
 - Package you can't change
 - Legacy package you don't want to change

There is github repos hosting stub for stdlib
[[https://github.com/python/typeshed][python/typeshed]]

* What type hints are not

- Not about code generation
- Your code can still break after type checking
- No runtime type checking, No performance overhead

* How you can find out function annotation in your code

You can use [[https://docs.python.org/3.5/library/inspect.html][inspect]] to get back the annotation.

.code py-type/complex.py /4 OMIT/,/5 OMIT/

i.e. You may abuse it to do type conversion or validation at runtime.

* Reference

- [[http://www.mypy-lang.org][mypy]]
- [[https://www.youtube.com/watch?v=Yqnrfa5ri7E][Guido van Rossum - Type Hints for Python 3.5]]
- http://blog.pirx.ru/media/files/2015/type-hinting-talk/type-hinting.html