Advanced Python Materials

In the world, everything is roughly distributed according to the Pareto principle. The smaller part is the rich, the majority is the poor (reading, you enter the golden billion). The same applies to materials about programming. Sometimes it is very difficult to find at least something not entry level.

After reading Dive into Python or similar and reading the documentation, the question arises, what to read next? You can refer to the list of books on python.org. There is a section of Advanced Books , but there are only 6 books in it (the seventh did not come out), and I would call only one really worthwhile.

Fortunately, Python has very detailed and high-quality documentation. But even in it, many topics are either only superficially affected, or they are very difficult to find (because the documentation is large, and if you do not know where to look, you will not find it).
')
Below are collected complex materials about Python, its structure and capabilities. Everything is in English (sin, do not know technical English). About Dive into Python, I went astray. Most of these materials require a good knowledge of Python and programming experience on it.

Articles from official documentation

Data model
Basic article about objects in Python. It is necessary to know by heart (truth-truth). Without it, further understanding how Python works is almost impossible. I would also like to focus on:

• Implementing Descriptors . In my opinion, descriptors are the best, unsurpassed, possibility of Python, but at the same time the most underestimated. Now they are mostly used only in frameworks. It will be necessary, somehow, to write how descriptors simplify development and improve the quality of the code.
• Customizing attribute access . And the first thing you need to read about is the difference __getattribute__ and __getattr__ .
• Special method lookup for new-style classes . Important: access to magic methods happens to bypass __getattribute__ if you just do not call them directly ( len(obj) and obj.__len__() ).

Exceptions
Exceptions must deal with inheritance ( PEP 352: Exceptions as New-Style Classes ). In Python 2.4, Exception is a classic class, not the object 's subclass. In Python 2.5, Exception already inherits BaseException(object) , i.e. this is a new class. Also note that in version 2.4 KeyboardInterrupt and SystemExit inherit Exception , and in 2.5 - BaseException .

I also recommend to directly take and compare (put two windows side by side) the hierarchy of exceptions in Python 2.4 and the hierarchy in Python 2.6 (compared to 2.5, only a few classes were added).

Articles about classes, attributes and methods

Unifying types and classes in Python 2.2 , Guido van Rossum
An article from the creator of the language about what new-style classes are. I think it is not necessary to write why it should be read (better, several times).

The Python 2.3 Method Resolution Order , Michele Simionato
How is the calculation of the order of access to methods (MRO) and attributes in multiple inheritance. The new classes can be obtained through the attribute __mro__ .

Shalabh Chaturvedi:
Very good and great books-articles + beautifully composed.

Python Types and Objects
How type and object correlate.

Python Attributes and Methods
How is access to the attributes in the new classes, the difference of functions and methods, descriptions of descriptors and MRO.

Be pythonic
A brief, not so important article, how to write code in the style of Python (added only to list all the articles of the author).

Metaclasses

I would like to cite one [very famous] quote about metaclasses:

            English 
 [Metaclasses] are deeper magic Metaclasses are complex magic, 
 than 99% of users should ever about which 99% of developers even 
 worry about.  If you wonder not to worry.  If you 
 Whether you need them, you ask if you need them 
 don't (the people who are actually metaclasses, you know, they don’t
 with people who need them 
 that they need them, and don't really need them, absolutely 
 they are sure of this, and they do not need 
 Tim Peters (clp post 2002-12-22) explanation why).

                                     (very bad that you do not know 
                                     English, this article links
                                     only on English-language materials)

It is easier to insert as monospace than trying to portray a table with cells without the width attribute, not to mention the styles. I honestly tried for about 20 minutes.

So, it is not . If you don't know metaclasses, you don't know Python. Now, after motivation, you can return to the articles.

Customizing class creation
Official documentation.

Python Metaclasses: Who? Why? When? , Mike Fletcher , PDF
There are many practical examples of how and why you can [but often do not need the author.] To use metaclasses.

Metaclass programming in Python , Part 2 , Part 3 , David Mertz, Michele Simionato
Very detailed articles, all sorted out by bone. The only BUT: very boring and academic written. Although it is possible that this is just my prejudice to articles on IBM DeveloperWorks.

Metaclasses in Python 3000, PEP 3115
Metaclasses in the third Python. The main differences are the ability to pass kwargs directly into the class definition of class Foo(*bases, **kwargs) and support for the __prepare__ function, which should return mapping for class attributes, such as an ordered dictionary.

In general, about metaclasses: know, know and know again. But! If you decide to use them, think, then think again and, nevertheless, try to do without them ;-D.

Articles on other topics

Python 401: Some Advanced Topics , PDF
The best and most detailed explanation of string interpolation ( 'Hello, %s.' % username ). Plus iterators, generators, descriptors, metaclasses, why methods are slower in functions (because an object is created with each call), etc.

How-To Guide for Descriptors , Raymond Hettinger
How descriptrats work, how to invoke and use them.

David Beazley
He writes very good, and most importantly, practical articles. Only he can read how to parse the Apache log with the help of generators or how to write an “operating system” on quiche.

Inside the python gil
How Global Interpreter Lock works and why using threads in Python can slow down a program.

Inside the New GIL (Python 3.2)
Description and tests of the new GIL, which will be in Python 3.2 (and maybe will be ported to 2.7).

Generator Tricks for Systems Programmers , Version 2
Practical use of generators and their advantages (speed + low memory consumption).

A Curious Course on Coroutines and Concurrency
What is Korutin and why it is necessary to clearly distinguish them from conventional generators (they have different functions: some give out data, others process them). At the end, he will write an analogue of the operating system on them.

My recommendations
See the result dir(method) and dir(func) . Read the help for new , itertools , functools . Be sure, if you still do not know, get acquainted with the functions map , reduce , zip and filter (a map can take many iterators at once).

Read about Abstract Base Classes and Collections. And not only in Python, but also in Java . ABC largely follows its trail.

Visit the site of PyCon , the annual developer conference. Materials of past conferences are also available there (although often not available). It is necessary to look in Conference / Schedule .

I almost forgot, be sure to read the Python Cookbook . This practical book is simply superb. How to quickly sort using the decorate-sort-redecorate method, which sorting algorithm Python uses (insertion sort for a small number of elements and timsort for a large one), how to work with XML, dynamically create modules and much more.

Conclusion

Wrote that he remembered or found in bookmarks. I hope many will help. Also do not forget ipython and it? and ?? .. (for those who do not know, this is a very powerful interactive shell for Python, and the questions show information about the objects, for example, import sys; sys? ). May the power come to you.