Python

iPython tricks

Use the cpaste command to copy in blocks of raw python code, even if the indenting is a little weird.

Style

Python PEP-008: Style Guide for Python Code

pylint, a Python syntax checker. Very verbose, use pylint -E (errors only) or at least pylint -r no (no report). Eg, pylint -r no file.py -d W0614 -d C -d R.

For docstring documentation, refer to PEP-257 and the Sphinx documentation; specifically, document script functionality in a top level (above imports, below any hashbang) docstring.

Use leading #: style comments to document important non-object/non-function element definitions (eg, static variables) in a way that will get pulled out into Sphinx. Use "Google-style" function argument/return documentation instead of "Sphinx style". For example:

def public_fn_with_googley_docstring(name, state=None):
    """This function does something.

    Args:
        name (str):  The name to use.

    Kwargs:
        state (bool): Current state to be in.

    Returns:
        int.  The return code::

            0 -- Success!
            1 -- No good.
            2 -- Try again.

    Raises:
        AttributeError, KeyError

    A really great idea.  A way you might use me is

    >>> print public_fn_with_googley_docstring(name='foo', state=None)
    0

    BTW, this always returns 0.  **NEVER** use with :class:`MyPublicClass`.

    """
    return 0

autopep8 is a tool to automatically pep8-ify a file. Use it like:

sudo pip install autopep8
autopep8 --in-place --select=W293,W191,W291 *.py

pep8radius is sort of similar, but only applies to code that you are going to commit (using VCS info).

Packaging

Flask app packaging advice, including MANIFEST.in and non-PyPi dependancy advice: http://flask.pocoo.org/docs/patterns/distribute/

Use console_scripts in setup.py to install system-wide scripts: http://packages.python.org/distribute/setuptools.html#automatic-script-creation

For debian packaging, use stdeb (via stackoverflow thread).

For notes on pip vs. setup.py dependencies: https://caremad.io/blog/setup-vs-requirement/

"Fucking" String Encoding

(str/unicode errors in python are very prevalent and give me the rage)

The codecs package has some helpers; see for example open(f,mode,encoding).

ASCII

'ord' is the function that takes a single ASCII character and returns the value number (as an int).

RunSnakeRun

Runsnake seems to be unmaintained... snakeviz is the new thing?

Example session:

$ python -m cProfile -o ./dump.profile myscript.py --script-option blah
$ # run to completion or Ctrl-C, then
$ runsnakerun ./dump.profile
# or
$ snakeviz ./dump.profile

nosetests

NOTE: by default I now use pytest instead of nose

To do minimal tests without wrapping everything in a class, import assert functions from nose.tools, eg:

from nose.tools import assert_raises, assert_equal

To do interactive pdb debugging, simply:

$ nosetests --pdb
# or sometimes:
$ nosetests --pdb-failures

pdb

To debug a script (interactive prompt on exception):

python -m pdb myscript.py

or in-line, always at a particular point:

import pdb; pdb.set_trace()

Use ipdb (debian: python-ipdb) instead of pdb to get a nicer IPython prompt.

Python 3 Porting

To help port and support both Python 2.7 and 3.0+, start with an import:

from __future__ import absolute_import, division, print_function

str/bytes/unicode is indeed the major porting challenge. Using bytearrays helps. Use b'asdf' style byte array definitions for most low-level constants.

struct.unpack() wants doesn't allow bytearray(); use bytes() instead.

Make sure rase Exception ("Message here") style is used everywhere, instead of raise Exception, "Message here".

There was some change in comparison between None and integers, which makes if args.verbose > 0 break. Use if args.verbose and args.verbose > 1 instead.

Debian Default Python

Keep in mind that this can break some system things (eg, arandr, some cups things).

To make Python 3.4 the default for python system-wide, do something like:

sudo update-alternatives --install /usr/bin/python python /usr/bin/python3.4 2

For an individual user you can also just create a python alias. See also:

http://linuxconfig.org/how-to-change-from-default-to-alternative-python-version-on-debian-linux

Nice Packages

subprocess/pexpect: https://github.com/kennethreitz/delegator.py

time/datetime: https://github.com/crsmithdev/arrow

tsv: https://github.com/kennethreitz/tablib

simple HTML scraping: https://github.com/michaelhelmick/lassie

sqlite3dbm is a library to back a python dict with sqlite3 on disk

pytest

pytest sort of "just works" if you put test files under ./tests/. If you want crude file imports, skipping directories, and including test_* functions from any python file (not just those with test_*.py, install pytest-pythonpath and create a pytest.ini like:

[pytest] python_paths = . python_files = .py norecursedirs = .svn _build tmp

Need to mock? <https://blog.fugue.co/2016-02-11-python-mocking-101.html>

Debugging Memory Usage

Most helpful tools I found were psutil and pympler (both need to be installed).

import os, psutil process = psutil.Process(os.getpid()) print(process.memory_info().rss) # ... do some stuff ... print(process.memory_info().rss)

and

from pympler import tracker tr = tracker.SummaryTracker() tr.print_diff()

# ... do some stuff ... tr.print_diff()

Canonical Timestamp

As a terse one-liner (with datetime imported):

f"{datetime.datetime.utcnow().isoformat()[:-3]}Z"