Short Notes on Python

Timing, and memory, on Linux

Timing

On Linux, it's safer to use time.time()

import time
t = time.time()
# do some stuff
print "stuff took %1.3f", time.time() - t, "seconds"

On Windows, AFAIK, it's safer to use time.clock()

Memory

For me, the following does a good job getting memory usage (in kB) on Linux:

import resource
print resource.getrusage(resource.RUSAGE_SELF).ru_maxrss

Since resource is standard package, it should work on Windows too, but I don't know if it does, or what units are used if it works.

Importing Files

If you need to import a file '../mylib/commons.py', you can use the following snippet:

import sys, os
fld = os.path.realpath(os.path.abspath(os.path.join('..', 'mylib')))
if fld not in sys.path:
    sys.path.insert(0, fld)
import commons

# use your commons.py module now...
commons.super_function()

uWSGI, nginx, Flask

• install uwsgi (incl. uwsgi python plugin), python flask, and nginx,

Setting Up uWSGI

• create main.py file that will hold the server logic, for instance:

from flask import Flask

app = Flask(__name__)


@app.route("/")
def hello():
    return "hello there!"

• create uwsgi config file, wsgi.ini (minimal version here; read uwsgi docs for head-spinning array of configurables):

[uwsgi]
module          = main:app
master          = true
processes       = 5
socket          = 127.0.0.1:8000
protocol        = http
plugin          = python

• run uwsgi

uwsgi --ini wsgi.ini

• check that all works on http://localhost:8000/

Adding nginx Layer

• remove the "protocol" directive from wsgi.ini, and add "die-on-term":

[uwsgi]
module          = main:app
master          = true
processes       = 5
socket          = 127.0.0.1:8000
plugin          = python
die-on-term     = true

• add a new vhost to nginx - /etc/nginx/sites-available/app.nginx:

server {
    listen 80;
    server_name my.awesome.domain;
    location / {
        include uwsgi_params;
        uwsgi_pass 127.0.0.1:8000;
    }
}

• • communication through socket is also possible (see socket, chmod-socket, vacuum and other directives for uWSGI)
  • of course, create link in /etc/nginx/sites-enabled/, and restart nginx,

Run uWSGI daemon on boot - supervisor

• install supervisor

apt-get install supervisor

• add/edit /etc/supervisord.conf with content like this:

[unix_http_server]
file=/run/supervisor.sock

[supervisord]
logfile=/var/log/supervisord.log
logfile_maxbytes=1MB
logfile_backups=1
loglevel=info ; (others: warn,debug,trace)
pidfile=/run/supervisord.pid
nodaemon=false
minfds=1024
minprocs=200

[supervisorctl]
serverurl=unix:///run/supervisor.sock

[program:your app]
directory=/path/to/app
command=/usr/bin/uwsgi --ini wsgi.ini
autostart=true
autorestart=true
stdout_logfile=/var/log/uwsgi.log
redirect_stderr=true
stopsignal=QUIT

Run uWSGI daemon on boot - systemd

• create systemd file for uWSGI, /etc/systemd/system/uwsgi-app.service:

[Unit]
Description=Job that runs the uWSGI app

[Service]
Type=simple
WorkingDirectory=/home/project/flask-test/
ExecStart=/usr/bin/uwsgi --ini wsgi.ini

[Install]
WantedBy=multi-user.target

Then you can start and stop the uwsgi service using:

# systemctl start uwsgi-app.service
# systemctl stop uwsgi-app.service

Once you're happy with the settings, enable the daemon to be run on boot:

# systemctl enable uwsgi-app.service

Decorators

based on.

Decorators are simple and expressive way to modify function without editing the source of the function itself. Or, the other way around, to modify multiple functions in the same way, without code duplication (DRY).

Decorators can be spotted in the code by starting with @ character. Decorator is a function (or class) that can do some additional work before or after the call to the decorated function. It can even call the decorated function multiple times, or not at all.

Decorators can be implemented as closures (my fave), or as classes; the following approaches are equivalent:

# using a function (closure) as a decorator
def beforeAndAfter(f):
    def decorated_fn():
        print("Before", f.__name__)
        f()
        print("After", f.__name__)
    return decorated_fn

@entryExit
def func():
    print "func() is in da' house!"

func()

# using a class as a decorator
class beforeAndAfter(object):

    def __init__(self, f):
        self.f = f

    def __call__(self):
        print("Before", self.f.__name__)
        self.f()
        print("After", self.f.__name__)

@beforeAndAfter
def func():
    print "func() is in da' house!"

func()

Note that in functional decorator, anything outside the body of decorated_fn() is equivalent to content of the __init__() constructor of the class-based decorator. This code is run during the initialization, only once for each decorated function, regardless of whether the decorated function is ever called in the code - you should avoid any heavy lifting there.

Decorators can also accept arguments, and (obviously should) forward arguments to the decorated function:

def beforeAndAfter(p1, p2):
    def wrap(f):
        def wrapped_f(*args, **kwargs):
            print "Decorator arguments:", p1, p2
            print("Before", f.__name__)
            f(*args, **kwargs)
            print("After", f.__name__)
        return wrapped_f
    return wrap

@beforeAndAfter("hello", "world")
def func(a, b=2):
    print "func() is in da' house,", a, b

func(1)