I’m in a way of releasing few of my projects to bitbucket.
Starting now I’m releasing the Sqlalchemy model generator. A quick reminder, it’s a project that you
can generate models based on your database. So you don’t need to write models by yourself.
You can see the project here: http://bitbucket.org/marcinkuzminski/sqlalchemy_model_generator/
Any participants are welcome.

I’d created another simple application in pylons, final step was to translate it from English to Polish.

I used the babel extractor recommended by pylons and that went really smooth.

My application have a language selector available anytime from menu.
I saw that even this switch worked very well for mako templates and string inside controllers,
it did not work when displaying my custom messages that i over rid in validators
and also did not work in my own custom validators that inherit from FancyValidator.
So i started to investigate this.

The first problem of not translating custom error messages in the regular formencode validators was easily being fixed by using lazy_ugettex method.

def _(s):
    #fix for translation error messages
    return lazy_ugettext(s)

This method allowed me to use lazy_ugettext to translate.
So now when in the validators your translating a string using _(‘this string is for translate’) method it’s actually wrapper for lazy_ugettext function.
Lazy translation is when you translate a string when it is accessed, not when the _() or other functions are called.

The second problem of translating custom error messages in your own validator was a bigger problem

I was looking for a solution for some time, and found a nice trick. You have to use your own
state_obj with static _() function. As example below for custom validation function.

class ValidLoginNames(formencode.validators.FancyValidator):

    messages = {'invalid_name':_('you cannot use %(username)s as login')}

    def validate_python(self, value, state):
        banned_names = ['admin', 'administrator', 'root']
        if value in banned_names:
            raise formencode.Invalid(self.message('invalid_name', state = State_obj, username = value), value, state)

#this is needed to translate the messages using _()
class State_obj(object):
    _ = staticmethod(_)

Those two trick now let’s you translate custom validators without a problem.

So again the blog was having some mail problems which I just fixed.
So anyone can register, post to the blog again.

Sorry for the problems.

I was browsing a web for some videos about stackless python implementation.
And i found a Montreal Python blip tv channel. montreal python blip tv
They have few interesting podcast which i recommend everyone to watch. The two last ones,
are dedicated to Stackless python and C extensions. Those are montreal python #8 and #9.

Be sure to check them out.

Reading a huge file (ie. 70gb) in python just using open() could get you in troubles :).
There is a clean and nice solution for reading a complex huge files or list.
Using the new with open() as statement and python generator function you could write an easy functions
which will read such files without taking whole computer resources. Each iteration on such a function will perform a read of
given size, and the with statement will make sure that file is closed when iteration is finished…

An example of such a function:

def read_large_file(filename, mode = 'rs', size = '1024'):
    '''
    A lazy generator functions that reads a file with a given chunk of data
    USAGE:
    for data_chunk in read_large_file('/tmp/huge.file','rb',10240):
        print data_chunk
    @param filename: a filename
    @param mode: read mode
    @param size: size to read at one iteration
    '''
    with open(filename, mode) as f:
        while 1:
            data = f.read(size)
            if not data:
                break
            yield data

Remember that with statement is supported out of the box from python 2.6 in python
2.5 you have to do

from __future__ import with_statement

before using it. If you have troubles using with statement i recommend reading
this link

I always loved how the print_r() function worked in php. In python print does print everything but when large collections (ie huge dicts) are printed using print you could get lost in the output. I just discovered a little smarter print function called pprint from module pprint

from pprint import pprint

This is great for debugging something in collections. Pprint prints in a nicer human readable form types of data like : list,dict,tuples.

l = [x for x in range(100)]
l.insert(0, l[:])
pprint(l)
l = dict([(x, x) for x in range(100)])

pprint(l)
#this will print this list in more convinient way
{0: 0,
 1: 1,
 2: 2,
 3: 3,
 4: 4,
 5: 5,
 6: 6,
 7: 7,
 8: 8,
 9: 9,
 ...
 98: 98,
 99: 99}

There are two interesting arguments in pprint function the:
width
Attempted maximum number of columns in the output.
width=100 will wrap everything that outputs longer than 100 chars.
and
depth
The maximum depth to print out nested structures.
ie:

pprint.pprint(stuff, depth=3)
['aaaaaaaaaa',
 ('spam', ('eggs', (...))),
 ['aaaaaaaaaaaaaaaaaaaaaaaaaaaaaa', 'bbbbbbbbbbbbbbbbbbbbbbbbbbbbbb'],
 ['cccccccccccccccccccc', 'dddddddddddddddddddd']]
#now as you can see the tuples at index 1 have a depth of more than 3 and #they are being reduced by the pprint. Imagine how this helps with debuggin a
#really complicated structures.

Few days ago there was again a major update on pydev eclipse plugin.
The version 1.5.1. Comes with new refactoring engine PEPTIC, it’s way better than any one before.
Complete list below: visit http://pydev.org for download.

I installed new eclipse on ubuntu 9.10, and noticed that manu of FINISH,
NEXT buttons are not working correctly ( you can’t click them).

So it found out it’s eclipse bug, but fortunately there is a simple
solution to this problem.

Write a startup script:

#!/bin/sh
export GDK_NATIVE_WINDOWS=1
/path_to_eclipse/eclipse

Save it as eclipse-run.sh make it +x, and the buttons will work now if you
run your eclipse via the run script.

Today is a great day ;) i switched my repos from svn to mercurial.
I was still thinking about GIT but there where three key points that made me choose mercurial

• hg serve ( out of the box http server)
• it’s written in python
• has a very good eclipse plug-in (http://www.vectrace.com/mercurialeclipse/)

Cons: it’s slower than GIT ( but who cares :D )

So after installing mercurial and moving my svn repos with history with hgsvn ( which can also do push to svn )
I started to setup easy access to the repository using nginx. I’ll show you how to setup an SSL http based repository to do PUSH/PULL/CLONE what ever.

Firstly we have to make self signed ssl certificates.
I found on of the easiest tutorials available you can find it here: http://www.akadia.com/services/ssh_test_certificate.html

OK when we have the ssl certificates for our server. Put it somewhere on the server so nginx can access it.
We have to setup a new virtual host for nginx that will only do ssl connections
and have basic auth additionally.

Here’s the example config:

server {
    listen          443;
    server_name     hg.yourserver.com;
    ssl    on;
    ssl_certificate    /home/ssl_certs/hg_cert.crt;
    ssl_certificate_key     /home/ssl_certs/hg_cert.key;
    access_log      /var/log/nginx/hg.log;
    auth_basic      "mercurial server";
    auth_basic_user_file    /etc/nginx/.htpasswd;

    location / {
        proxy_pass      http://127.0.0.1:8001;
        #here's where the hg server runs
        include         /etc/nginx/proxy.conf;
    }
}

Few thing to notice.
.htpasswd file has to be in a format <username>:<cryptPassword> if you don’t have apache
installed you can use my password generator for generating crypt password. This username
and password will be used to do pull/push from console and eclipse.
Another important thing is that when you run hg serve you must specify the -a 127.0.0.1 option which is
for the address the mercurial server runs. Hg serve default is to start at all interfaces
so you have this port open outside and your ssl/passwd protection is for nothing…

I run my using hg serve –webdir-conf=/etc/hg/hgweb.conf -d -p 8001 -a 127.0.0.1

The hgweb.conf should be with

push_ssl = false
#since nginx is doing the SSL
allow_push = * #NGINX  is doing the auth
style = gitweb

O and one more thing remember that your repo should be accessible to hg serv.
I made a mistake and run hg serve as www-data and my repo was to my home user,
and i had internal server error when trying to do push to server.

So now you can have your repo via http with SSL and nginx authentication.

During the weekend I’ve created Djalog project hosted at the http://code.google.com/p/djalog/. It is simplest possible logging app aimed to work within Django context (or for prototyping as it is also logging-ready out of the box after import in any scripts).

I’ve also put it on the PyPi (http://pypi.python.org/pypi/Djalog) so in order to install it you would execute:

easy_install Djalog

and you are ready. If you want to quickly check how it works just type:

import djalog
import logging

logging.debug("Log some debug information.")
logging.info("Log some information.")
logging.warn("Log some warning.")
logging.error("Log some error.")
logging.critical("Log some critical information.")

from djalog import configure
configure(LOG_USE_COLORS=True, LOG_SQL=True, LOG_LEVEL=5)

logging.debug("Log some debug information.")
logging.info("Log some information.")
logging.warn("Log some warning.")
logging.error("Log some error.")
logging.critical("Log some critical information.")

# Log sql query
logging.sql("SELECT * FROM users WHERE username = 'admin'")

This is beta but it should be changed to “final” in two weeks. Remember that you may update projects from CheeseShop using “-U” option:

easy_install -U Djalog

Enjoy!