The Problem:

When a python script starts, the main script is loaded and given the __name__ of ‘__main__’ and it is a module unto itself.  If you import a file, it creates a module with its own namespace and the two do not share, unless explicitly told to share.

What must be done is to develop a pattern to forces the modules to share.  But as an application developer, you must decide if the classes or instantances are to be shared and if the shared object is a global within the module (refenced by name) or within the class (referenced by self.name).

Another concern is that modules and classes should be stand-alone and if there are dependencies, these need to be explicitly defined in module. That way, everything needed for the code to run is defined in the code. In the long run, it should make debugging much easier.

The code samples that follow passes the instance, but the commented code passes the class.

The structure is:
db_example.py
|-dbi.py   (import the db module)
|-dbfunct  (uses db module)

db_example.py

#!/usr/bin/python
'''
test of scope
'''
import classes.dbi as dbi       #get db_util based on dbtype
#du=dbi.db_util               #add class to module global
du=dbi.du                       #add instance du to module global
from classes.dbfunc import dbfunc as fn

print 'globals:',dir()
print 'modules','fn:',fn,'du:',du
print ''
print 'from main'
du.m1()                  #method 1
du.m2()                  #method 2

print ''
print 'from imports'
fd=fn()
fd.ref_1()
fd.ref_2()

dbi.py

#database import
#put logic to get db type
#db='MySQL'
db='Postgres'
if db == 'MySQL':
from mysql_db import db_util
elif db == 'Postgres':
from postgres_db import postgres_db as db_util
du=db_util()            #create the instance

print 'dbi globals:',dir()
print ''

mysql_db.py

class db_util(object):
    _db='mysql'
    def __init__(self):
        print 'created',self._db,'class'
        pass

    def m1(self):
        print "this is m1 from",self._db
    def m2(self):
        print "this is m2 from",self._db

postgres_db.py

class postgres_db(object):

    _db='postgres'

    def __init__(self):
        print 'created',self._db,'class'

    def m1(self):
        print "this is m1 from",self._db
    def m2(self):
        print "this is m2 from",self._db

dbfunc.py

import dbi as dbi       #get db_util based on dbtype
#db_util=dbi.db_util       #add class to module global
du=dbi.du               #add the instance as a global
class dbfunc(object):
''' a unique function or process that justifies being a class and needs the db_util instance '''

    def print_globals(self):
        print 'from classc'
        print 'globals:',dir()
        print 'modules','ca:',ca,'CA:',ClassA
        print ''

    def ref_1(self):
        print 'from dbfunc ref_1  ==>',
        du.m1()
    def ref_2(self):
        print 'from dbfunc ref_2  ==>',
        du.m2()

The main program imports the db import module which determines the database type and imports the appropriate db_util file. Then creates an instance of db_util as du. In the module that imports the dbi.py module, the db_util instance defined in dbi is defined in the globals for the module with the command du=dbi.du.

Once the concept of each module is a namespace and there is not a ‘global’ as in C/C++, this makes sense.

#set timezone
date_default_timezone_set('America/Denver');

require_once($yii);
Yii::createWebApplication($config)->run();

for error CException:

Application runtime path "/var/www/html/yiitest/protected/runtime" 
is not valid. Please make sure it is a directory writable by the Web 
server process.

temp setting

chcon -Rv --type=httpd_sys_content_t /var/www/html

To set the context permanently:

semanage fcontext -a -t httpd_sys_content_t "/path to/html(/.*)?"

For full exlplanation

http://wiki.centos.org/HowTos/SELinux

SELinux HowTo

http://docs.fedoraproject.org/en-US/Fedora/13/html/Security-Enhanced_Linux/sect-Security-Enhanced_Linux-Working_with_SELinux-SELinux_Contexts_Labeling_Files.html

After installing Postgres and running benchmarks, I was unable to get the same transaction through-put as I had achieved with MySQL. This may be a tuning issue, but after a week of working on it the best I could get was 20,000 rows per minute in Postgres versus 350,000 per minute rows in MySQL. So back to MySql and rereading the documentation for the MySQLdb interface for Python. Quite by accident I read about SSCursor and SSDictCursor in the Using and Extending section of the documentation. These cursor types will open a server side cursor and will process the large transaction set. As a bonus, it performs almost as well as the regular cursor.

On my i-3, 3gig memory Fedora 17 box, I am achieving over 300,000 rows per minute which is acceptable for the processes being performed.

PHP Strict Standards:  Only variables should be passed
 by reference in /usr/share/phpPgAdmin/redirect.php on line 14

Finally, I thought about SELinux and in the boolean options section there is an option

apache - Allow HTTPD scripts and modules to access databases

I enabled this option and the problem is solved.

For those who use the command line

setsebool -P httpd_can_network_connect_db 1

in functions.php

at line 450

if ( STYLESHEETPATH == TEMPLATEPATH ) {
define('OPTIONS_FRAMEWORK_URL', TEMPLATEPATH . '/admin/');
define('OPTIONS_FRAMEWORK_DIRECTORY', get_template_directory_uri() . '/admin/');
} else {
define('OPTIONS_FRAMEWORK_URL', STYLESHEETPATH . '/admin/');
define('OPTIONS_FRAMEWORK_DIRECTORY', get_stylesheet_directory_uri() . '/admin/');
}

change to

if (is_dir(STYLESHEETPATH.'/admin/')) {
define('OPTIONS_FRAMEWORK_URL', STYLESHEETPATH . '/admin/');
define('OPTIONS_FRAMEWORK_DIRECTORY', get_stylesheet_directory_uri() . '/admin/');
} else {
define('OPTIONS_FRAMEWORK_URL', TEMPLATEPATH . '/admin/');
define('OPTIONS_FRAMEWORK_DIRECTORY', get_template_directory_uri() . '/admin/');
}

in options.php

at line 79

$imagepath = get_stylesheet_directory_uri() . '/admin/images/';

change to

$imagepath = OPTIONS_FRAMEWORK_DIRECTORY . 'images/';

in admin/options_framework.php

at line 194

jQuery('#of-option-documentation').load('/admin/documentation.php', function(){
jQuery.getScript("/admin/js/external.js", function(){});
});
});
jQuery('#of-option-about').load('/admin/about.php')
jQuery('#of-option-upgrade').load('/admin/upgrade.php')
});

change to

jQuery('#of-option-documentation').load('documentation.php', function(){
jQuery.getScript("js/external.js", function(){});
});
});
jQuery('#of-option-about').load('about.php')
jQuery('#of-option-upgrade').load('upgrade.php')
});

In header.php line 8 change (thanks to Arne for this update) :

<link rel="stylesheet" media="only screen and (max-width: 480px)" href="<?php echo get_template_directory_uri();?>/mobile.css" />

to:

<link rel="stylesheet" media="only screen and (max-width: 480px)" href="<?php echo get_stylesheet_directory_uri();?>/mobile.css" />

That should do it.

One challenge of the project was to cleanse the addresses to reduced the number of returns from the mailing due to invalid addresses.  As part of the cleansing process, the address was run through a parser written in PHP.  This code was found online and was a good starting spot. However, there were some problems with parsing streets, cities and states that had two words.  Several modifications later, some of the problems were cleared up, but several adjustments are still pending.

I made the changes needed for the project, but time did not permit correcting all the issues.

The updates address parse routine can be located at http:github.com/cswaim.  Let me hear how you are using this.

I have posted the class in a github repository and have included as sample driver.  You can find the repository at http://github.com/cswaim/gps.  If you use this, let me hear how you are using it.

The change was to allow any of the get_posts options to be used in the shortcode, instead of just the ones predefined in the default table.  Simple.  For the record, I had tested the change and it was working on two sights. So I pushed it to the WordPress plugin site.

I started getting messages the plugin was throwing errors, but I could not recreate the problem.  Finally, someone sent me the shortcode being used that was causing the problem.  Just the shortcode, no parameters.  My tests did not include testing with no parameters being passed in the shortcode.  But once I saw the shortcode, the problem was obvious.  By the way, this no parameter test has now been added to my test site.

Not the best way to meet new friends, but it was nice to hear that people were actually using the plugin.

Error connecting SSH tunnel: Could not open socket to *addr*
-or-
Can't connect to MySQL server on 127.0.0.1 (10061)

which may be a result of my not knowing how to read them – there was nothing to indicate what was causing the problem.

I added the following to the /etc/sysconfig/iptables

 -A INPUT -m state --state NEW -m tcp -p tcp --dport 3306 -j ACCEPT

but the connection still failed.

I finally found a clue when I tried to connect locally to a Fedora system and saw security message pop up on the Fedora box.  Of course I did not write it down, but it pointed me to the setting for  SELinux Management.  Going to System/Administration/SELinux Management on the Fedora computer hosting MySQL, I started looking at the ssh options under the Boolean selection menu.  I turned on the

 allow sshd to forward port connections

and that was all it took.

I am unsure what security holes this opens, but it allows Workbench to connect and that helps me manage the remote database.

Create Public Key Authentication

# ssh-keygen -t rsa

Press enter twice to create a key without a pass phrase. Then copy .ssh/id_rsa.pub to remote computer, placing it in the .ssh folder.

 # scp .ssh/id_rsa.pub userid@xxxxxxx:.ssh/id_ras.pub

Logon to remote computer and build authorized key ring

To add the new key to the authorized keys, append the new key to authorized_keys file.  Note the access level of the file should be 600.  SSH is very picky about this and may not accept a less restrictive access setting.

# cd .ssh
# cat id_ras.pub >> authorized_keys
# chmod 600 authorized_keys

Finally, logout of the remote computer and logoff and log back on local/client computer – do not skip this step if you generated the key in this session.

Test the change by ssh to site.  If no password is requested, it worked.

Debugging

If the ssh test prompts for the password,  then rerun ssh with the option -vvv to get the detailed debugging information.

1) Make sure you logged out of the local account and back in.  This reloads the credentials.

2) Make sure the id_rsa.pub file was appended to the authorized_keys

3) Verify the permissions on the file (600 works for me on Fedora 14) and 700 on the .ssh folder.