Support python-2.6

• Adds a tox.ini to facilitate testing supported py_versions
• Add a requirements-optional-26.txt file
• Update fauxfactory/init.py to support 2.6 syntax
• updates tests/test_*.py to conditionally import unittest2
• Add 2.6 to .travis.yml

I hope tests say all, intended to replace https://github.com/RedHatQE/cfme_tests/blob/master/utils/randomness.py#L5

btw. what's wrong on map(str, prefix_ipv4 or []) ? Shorter than the list comprehension :smile:

Using Win64 Python2.7, sys.maxsize returns a long and the isinstance(maxsize, int) test fails. I updated this to a tuple to consider the differences with Python3 (which doesn't really use long()) by adding an integer types tuple (int, long,) depending on the platform.

This generator is a helper for the gen_utf8 function which will provide the system supported list of unicode letters. This will avoid generating unicode string with control characters and other non letters characters.

Also adds tests for the generator in order to ensure it is not generating unwanted characters.

Closes #69

I was motivated to create this because this regex used to validate netmasks https://github.com/theforeman/foreman/blob/develop/lib/net/validations.rb#L8.

Also got some info at http://www.iplocation.net/tools/netmask.php.

Methods like generate_alphanumeric and generate_alpha allow the user to choose how long the resultant string should be. Each of those string generation methods checks length to ensure that the value is an integer and is not too short. There are two problems here:'

1. A length of zero is not allowed. That doesn't make sense. A user should be able to generate a zero-length string if they so desire.
2. The validation logic in each method is identical. It should be refactored out into a single private method.

this tries to install a "tests" python package, which we don't own. at the same time, also exclude docs and contrib as done in the PyPA sample project [1]

[1] https://github.com/pypa/sampleproject/blob/master/setup.py

For discovery feature, I need a valid mac that I can use for VM provisioning. However the current gen_mac() doesn't generate a valid mac for VM's on QEMU/KVM. The mac address must start with sequence: 54:52:00 otherwise VM creation fails with error:

ERROR XML error: expected unicast mac address, found multicast '63:8e:b3:53:77:b1'

So I added a new method to generate vm mac. gen_vm_mac(). We can update existing one too if that's the best option ?

Introduce fauxfactory.generate() which takes a string with formatting similar to "".format() but instead of inserting strings, it randomizes the values based on what is located inside the braces.

This pull requests proposes to simplify the definition of the FauxFactory class for backwards compatibility by looking into the module's own set of functions and individually adding them to the class using setaatr, and removing a long list of function definitions that were calling the newer set of functions.

Fauxfactory appears to be incompatible with Python 3.

Why do I say this? When executing one of the example lines of code displayed in the readme (FauxFactory.generate_alphanumeric()), I get an error stating name 'unicode' is not defined. This is probably because the application assumes that it is being run under Python 2. However, unicode support has been more fully integrated in Python 3, and the "unicode" function has been removed from the standard library.

Here's some copy-pasta from my machine:

Python 3.4.0 (default, Mar 17 2014, 23:20:09)·                                                                        
[GCC 4.8.2 20140206 (prerelease)] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> from fauxfactory import FauxFactory
>>> FauxFactory.generate_alphanumeric()
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/usr/lib/python3.4/site-packages/fauxfactory/__init__.py", line 102, in generate_alphanumeric
    return unicode(output_string)
NameError: name 'unicode' is not defined

We already have gen_cjk() and per pull #63 might have gen_cyrillic.

If we wanted to, in the future, support other methods (Tamil, Telugu, etc.), we can see where this would get very cumbersome/duplicitous, very quickly.

It might be good to have some generic function that takes any specific range and plugs it in, and then wrap that with a function specific to the unicode block you want to test.

e.g., instead of

     codepoints = [random.randint(0x4E00, 0x9FCC) for _ in range(length)]
     try:
         # (undefined-variable) pylint:disable=E0602
         output = u''.join(unichr(codepoint) for codepoint in codepoints)
     except NameError:
         output = u''.join(chr(codepoint) for codepoint in codepoints)
     return _make_unicode(output)

...put this into a generate_unicode_range() function that can have codepoint values passed to it, and then use that inside a function for any desired unicode block...

gen_bengali() gen_hebrew() gen_hiragana()

Now, there is a sticky wicket in all this. Some character sets span multiple, non contiguous blocks. More details here:

http://en.wikipedia.org/wiki/Unicode_block

So really, we should be able to pass all desired blocks into a python list, and then either make a single range to rule them all, or simply the ability to choose a random character out of each block within the list.

As described in RFC 2606, the IETF has reserved several domains for use within documentation and example code. The following top-level domains are reserved:

• test
• example
• invalid
• localhost

Additionally, the following second-level domains are reserved:

• example.com
• example.net
• example.org

Method FauxFactory.generate_email should, by default, use these IETF-sanctioned domains.