Drop support for IPython < 3.0.

[moorepants]

Fixes #235.

Also add PR links to some items in the release notes.

I've tested each of the import errors and warnings with no IPython, IPython < 3.0, and IPython 3.0+. Everything works for me locally.

• There are no merge conflicts.
• If there is a related issue, a reference to that issue is in the
  commit message.
• Unit tests have been added for the bug. (Please reference the issue #
  in the unit test.)
• The tests pass both locally (run nosetests) and on Travis CI.
• The code follows PEP8 guidelines. (use a linter, e.g.
  pylint, to check your code)
• The bug fix is documented in the Release
  Notes.
• The code is backwards compatible. (All public methods/classes must
  follow deprecation cycles.)
• All reviewer comments have been addressed.

[moorepants]

@pydy/pydy-developers This needs a review for the 0.3.0 release.

[oliverlee]

Don't use catch_warnings(). It should be used if you expect someone else's code to throw warnings.

Also why do you need to specify once? This module shouldn't be imported multiple times.

[moorepants]

By default Python does not show warnings and I only want these warnings to show once per python session. So if you happen to import this multiple times in single session it will only issue the import warning once.

catch_warnings is a nice way to set the filters on a per warning level. I'd also be fine with setting the warning filter to once globaly for PyDy. If you have a suggestion of a better pattern for per warning filter control, I can change it.

[oliverlee]

Just use the warning filter with once. The way you have it written was causing the Deprecation Warning to be displayed every single time the method was called when using Python 3.

[moorepants]

So what is the difference in Python 3 about this. I don't understand.

[moorepants]

The reason I used this way of setting the filter was because if you set the warnings filter globally in a module, you'll turn on all warnings for that python session, including, I believe, all the warnings that are normally hidden in any external module you import. Maybe this was a problem when we were testing against NumPy 1.6 and SciPy 0.9.

[moorepants]

You suggestion requires creating sublcassed warnings. I see why you did that in your other PR. I guess I never did that because it seems silly to create a subclass that doesn't have anything overridden. I'll create a PyDyImportWarning that subclasses ImportWarning.

[oliverlee]

In Python 3, the filter is only applied in the scope of the with warnings.catch_warnings() block.

Here's a short example:

import warnings

def func():
    with warnings.catch_warnings():
        warnings.filterwarnings('once'):
        warnings.warn('asd', ImportWarning)
        warnings.warn('asd', ImportWarning)

for i in range(2):
    func()

If you run with Python 2 and Python 3, it prints the warning a different number of times.

[moorepants]

Thanks, I'll change to what I said above.

[oliverlee]

Please change the commit title to:
Drop support for IPython < 3.0

[moorepants]

Ok, things are fixed.

[oliverlee]

Why is this here? I thought it was already merged with Master.

[moorepants]

I rebased and then changed it here to point to the PR instead of the original issue to follow the same pattern as the others.

[oliverlee]

cool thanks