test against python 3.5, pandas 0.17

[wholmgren]

Python 3.5 was released yesterday (Sep 14, 2015) and Pandas 0.17 is scheduled to be released sometime this week. I'll add them to the travis build matrix later this week. I don't expect any problems.

[wholmgren]

Pandas 0.17 is now out, so I'm back to trying to make this work. There are no new real code compatibility problems introduced by Pandas 0.17 and Python 3.5, but there are new packaging problems.

• The Linux/Python3/SciPy issue singlediode and I_from_V do not work on Linux/Scipy 0.15/Python 3.4 #72 continues to exist on the most recent versions of those packages. On Python 3.4, you could hack around it by downgrading the numpy and scipy builds, but this is not easily possible on Python 3.5 for two reasons. 1) Continuum's Pandas 0.17 is only built against the new numpy 1.10, and 2) Continuum's SciPy 0.14 is not available for Python 3.5.
• Pandas 0.17 is not available for Python 3.3. I suggest that we remove 3.3 from the supported distributions. I don't know of any reason why one would stick with 3.3 over 3.4 or 3.5, and if you're using Python 3 then you're probably a person that likes to keep things up to date.

I just downloaded scipy 0.17dev and I'm still having problems, so this linux issue isn't going to magically disappear anytime soon.

[jforbess]

Seems reasonable to drop python 3.3, given the development level of pvlib,
and the ability to use virtual environments if absolutely necessary.

(Speaking as someone who prefers to not update until forced, but recognizes
the value of being pushed to upgrade somewhat regularly.)

On Sat, Oct 10, 2015 at 12:51 PM, Will Holmgren notifications@github.com
wrote:

Pandas 0.17 is now out, so I'm back to trying to make this work. There are
no new real code compatibility problems introduced by Pandas 0.17 and
Python 3.5, but there are new packaging problems.

• The Linux/Python3/SciPy issue singlediode and I_from_V do not work on Linux/Scipy 0.15/Python 3.4 #72
  singlediode and I_from_V do not work on Linux/Scipy 0.15/Python 3.4 #72 continues to exist
  on the most recent versions of those packages. On Python 3.4, you could
  hack around it by downgrading the numpy and scipy builds, but this is not
  easily possible on Python 3.5 for two reasons. 1) Continuum's Pandas 0.17
  is only built against the new numpy 1.10, and 2) Continuum's SciPy 0.14 is
  not available for Python 3.5.
• Pandas 0.17 is not available for Python 3.3. I suggest that we
  remove 3.3 from the supported distributions. I don't know of any reason why
  one would stick with 3.3 over 3.4 or 3.5, and if you're using Python 3 then
  you're probably a person that likes to keep things up to date.

I just downloaded scipy 0.17dev and I'm still having problems, so this
linux issue isn't going to magically disappear anytime soon.

—
Reply to this email directly or view it on GitHub
#87 (comment).