Added sum combiner (#500)

[hletrd]

Please have a look at the following list and replace the "[ ]" with a "[x]" if
the answer to this question is yes.

• For new contributors: Did you add yourself to the "Authors.rst" file?

For documentation changes:

• For documentation changes: Does your commit message include a "[skip ci]"?
  Note that it should not if you changed any examples!

For bugfixes:

• Did you add an entry to the "Changes.rst" file?
• Did you add a regression test?
• Does the commit message include a "Fixes #issue_number" (replace "issue_number").
• Does this PR add, rename, move or remove any existing functions or parameters?

For new functionality:

• Did you add an entry to the "Changes.rst" file?
• Did you include a meaningful docstring with Parameters, Returns and Examples?
• Does the commit message include a "Fixes #issue_number" (replace "issue_number").
• Did you include tests for the new functionality?
• Does this PR add, rename, move or remove any existing functions or parameters?

Please note that the last point is not a requirement. It is meant as a check if
the pull request potentially breaks backwards-compatibility.

---

[hletrd]

Appveyor build fails on Python 3.6 because it fails to build Scipy. (Appveyor build and test on Python 2.7 succeed)

[MSeifert04]

In general this looks really good, I only had a few minor comments.

However I think it should also include a test that tests for the correct uncertainty especially with some combinations of masked and unmasked pixels. Like:

• none masked except one
• none masked (already present but uncertainty isn't checked)
• all masked (already present but uncertainty isn't checked)
• all masked except one

[MSeifert04]

Could you also include the PR number [#500, #508]?

[MSeifert04]

Maybe include a sentence that if a re-scaled average sum is needed one should do average_combine and multiply it with the number of images combined.

[MSeifert04]

I think there should be another line split. Even though 80 is just a "soft limit" I see no reason to make it much longer here.

[hletrd]

• Uncertainty test for partially masked image added in 2a7e2e6
• Modified comments in 759e899

[hletrd]

AppVeyor build only fails on Python 3.6 because it fails to build astroscrappy.

[MSeifert04]

There seem to be 3 trailing whitespaces:

ccdproc/combiner.py:462:78: W291 trailing whitespace
        Because sum_combine returns 'pure sum' with masked pixels ignored, if
                                                                             ^
    Trailing whitespace is superfluous.


ccdproc/combiner.py:463:70: W291 trailing whitespace
        re-scaled sum is needed, average_combine have to be used with
                                                                     ^
    Trailing whitespace is superfluous.



ccdproc/combiner.py:507:77: W291 trailing whitespace
        # Commented out because it cancels the effect of 'Divide uncertainty
                                                                            ^
    Trailing whitespace is superfluous.

If these are removed that should fix the Travis build.

[MSeifert04]

I'm still a bit confused about uncertainty for sum_combine:

>>> from ccdproc import CCDData, Combiner
>>> import numpy as np
>>> from astropy.nddata import StdDevUncertainty

>>> ccd1 = CCDData(np.arange(10), unit='', uncertainty=StdDevUncertainty(np.ones(10)))
>>> ccd2 = CCDData(np.ones(10), unit='', uncertainty=StdDevUncertainty(np.ones(10)))

>>> Combiner([ccd1, ccd2]).average_combine().uncertainty
StdDevUncertainty([ 0.35355339,  0.        ,  0.35355339,  0.70710678,
                    1.06066017,  1.41421356,  1.76776695,  2.12132034,
                    2.47487373,  2.82842712])

>>> Combiner([ccd1, ccd2]).sum_combine().uncertainty
StdDevUncertainty([ 0.5,  0. ,  0.5,  1. ,  1.5,  2. ,  2.5,  3. ,  3.5,
                    4. ])

It seems somewhat strange that the sum uncertainty isn't a multiplicative of the average combine. In case there is no masked pixel shouldn't it be the uncertainty from average-combine times the number of "combined" pixels?

[hletrd]

I was wrong. Changed in 25c4e68.

[crawfordsm]

Thanks @hletrd !