allow ordered dict export to yaml

[bridwell]

Currently, in utils.yamlio, dictionaries exported via convert_to_yaml use a hard-coded list to control the ordering of keys in the output file. This has two drawbacks (1) the list has to be updated when new parameters are added and (2) only the outer dictionaries ordering can be controlled, the ordering of inner dictionaries will not follow the order list.

This PR adds support for exporting ordered dictionaries to yaml in utils.yamlio. For example, the following:

inner_dict = OrderedDict()
inner_dict['z'] = 'had'
inner_dict['a'] = 'a'
inner_dict['f'] = 'little'

outer_dict = OrderedDict()
outer_dict[10] = 'marry'
outer_dict['inner'] = inner_dict
outer_dict['a'] = 'lamb'

yamlio.convert_to_yaml(outer_dict, my_file.yaml)

Will have the following yaml:

10: marry

inner:
    z: had
    a: a
    f: little

a: lamb

Note that if a ordered dictionary is provided (as in the current example), its ordering will override the hard-code ordering the yamlio file.

[coveralls]

Coverage increased (+0.04%) to 94.327% when pulling ba03b34 on AZMAG:odict_to_yaml into a14b287 on UDST:master.

[pksohn]

Great idea Scott - I'll review this later today

[coveralls]

Coverage decreased (-0.2%) to 94.129% when pulling aa9406f on AZMAG:odict_to_yaml into a14b287 on UDST:master.

[coveralls]

Coverage decreased (-0.3%) to 94.036% when pulling 542f926 on AZMAG:odict_to_yaml into a14b287 on UDST:master.

[coveralls]

Coverage increased (+0.04%) to 94.334% when pulling 058a06f on AZMAG:odict_to_yaml into a14b287 on UDST:master.

[pksohn]

This looks good other than a very minor typo. I suppose this opens us up to moving the DCM and regression YAML methods to OrderedDicts instead of relying on the list in this module. Are you planning to work on that too, or just using this for something external?

Fun sidenote: Python 3.6 now has order preserved in default dictionaries, which probably won't be fully relevant for us for many years, but good to be aware of.

[pksohn]

Just a minor typo here

[bridwell]

I was planning on using this for DCM updates. It's really just cosmetic, but I wanted a little more control over the look of the yaml.

One thing to note is that constructing a pandas DataFrame with an ordered dict will allow for controlling the column order, but not necessarily the row/index order. However, I think in most cases this can be handled with some re-indexing from the dictionary keys:

 pd.DataFrame.from_dict(my_odict, orient='index').reindex(my_odict.keys()).T

[coveralls]

Coverage increased (+0.04%) to 94.334% when pulling 9abd5f7 on AZMAG:odict_to_yaml into a14b287 on UDST:master.