xarray to and from iris #814

nparley · 2016-04-01T10:07:44Z

Added a first go at a converter from xarray.DataArrays objects to Iris.cube.Cube objects. Uses the same template as the cdms2 conversion.

…s.cube.Cube objects. Uses the same template as the cdms2 conversion.

nparley · 2016-04-01T10:09:31Z

Reference #621 and #37. Uses template from #236.

nparley · 2016-04-01T10:21:23Z

docker pull nparley/xarray_test is a python 2.7 environment with Iris etc. which mirrors the travis CONDA_ENV=py27-cdat+iris+pynio.

It doesn't do everything, but it's a good first go. Here is an example:

>>> import xarray
>>> import iris
>>> nc_file = 'E1_north_america.nc'
>>> cube = iris.load_cube(nc_file)
>>> darray = xarray.open_dataset(nc_file)['air_temperature']
>>> print(darray)
<xarray.DataArray 'air_temperature' (time: 240, latitude: 37, longitude: 49)>
[435120 values with dtype=float32]
Coordinates:
  * time                     (time) datetime64[ns] 1860-06-01 1861-06-01 ...
  * latitude                 (latitude) float32 15.0 16.25 17.5 18.75 20.0 ...
  * longitude                (longitude) float32 225.0 226.875 228.75 ...
    forecast_period          (time) timedelta64[ns] 449 days 18:00:00 ...
    forecast_reference_time  datetime64[ns] 1859-09-01T06:00:00
    height                   float64 1.5
Attributes:
    standard_name: air_temperature
    units: K
    Model scenario: E1
    ukmo__um_stash_source: m01s03i236
    source: Data from Met Office Unified Model 6.05
    cell_methods: time: mean (interval: 6 hour)
>>> print(xarray.DataArray.from_iris(cube))
<xarray.DataArray u'air_temperature' (time: 240, latitude: 37, longitude: 49)>
[435120 values with dtype=float32]
Coordinates:
  * time                     (time) datetime64[ns] 1860-06-01 1861-06-01 ...
  * latitude                 (latitude) float32 15.0 16.25 17.5 18.75 20.0 ...
  * longitude                (longitude) float32 225.0 226.875 228.75 ...
    forecast_reference_time  datetime64[ns] 1859-09-01T06:00:00
    height                   float64 1.5
    forecast_period          (time) timedelta64[ns] 449 days 18:00:00 ...
Attributes:
    Model scenario: E1
    STASH: m01s03i236
    Conventions: CF-1.5
    source: Data from Met Office Unified Model 6.05
    standard_name: air_temperature
    units: K
>>> print(cube)
air_temperature / (K)               (time: 240; latitude: 37; longitude: 49)
     Dimension coordinates:
          time                           x              -              -
          latitude                       -              x              -
          longitude                      -              -              x
     Auxiliary coordinates:
          forecast_period                x              -              -
     Scalar coordinates:
          forecast_reference_time: 1859-09-01 06:00:00
          height: 1.5 m
     Attributes:
          Conventions: CF-1.5
          Model scenario: E1
          STASH: m01s03i236
          source: Data from Met Office Unified Model 6.05
     Cell methods:
          mean: time (6 hour
>>> print(darray.to_iris())
air_temperature / (K)               (time: 240; latitude: 37; longitude: 49)
     Dimension coordinates:
          time                           x              -              -
          latitude                       -              x              -
          longitude                      -              -              x
     Auxiliary coordinates:
          forecast_period                x              -              -
     Scalar coordinates:
          forecast_reference_time: 1859-09-01 06:00:00
          height: 1.5 m
     Attributes:
          Model scenario: E1
          source: Data from Met Office Unified Model 6.05
          ukmo__um_stash_source: m01s03i236

shoyer · 2016-04-01T15:39:10Z

xarray/test/test_dataarray.py

+                                                            'Fire Temperature'},
+                             dims=('distance', 'time'))
+
+        expected_coords = [Coordinate('distance', [-2, 2]),


Maybe better to just pull these out of original.coords instead?

…ributes

shoyer · 2016-04-02T00:35:01Z

xarray/convert.py

+        coord = encode(dataarray.coords[coord_name])
+        coord_args = get_args(coord.attrs)
+        coord_args['var_name'] = coord_name
+        iris_coord = iris.coords.DimCoord(coord.values, **coord_args)


You are probably going to need to use AuxCoord for some coordinates -- maybe anything that doesn't match a dimension name? In principle, DimCoord instances must be 1d, numeric and strictly monotonic. It does seem like this works for your test examples, though, so I'm not sure what's going on.

Yes I had that thought too last night. I started off with the code only working for dimensional coordinates and then extended to auxiliary later. It's because add_aux_coord can accept DimCoord or AuxCoord and my test is only 1D. I think it'll probably fail if the coord is 2D with the way it's currently written. I will check.

shoyer · 2016-04-02T00:37:47Z

This is awesome!

Let's try to add some documentation -- at least a mention in What's New.

CC @pelson

shoyer · 2016-04-02T00:41:45Z

xarray/convert.py

 from .conventions import (
    maybe_encode_timedelta, maybe_encode_datetime, decode_cf)

-ignored_attrs = set(['name', 'tileIndex'])
+cdms2_ignored_attrs = {'name', 'tileIndex'}


Currently we support Python 2.6, so set literals will cause test failures.

I think we can change this for the next major release of xarray though -- I just sent out a mailing list post about this.

I can change these, I wasn't doing dictionary comprehension with the new syntax for this reason but forgot about sets. It would not surprise me if quite a few scientist were still using python 2.6. Why did this not get picked up in the travis py26 tests?. Actually I can answer my own question here, I guess it's because no tests hit xarray/convert.py if Iris or cdms are not installed.

pelson · 2016-04-02T10:40:59Z

Thanks for pinging @shoyer. I'll also ping @rhattersley, and one of us will take a good look in the next few days. 👍

…rd and DimCoord correctly so 2d coords will work. Use dims variable to convert dimension numbers into names. Add 2d coord to test DataArray.

nparley · 2016-04-02T17:48:01Z

There is a problem when it comes to bounds. Xarray adds the bounds as another DataArray to the DataSet, i.e.:

Data variables:
    votemper      (dim0, dim1) float64 nan nan nan nan nan nan nan nan nan ...
    deptht_bnds   (bnds) float64 0.0 10.0

I can get the name of this variable from the Coord attribute in the DataArray for another variable, i.e. ds['votemper'].coords['deptht'].bounds but can't get the bound array from the DataArray? is that correct?

shoyer · 2016-04-02T20:41:21Z

There's no good solution for representing bounds in xarray -- we can put these as coordinates on Dataset objects, but they aren't supported on DataArray objects. I had some plans to add an IntervalIndex type to pandas (pandas-dev/pandas#8707) which would let us support bounds, but honestly I'm not sure if/when I'll finish that.

nparley · 2016-04-03T11:53:25Z

Ok, bounds can just be something to add to the converter if / when it's supported.

nparley · 2016-04-04T09:17:20Z

Latest commit fixes bug if the Iris cube does not have any dimension coordinates. What do people think about keeping or dropping the automatic dimension coordinates that xarray creates? And if dropping should be done should all coordinates named dim{x} be dropped? Or are we happy with just leaving them there?

For example at the moment:

>>> print(cube)
sea_water_potential_temperature / (degC) (-- : 148; -- : 180)
     Auxiliary coordinates:
          latitude                           x         x
          longitude                          x         x
     Scalar coordinates:
          depth: 4.99994 m, bound=(0.0, 10.0) m
          time:    1-01-01 12:00:00
     Attributes:
          Conventions: CF-1.5
     Cell methods:
          mean: time_counter
>>> print(xarray.DataArray.from_iris(cube).to_iris())
sea_water_potential_temperature / (degC) (dim0: 148; dim1: 180)
     Dimension coordinates:
          dim0                                x          -
          dim1                                -          x
     Auxiliary coordinates:
          latitude                            x          x
          longitude                           x          x
     Scalar coordinates:
          depth: 4.99993801117 m
          time:    1-01-01 12:00:00
     Attributes:
          Conventions: CF-1.5

rhattersley · 2016-04-04T11:49:23Z

What do people think about keeping or dropping the automatic dimension coordinates that xarray creates?

Is there any way to know for sure whether a coordinate has been automatically created? Or is the name the only clue?

rhattersley · 2016-04-04T13:00:04Z

Uses the same template as the cdms2 conversion.

If this is the way to do things then does this comment need deleting?

nparley · 2016-04-04T13:16:45Z

The comment was just to mention that I had based my code on the code from cdms2 from #236 as suggested by @shoyer. I guess this comment would not be needed in the final merge.

shoyer · 2016-04-04T16:02:38Z

Is there any way to know for sure whether a coordinate has been automatically created? Or is the name the only clue?

I'm afraid the name is the only clue. My inclination would be not to do anything special here to "un-name" default coordinates, assuming Iris has a straightforward way to get rid of them. Dimensions like dim0 come up quickly in demo code but rarely in practice -- it's mostly a signal to the user that the should pick a meaningful name.

If this is the way to do things then does this comment need deleting?

Yes, that comment is out of date. I'm currently refactoring that code though, so let me take care care of it.

There's a separate argument about whether or not xarray should automatically convert cubes when passed into the xarray.DataArray constructor. We do this for pandas objects, but not any other labelled array types, currently.

shoyer · 2016-04-04T16:04:26Z

One other thing that I noticed from the latest example -- it would be nice to covert cell methods from Cubes into string attributes on DataArrays.

nparley · 2016-04-08T14:47:08Z

OK the latest commit tries to do the cell methods:

>>> print(cube)
sea_water_potential_temperature / (degC) (-- : 148; -- : 180)
     Auxiliary coordinates:
          latitude                           x         x
          longitude                          x         x
     Scalar coordinates:
          depth: 4.99994 m, bound=(0.0, 10.0) m
          time:    1-01-01 12:00:00
     Attributes:
          Conventions: CF-1.5
     Cell methods:
          mean: time_counter
>>> print(xarray.DataArray.from_iris(cube))
<xarray.DataArray u'votemper' (dim0: 148, dim1: 180)>
array([[ nan,  nan,  nan, ...,  nan,  nan,  nan],
       [ nan,  nan,  nan, ...,  nan,  nan,  nan],
       [ nan,  nan,  nan, ...,  nan,  nan,  nan],
       ..., 
       [ nan,  nan,  nan, ...,  nan,  nan,  nan],
       [ nan,  nan,  nan, ...,  nan,  nan,  nan],
       [ nan,  nan,  nan, ...,  nan,  nan,  nan]], dtype=float32)
Coordinates:
    deptht        float64 5.0
    time_counter  object    1-01-01 12:00:00
    nav_lat       (dim0, dim1) float32 -78.1906 -78.1906 -78.1906 -78.1906 ...
    nav_lon       (dim0, dim1) float32 80.0 82.0 83.9999 85.9999 87.9999 ...
  * dim0          (dim0) int64 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 ...
  * dim1          (dim1) int64 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 ...
Attributes:
    units: degC
    long_name: Temperature
    standard_name: sea_water_potential_temperature
    cell_methods: time_counter: mean
    Conventions: CF-1.5
>>> print(xarray.DataArray.from_iris(cube).to_iris())
sea_water_potential_temperature / (degC) (dim0: 148; dim1: 180)
     Dimension coordinates:
          dim0                                x          -
          dim1                                -          x
     Auxiliary coordinates:
          latitude                            x          x
          longitude                           x          x
     Scalar coordinates:
          depth: 4.99993801117 m
          time:    1-01-01 12:00:00
     Attributes:
          Conventions: CF-1.5
     Cell methods:
          mean: time_counter

shoyer · 2016-04-11T02:20:04Z

xarray/convert.py

+                skip = False
+                continue
+            # If this word is an axis
+            if word not in cell_methods_strings | set(['interval', 'comment']):


can you give an example of what you're trying to catch here? this looks very complicated to me -- maybe there is an more straightforward way to do this

Yes there may, but it is actually a little complicated to parse these strings. The format is:

"coord1: [coord2:] method: [comment: a comment] [interval: a interval]"

I.e. here are some examples:

>>> get_cell_methods('dim1: dim2: mean: comment: A comment: interval: A interval') [CellMethod(method='mean', coord_names=('dim1', 'dim2'), intervals=('A interval',), comments=('A comment',))] >>> get_cell_methods('dim1: dim2: mean: dim2: sum:') [CellMethod(method='mean', coord_names=('dim1', 'dim2'), intervals=(), comments=()), CellMethod(method='sum', coord_names=('dim2',), intervals=(), comments=())] >>> get_cell_methods('dim1: mean: dim2: dim3: sum: comment: A comment') [CellMethod(method='mean', coord_names=('dim1',), intervals=(), comments=()), CellMethod(method='sum', coord_names=('dim2', 'dim3'), intervals=(), comments=('A comment',))]

So that line is looking to see if the word is a method or "comment" or "interval", if not it's a coord and so checks to see if it's a new cell method or another coord for the current one.

Is there something like iris.coords.CellMethod.from_string which handles this logic? It might be a happier path to expose such a method from Iris (which already is using this sort of logic internally) rather than incorporating it into xarray.

I haven't looked closely enough, but maybe https://github.com/SciTools/iris/blob/master/lib/iris/fileformats/netcdf.py#L1422

@pelson that's going the other direction: iris coordinate -> string (which I think is easier to get right)

@shoyer I had a look before I wrote that code but could not see one. Is there one in the netcdf library?

Looks like the code is horrifically obfuscated in Pyke rules: https://github.com/SciTools/iris/blob/master/lib/iris/fileformats/_pyke_rules/fc_rules_cf.krb#L1556-L1619

I'd be completely happy if this were moved into a more useful place. I think at the moment, it is reachable through iris.fileformats._pyke_rules.compiled_krb.fc_rules_cf_fc, but I haven't tested it, and can't guarantee its stability there.

nparley · 2016-04-20T13:05:07Z

OK this update uses iris.fileformats._pyke_rules.compiled_krb.fc_rules_cf_fc as @pelson mentioned in the comment above. I also added a cell method to the test

shoyer · 2016-04-20T16:05:00Z

xarray/convert.py

+    """
+    # Iris not a hard dependency
+    import iris
+    import iris.fileformats._pyke_rules.compiled_krb.fc_rules_cf_fc \


Can we add some sort of guard so this doesn't fail if/when Iris cleans up it's internals and moves this function? We've been bitten by using private APIs before.

Ideally, @pelson would add something exposing this to the public API for the next release, and we could use something like:

try: from iris.fileformats import parse_cell_methods except ImportError: # prior to Iris 2.0 (or whatever) from iris.fileformats._pyke_rules.compile_krb.fc_rules_cf_fc \ import _parse_cell_methods as parse_cell_methods

Yes I can add that. @pelson how do you think this might be exposed by Iris?

I'll put up a PR for iris.fileformats.netcdf.parse_cell_methods

See SciTools/iris#1981

nparley · 2016-06-26T11:35:36Z

@shoyer I have not forgotten about this PR, I just been working a lot on a pandas PR updating the ci and it's been taking up a lot of my spare time.

shoyer · 2016-06-26T16:04:11Z

No worries, no rush here from my end :).
On Sun, Jun 26, 2016 at 1:35 PM Neil Parley [email protected]
wrote:

@shoyer https://github.com/shoyer I have not forgotten about this PR, I
just been working a lot on a pandas PR updating the ci and it's been taking
up a lot of my spare time.

—
You are receiving this because you were mentioned.

Reply to this email directly, view it on GitHub
#814 (comment), or mute
the thread
https://github.com/notifications/unsubscribe/ABKS1tPi3K7IyPa2MwwQWDfnrwMemmetks5qPmQJgaJpZM4H9jGH
.

pelson · 2016-06-27T10:00:54Z

@pelson @rhattersley any tips on handling the edge cases of cubes with some, but not all missing dim_coords?

I might consider something like (untested):

for dim in xrange(cube.ndim):
    try:
        dim_coord = cube.coord(dim_coords=True, dimensions=(dim,))
    except iris.exceptions.CoordinateNotFoundError:
        index_coord = range(cube.shape[dim])

# Conflicts: # xarray/test/test_dataarray.py

nparley · 2016-08-09T15:15:53Z

This adds the extra code that @pelson suggested but it would be nice to create a little unit test with a mixed Dimension and Auxiliary coordinates. Is there an easy way to create a simple cube in Iris that could test this?

shoyer · 2016-08-09T16:32:09Z

xarray/convert.py

+            dims.append(dim_coord.var_name)
+        except iris.exceptions.CoordinateNotFoundError:
+            index_coord = range(cube.shape[dim])
+            dims.append("dim{}".format(index_coord))


Use dim_{} to be consistency with our normal rule for default dimensions

This should actually be something like dims.append('dim_{}'.format(dim)) -- index_coord is currently the default coordinate labels, not the dimension name.

shoyer · 2016-08-10T20:35:42Z

xarray/convert.py

+        coord_attrs = _iris_obj_to_attrs(coord)
+        coord_dims = [dims[i] for i in cube.coord_dims(coord)]
+        if coord_dims:
+            coords[coord.var_name] = (coord_dims, coord.points, coord_attrs)


I think it's also possible to coord.var_name be None. We should probably raise an error in that case.

Is there an xarray exception that might make sense for the error? Or just nameerror?

We usually just raise ValueError for cases like this where we cannot handle
the input.

On Wed, Aug 10, 2016 at 2:01 PM, Neil Parley [email protected]
wrote:

In xarray/convert.py
#814 (comment):

try:

dim_coord = cube.coord(dim_coords=True, dimensions=(i,))

dims.append(dim_coord.var_name)

except iris.exceptions.CoordinateNotFoundError:

dims.append("dim_{}".format(i))

dims = [dim.var_name for dim in cube.dim_coords]

if not dims:

dims = ["dim{}".format(i) for i in range(cube.data.ndim)]

coords = OrderedDict()

for coord in cube.coords():

coord_attrs = _iris_obj_to_attrs(coord)

coord_dims = [dims[i] for i in cube.coord_dims(coord)]

if coord_dims:

coords[coord.var_name] = (coord_dims, coord.points, coord_attrs)

Is there an xarray exception that might make sense for the error? Or just
nameerror?

—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
https://github.com/pydata/xarray/pull/814/files/338ef6b7242c06d5107d2224c1df4cdb2a4b951c#r74329887,
or mute the thread
https://github.com/notifications/unsubscribe-auth/ABKS1vRRJWIawQDcDJYZxXjQEcbQJIjkks5qejw5gaJpZM4H9jGH
.

pelson · 2016-11-22T20:39:23Z

This is looking pretty comprehensive to me. Could maybe do with a couple of tests perhaps, but otherwise 👍 .

shoyer · 2016-11-22T21:32:57Z

This is looking pretty comprehensive to me. Could maybe do with a couple of tests perhaps, but otherwise 👍 .

Indeed, merging this is really just blocked on tests right now. We need at least something so we can verify that this basically works.

nparley · 2016-11-22T22:21:15Z

This did used to have some tests, but they seem to have got lost somewhere along the way. I'll add the old tests back in, which would be a start.

DWesl · 2017-06-19T15:58:27Z

If you're still looking for the old tests, it looks like they disappeared in the last merge commit, f48de5.

shoyer · 2017-06-19T16:02:00Z

Yes, we would still love to get this in, though it might need a few other changes to work now. Could you open a new PR with the tests added back in? (Assuming Neil is not interested in picking this back up)

…

On Mon, Jun 19, 2017 at 8:58 AM, DWesl ***@***.***> wrote: If you're still looking for the old tests, it looks like they disappeared in the last merge commit, f48de5 <f48de5a#diff-48ab4ba033ad06981f566a2e2f561f5aL1819> . — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#814 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ABKS1gi9-YLVTLayYcMiFTqVSUy1v729ks5sFpqkgaJpZM4H9jGH> .

nparley · 2017-06-20T07:25:11Z

Sorry guys I don't have time to do anything on this (or any open source stuff) at the moment.

shoyer · 2017-06-20T16:14:36Z

No worries! The great thing about GitHub is that it's pretty easy for someone else to pick up your work :)

…

On Tue, Jun 20, 2017 at 12:25 AM, Neil Parley ***@***.***> wrote: Sorry guys I don't have time to do anything on this (or any open source stuff) at the moment. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#814 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ABKS1jC8Vrt_NKYbc4Z3-sgDpRAfaDqsks5sF3PXgaJpZM4H9jGH> .

duncanwp · 2017-11-30T15:25:44Z

I'm happy to pick this up - I'm not sure how best to pick up someone else's pull request though. Am I best off forking nparley:master then resubmitting, or creating a new one and merging in the changes from nparley:master?

nparley · 2017-11-30T15:38:57Z

I think also I could make you a contributor or transfer the branch to someone. So that might be an option too.

shoyer · 2017-11-30T23:41:55Z

I think either way works. Probably easiest to fork this branch and then merge in the latest from master.

…

On Thu, Nov 30, 2017 at 3:38 PM Neil Parley ***@***.***> wrote: I think also I could make you a contributor or transfer the branch to someone. So that might be an option too. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#814 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ABKS1oT1qWWy-o5CJe6Y45Y-DsGhjTucks5s7swSgaJpZM4H9jGH> .

pelson · 2018-01-03T10:33:21Z

Closed by #1750?

fmaussion · 2018-01-03T13:30:25Z

Indeed. Thanks to all people involved!

Added a first go at a converter from xarray.DataArrays objects to Iri…

f7842b3

…s.cube.Cube objects. Uses the same template as the cdms2 conversion.

shoyer reviewed Apr 1, 2016
View reviewed changes

Update tests to use original.coords and add extra tests for coord att…

e0498c5

…ributes

shoyer reviewed Apr 2, 2016
View reviewed changes

Remove set literals just in case. Replace has_key with in. Use AuxCoo…

c261845

…rd and DimCoord correctly so 2d coords will work. Use dims variable to convert dimension numbers into names. Add 2d coord to test DataArray.

Create dimensions if the Iris cube does not have any

edae053

Add code to convert cell_methods

cd92bca

shoyer reviewed Apr 11, 2016
View reviewed changes

nparley added 2 commits April 11, 2016 10:25

Don't append blank cell method

44930af

Update cell method code to use internal Iris functions. Also add tests.

6bed306

shoyer reviewed Apr 20, 2016
View reviewed changes

Update the API for IRIS change

cd06a2e

nparley added 3 commits August 9, 2016 15:49

Update to build dims with mix of Dimension and Auxiliary coordinates

877d06f

Merge remote-tracking branch 'upstream/master'

f48de5a

# Conflicts: # xarray/test/test_dataarray.py

Fix import after merge

e42aeb2

shoyer reviewed Aug 9, 2016
View reviewed changes

Bug fix / refactoring

338ef6b

shoyer reviewed Aug 10, 2016
View reviewed changes

Change the dencode_cf method and raise error if coord has no var_name

46f68ff

jhamman added contrib-help-wanted needs work labels Jul 13, 2017

duncanwp mentioned this pull request Nov 30, 2017

xarray to and from Iris #1750

Merged

4 tasks

fmaussion closed this Jan 3, 2018

xarray to and from iris #814

xarray to and from iris #814

Conversation

nparley commented Apr 1, 2016

nparley commented Apr 1, 2016

nparley commented Apr 1, 2016

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

shoyer commented Apr 2, 2016

Choose a reason for hiding this comment

Choose a reason for hiding this comment

pelson commented Apr 2, 2016

nparley commented Apr 2, 2016

shoyer commented Apr 2, 2016

nparley commented Apr 3, 2016

nparley commented Apr 4, 2016

rhattersley commented Apr 4, 2016

rhattersley commented Apr 4, 2016

nparley commented Apr 4, 2016

shoyer commented Apr 4, 2016

shoyer commented Apr 4, 2016

nparley commented Apr 8, 2016

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

nparley commented Apr 20, 2016

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

nparley commented Jun 26, 2016

shoyer commented Jun 26, 2016

pelson commented Jun 27, 2016

nparley commented Aug 9, 2016

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

dims = [dim.var_name for dim in cube.dim_coords]

if not dims:

dims = ["dim{}".format(i) for i in range(cube.data.ndim)]

pelson commented Nov 22, 2016

shoyer commented Nov 22, 2016

nparley commented Nov 22, 2016

DWesl commented Jun 19, 2017

shoyer commented Jun 19, 2017 via email

nparley commented Jun 20, 2017

shoyer commented Jun 20, 2017 via email

duncanwp commented Nov 30, 2017

nparley commented Nov 30, 2017

shoyer commented Nov 30, 2017 via email

pelson commented Jan 3, 2018

fmaussion commented Jan 3, 2018