Multiple .close(varid, preserve=True) causes a growing 'coordinates' list

[jkettleb]

It appears that closing files for a variable, but then opening a new file to continue the timeseries can result is a growing coordinates attribute:

$ ncdump -h  /data/local/hadju/test/rlut_Amon_HadCM3_piControl_r1i1p1_200001\-200001.nc  | grep coordinates
  rlut:coordinates = "lat lon" ;
$ ncdump -h  /data/local/hadju/test/rlut_Amon_HadCM3_piControl_r1i1p1_200002\-200002.nc  | grep coordinates
  rlut:coordinates = "lat lon lat lon" ;

A possible fix is to identify in the code where coordinates is appended to and then to replace the append with the C equivalent of something like:

 def add_coord_att(new_coord, coordinates_att):
    '''
    append new_coord to the coordinates_att string

    Examples:
    ---------
    >>> add_coord_att('lat', '')
    'lat'
    >>> add_coord_att('lon', 'lat')
    'lat lon'
    >>> add_coord_att('lat', 'lat lon')
    'lat lon'
    '''
    coords = coordinates_att.split()
    if new_coord not in coords:
         coords.append(new_coord)
    return ' '.join(coords)

It also looks as though the history attribute is getting appended too many times - for instance 'Inverted lat' can appear multiple times. (This is also evident from a review of some CMIP5 data - so this bug has been around for a while).

[doutriaux1]

I believe this is fixed now that we merged Jamie's patch.