Export projections sometimes return Unicode, e.g. EPSG:42104

[Mike Taves]

With GDAL for Python 2, several "ExportTo..." methods sometimes return a Unicode result rather than a string. This may cause issues with SetProjection, which expects a string. Specifically, ​EPSG:42104: NAD83 / MTM zone 8 Québec.

Consider the following Python code:

from osgeo import osr
from osgeo import gdal
osr.UseExceptions()
gdal.UseExceptions()

srs = osr.SpatialReference()
srs.ImportFromEPSG(42104)

for m in dir(srs):
    if m.startswith('ExportTo'):
        print(m + ': ' + str(type(getattr(srs, m)())))

srs_wkt = srs.ExportToWkt()
print(srs_wkt)

drv = gdal.GetDriverByName('MEM')
ds = drv.Create('', 2, 3)
ds.SetProjection(srs_wkt)  # raises TypeError: in method 'Dataset_SetProjection', argument 2 of type 'char const *'

With output and error:

ExportToMICoordSys: <type 'str'>
ExportToPCI: <type 'list'>
ExportToPrettyWkt: <type 'unicode'>
ExportToProj4: <type 'str'>
ExportToUSGS: <type 'list'>
ExportToWkt: <type 'unicode'>
ExportToXML: <type 'unicode'>
PROJCS["NAD83 / MTM zone 8 Qubec",GEOGCS["GRS80",DATUM["GRS_1980",SPHEROID["GRS_1980",6378137,298.257222101]],PRIMEM["Greenwich",0],UNIT["Decimal_Degree",0.0174532925199433]],PROJECTION["Transverse_Mercator"],PARAMETER["latitude_of_origin",0],PARAMETER["central_meridian",-73.5],PARAMETER["scale_factor",0.9999],PARAMETER["false_easting",304800],PARAMETER["false_northing",0],UNIT["Meter",1],AUTHORITY["EPSG","42104"]]
Traceback (most recent call last):
  ...
  File "/usr/local/lib64/python/site-packages/GDAL-1.11.1-py2.6-linux-x86_64.egg/osgeo/gdal.py", line 722, in SetProjection
    return _gdal.Dataset_SetProjection(self, *args)
TypeError: in method 'Dataset_SetProjection', argument 2 of type 'char const *'

Note affected methods are ExportToPrettyWkt(), ExportToWkt() and ExportToXML(). Inspecting their output, no Unicode characters were actually found! However, cubewerx_extra.wkt (line 26) appears to be Unicode, as I see "Québec" in the name.

Furthermore, should SetProjection accept Unicode inputs? A simple workaround is just to encode the string in ASCII, e.g. ds.SetProjection(srs_wkt.encode('ascii')).

Python 3 is not affected.

[Even Rouault]

Easiest fix was r28779 "cubewerx_extra.wkt: replace Latin1 eacute character (first e in Quebec) by ASCII e (#5891)"

It appears this was the only occurrence on a non-ASCII character in WKT

[Even Rouault]

Milestone renamed