DataElement_from_raw exception in 0.9.8

Hi,

I have written a module to hold the tags from a set of dicom files which worked 
with pydicom 0.9.7.  (Thanks!)  With 0.9.8,

value = dicom.dataelem.DataElement_from_raw(value)

throws (with debugging info appended):

TypeError: "'NoneType' object is unsubscriptable"
> /usr/lib/python2.6/site-packages/dicom/values.py(191)convert_value()
    190         # Text VRs use the 2nd specified encoding

--> 191         value = converter(byte_string, is_little_endian, 
encoding=encoding[1])
    192     elif VR != "SQ":

ipdb> byte_string
'<RegExProg>head\\ARIC_NCS_HUMAN_3T_SKYRA_VD11_D_MRQ\\32 coil Human</RegExProg>'
ipdb> is_little_endian
True
ipdb> encoding
ipdb> default_encoding
'iso8859'
ipdb> encoding
ipdb> u
> /usr/lib/python2.6/site-packages/dicom/dataelem.py(338)DataElement_from_raw()
    337     try:
2-> 338         value = convert_value(VR, raw, encoding)
    339     except NotImplementedError as e:

ipdb> encoding
ipdb> l
    333             elif raw.tag.element == 0:  # group length tag implied in versions < 3.0
    334                 VR = 'UL'
    335             else:
    336                 raise KeyError("Unknown DICOM tag {0:s} - can't look up VR".format(str(raw.tag)))
    337     try:
2-> 338         value = convert_value(VR, raw, encoding)
    339     except NotImplementedError as e:
    340         raise NotImplementedError("{0:s} in tag {1!r}".format(str(e), raw.tag))
    341     return DataElement(raw.tag, VR, value, raw.value_tell,
    342                     raw.length == 0xFFFFFFFF, already_converted=True)

What happened is that I loaded a DICOM file including the tag

(0010,4000) LT [<RegExProg>head\ARIC_NCS_HUMAN_3T_SKYRA_VD11_D_MRQ\32 coil 
Human</... #  76, 1 PatientComments

dataelem.DataElement_from_raw's signature is 
DataElement_from_raw(raw_data_element, encoding=None) and calls 
values.convert_value with convert_value(VR, raw, encoding).  The default 
encoding for values.convert_value is not None, however, and it can't deal with 
it.

I worked around this in my code by changing the DataElement_from_raw call to

from dicom.charset import default_encoding
value = dicom.dataelem.DataElement_from_raw(value, default_encoding)

(not being enough of an encoding expert to know what else to use).

Thanks,

Rob Reid

Original issue reported on code.google.com by barlen...@gmail.com on 5 Feb 2013 at 3:59

Rob,

This was introduced in 0.9.8 as we strive toward python 3 compatibility. The 
biggest thing in python 3 is converting between raw bytes and string literals. 
We actually perform this conversion inside of DataElement_from_raw.

Internally when we use DataElement_from_raw, the second argument is the 
specific character set for the containing Dataset (Dataset._character_set). In 
many cases, the default_encoding should work (iso8859) but it will definitely 
fail in some.

All that being said. Our goal with 0.9.8 was to add python 3 compatibility 
without drastically altering the expected behavior in python 2. In python 2, we 
don't actually convert from string literals to unicode in DataElement_from_raw, 
so really the encoding parameter is unnecessary. However, that also means that 
the default value for the encoding parameter should not result in an exception. 
I'll have to think about what the default value should be (because we want to 
force python 3 users to enter a value still), but in the meantime as 
work-around, you can pass either ds._character_set or 
dicom.charset.default_encoding as the second parameter.

-Jonathan

Original comment by Suever@gmail.com on 5 Feb 2013 at 4:41

ivoflipse / pydicom

DataElement_from_raw exception in 0.9.8 #122