Non-Latin symbols in channel names

[plankter]

I don't know if it is an issue with tifffile or xtiff library, but when channel names contain non-Latin symbols then to_tiff method crashes:

xtiff/xtiff.py

Line 311 in d0a7b5c

writer.save(data=img, photometric='MINISBLACK', compress=compression, description=description,

Traceback (most recent call last):
  File "/home/anton/bblab/imctools/imctools/converters/mcdfolder2imcfolder.py", line 59, in mcdfolder_to_imcfolder
    imc_writer.write_imc_folder(create_zip=create_zip)
  File "/home/anton/bblab/imctools/imctools/io/imc/imcwriter.py", line 77, in write_imc_folder
    acquisition_data.save_ome_tiff(
  File "/home/anton/bblab/imctools/imctools/data/acquisitiondata.py", line 136, in save_ome_tiff
    to_tiff(
  File "/home/anton/bblab/imctools/venv/lib/python3.8/site-packages/xtiff.py", line 311, in to_tiff
    writer.save(data=img, photometric='MINISBLACK', compress=compression, description=description,
  File "/home/anton/bblab/imctools/venv/lib/python3.8/site-packages/tifffile/tifffile.py", line 1836, in write
    addtag(270, 's', 0, description, writeonce=True)
  File "/home/anton/bblab/imctools/venv/lib/python3.8/site-packages/tifffile/tifffile.py", line 1772, in addtag
    value = bytestr(value, 'ascii') + b'\0'
  File "/home/anton/bblab/imctools/venv/lib/python3.8/site-packages/tifffile/tifffile.py", line 15622, in bytestr
    return s.encode(encoding) if isinstance(s, str) else s
UnicodeEncodeError: 'ascii' codec can't encode character '\u03b5' in position 1241: ordinal not in range(128)

[jwindhager]

Interesting. I'd vaguely guess that this is a problem with tifffile. Happy to take pull requests though, should you find a problem in our code, @plankter.

[plankter]

My assumption is that an encoding in this code

xtiff/xtiff.py

Line 301 in d0a7b5c

with BytesIO() as description_buffer:

should be ascii instead of utf-8. Then it works. This is due to https://github.com/cgohlke/tifffile/blob/75a11f2d781705a64151d3d329e47187e3be82a8/tifffile/tifffile.py#L1264 in tifffile, which states that "The subject of the image. Must be 7-bit ASCII." But I am also confused that it tells "Cannot be used with the ImageJ or OME formats."

with BytesIO() as description_buffer:
            ome_xml.write(description_buffer, encoding='ascii', xml_declaration=True)
            description = description_buffer.getvalue().decode('ascii')

[plankter]

Example MCD file with non-Latin channel name characters is available at server_homes/thoch/Data/Data_for_Anton/Slide_7_679.zip

[jwindhager]

But I am also confused that it tells "Cannot be used with the ImageJ or OME formats."

To my understanding, this only applies to situations in which the OME-XML is generated by tifffile itself. As xtiff specifies the OME-XML as a description for a "regular" tiff written by tifffile, this shouldn't be an issue.

[jwindhager]

with BytesIO() as description_buffer:
            ome_xml.write(description_buffer, encoding='ascii', xml_declaration=True)
            description = description_buffer.getvalue().decode('ascii')

Implemented this in v0.6.4, please upgrade

[jwindhager]

@plankter switched back to UTF-8 in v0.7.6 (just released). While the TIFF standard technically only supports ASCII, OME-XML requires UTF-8, see #8.. The error message reported in this issue should not occur anymore though, as the encoding is now handled by xtiff.