output_file_grp not set to ocrd-tool.json default for processors

[kba]

e.g.

{
  output_file_grp: "OCR-D-IMG-BIN",
}

but in processor's process method:

print(self.output_file_grp)

prints OUTPUT not OCR-D-IMG-BIN.

[kba]

@crater2150 reports the same problem for input_file_grp

[kba]

Related https://github.com/OCR-D/ocrd_tesserocr/pull/75/files#r321639953

But I agree with you all these errors should be reported properly and as early as possible. So either the Processor base class or process_cli / ocrd_cli_wrap_processor should check in ocrd-tool for the number of outputs. We discussed this in the spec, but it is not official, and not implemented yet. Also related to #296.

[bertsky]

I cannot find the reference, but I think we agreed that fileGrp names listed in the ocrd-tool.json are merely examples, not fallbacks. Users always have to specifiy these on the command line. So can we close this?

[kba]

I think we agreed that fileGrp names listed in the ocrd-tool.json are merely examples,

OCR-D/spec#121

-I/-O is currently only required in bashlib, the Python Processor implementations still have fallbacks (INPUT/OUTPUT). Once those are gone, this can be closed.