ocrd-dummy: make copying optional and disable by default, #803

[kba]

No description provided.

[bertsky]

Something is rotten in the state of Denmark.

The behaviour used to be:

• for each input file representing a physical page (which could be PAGE-XML, a single image file or a single file of some other type),
  • if PAGE-XML, just update the pcGtsId and append to Metadata, writing a copy in the output fileGrp
  • otherwise, make a byte-by-byte copy in the output fileGrp; further, if it was an image file, create a new PAGE-XML referencing that image file as original

We were dissatisfied with that, because copying images is generally not necessary and also not typical for OCR-D. (Images are always re-referenced by new PAGE-XML. Only new images must be added – if the processor operates on images. But ocrd-dummy is not such a processor. You can use ocrd-preprocess-image for that kind of thing.)

IMHO all that was necessary was to add a parameter copy_files=false and change the non-PAGE case such that the copying is dependent on that parameter (plus the image reference filename is, because you must use the new filename iff you created a new one).

However, the new behaviour here becomes:

• for each input file representing a physical page (which could be PAGE-XML, a single image file or a single file of some other type),
  • if PAGE-XML
    • if copy_files=true, then update the pcGtsId and append to Metadata, writing a copy in the output fileGrp
    • if copy_files=false, then do nothing (!)
  • otherwise,
    • if copy_files=true, then make a byte-by-byte copy in the output fileGrp
    • if copy_files=false, then do nothing
    • further, if it was an image file, create a new PAGE-XML referencing that image file as original (but always the old image file!)

Why would you want to do this?

[kba]

OK, I have revised the logic. Now:

• If input file is PAGE-XML:
  • set new pcGtsId
  • write output filegroup
• if input file is anything else except PAGE-XML
  • if copy_files=true:
    • copy input file to output filegroup, byte-by-byte
  • if input file is an image:
    • create PAGE-XML for it and reference either the original image (if copy_files=false), otherwise the copy we just created, in imageFilename.

I think that should be what you proposed and what @stefanCCS wants.

[kba]

Failing CI is because macos, test passes for linux.