Error splitting ALTO XML file during ocrWF

[estanfor]

Honeybadger alert: https://app.honeybadger.io/projects/52894/faults/123219313

XML file looks OK at first glance. The only weird thing I see is that in the ABBYY output folder, there are two XML files, one with the druid name and one with the image filename: S:\AbbyyShare\sdr-ocr-prod\OUTPUT\sh099tn7500

The SDR item sh099tn7500 is currently stuck in ocrWF in the split-ocr step. It looks like it might have started to split out the individual pages of OCR XML from the combined XML and ran into an error while it was doing that?

[edsu]

The split-ocr step in the workflow has some code to get that list out, and use it to name the individual OCR XML files that are extracted from the multipage Abbyy result XML document:

common-accessioning/lib/dor/text_extraction/abbyy/split_alto.rb

Lines 64 to 66 in 71afcdd

	def page_filenames
	@page_filenames ||= Nokogiri::XML(page_filename_nodes).css('//sourceImageInformation').children.select(&:element?).map(&:text)
	end

Usually it looks like this:

<sourceImageInformation>
  <fileIdentifier>bb222cc3333_00_0001.tif</fileIdentifier>
  <fileIdentifier>bb222cc3333_00_0002.tif</fileIdentifier>
  <fileIdentifier>bb222cc3333_00_0003.tif</fileIdentifier>
</sourceImageInformation>

However, I took at look at some other Abbyy output files that we successfully processed and these didn't have <sourceImageInformation> elements. So perhaps the code is mistakenly trying to split OCR that doesn't need to be split?

I've got a test that can reproduce the problem.

[edsu]

I had been looking at XML in /abbyy/RESULTXML instead of /abbyy/OUTPUT. I can see that in this case there is a <sourceImageInformation> element that contains one filename, but there are two <page> elements. The second looks virtually empty with the only text content being 1*.

The test in #1550 demonstrates the problem, but does not yet propose a fix, since I'm not sure what should happen. @estanfor & @andrewjbtw will probably have to weigh in here to decide what the best course of action is for these.

[estanfor]

Stopping when the count equals the number of files makes sense to me, but Andrew will probably have more perspective on how to handle this--whether we want to address it by discarding thumbnail layers before the images get to ABBYY.

[dnoneill]

@edsu Are you sure there wasn't just a re-versioning issue when things broke. The resultxml has two files but if you look at the item in argo there is only one file and the output only has one output. The output is correct. The resultxml is wrong but that might have had to do with the server going down

[edsu]

Perhaps we can try running this item through ocrWF in Stage and see if it repeats?

[dnoneill]

That is what I am thinking.

[edsu]

@estanfor did notice it was a multipage TIFF with Imagemagick, so perhaps that explains why this is happening if it is repeatable.