evaluate.py crashes after a detection with incorrect correction is found

[ebritoc]

The evaluation on the validation set crashes at line 62 when there is a misspelled word is correctly detected but incorrectly classified.
Example from page1037.json:

[DETECTION MATCH] page1037.text.list.1.item.6.s.1.w.3: hobbies -> Hobies
	[INCORRECT] Should be: hobby's
Traceback (most recent call last):
  File "./CLIN28_ST_spelling_correction/clin28tools/evaluate.py", line 93, in <module>
    main()
  File "./CLIN28_ST_spelling_correction/clin28tools/evaluate.py", line 62, in main
    print("[DETECTION WRONG] " + ";".join(outcorrection['span']) + ": " + " ".join([ outdata[wordid]['text'] for wordid in outcorrection['span'] ]) + " -> " + outcorrection['text'],file=sys.stderr)
  File "./CLIN28_ST_spelling_correction/clin28tools/evaluate.py", line 62, in <listcomp>
    print("[DETECTION WRONG] " + ";".join(outcorrection['span']) + ": " + " ".join([ outdata[wordid]['text'] for wordid in outcorrection['span'] ]) + " -> " + outcorrection['text'],file=sys.stderr)
  File "/home/ebritochac/anaconda3/lib/python3.6/site-packages/clin28tools/format.py", line 89, in __getitem__
    return self.data[key]
KeyError: 'page1037.text.list.1.item.1.s.1.w.1'

I changed that line by the following and it seems to work, but maybe I am misunderstanding what it is expected in the json file with the corrections:

print("[DETECTION WRONG] " + ";".join(outcorrection['span']) + ": " + " ".join([ refdata[wordid]['text'] for wordid in outcorrection['span'] ]) + " -> " + outcorrection['text'],file=sys.stderr)

[proycon]

I think this might have already been solved in the latest fix, can you try a pip install -U clin28tools and also try clin28-validator on your output?

[ebritoc]

I still get exactly the same KeyError with the original evaluate.py script after updating with pip.
My output JSON file seems to be valid:

$ ~/anaconda3/bin/python  validator.py ../data/results/page1037.json
valid <- ../data/results/page1037.json

[proycon]

Hmm, I'm still a bit puzzled why it goes wrong (hard to reproduce without having your output), but the way you changed the line is fine indeed (the reference text and output text should be the same, so this error shouldn't really happen). Anyway, I now included your fix in case anybody else stumbled on this problem.