Comparing the different typescripts
After a few different experiments, I’ve settled on the following modus opandi for comparing typescripts.
I work page by page.
This fits well with the images, which have one page in each JPEG file.
I keep a subdirectory (called original) in which I store the typescripts exactly as received.
Most typists send each page in a separate file.
I have made a utility which chops a multipage submission into separate pages for comparison purposes,
but the files in this subdirectory are kept exactly as submitted.
When the first copy of a page arrives (say page 123), I make a copy in the main directory with filename ttq123.txt.
I use my assembler to detect errors that generate invalid Usercode, and correct them, often by reference to the image file.
When a subsequent copy arrives, I rename ttq123.txt as okq123.txt, compare it with the new page.
Where a difference in flagged it is normally compared with the original image.
Where the new page has invalid Usercode and the error is obvious (e.g. − instead of =),
I do not feel the need to go back to the original.
I edit okq123.txt to conform with the photo where the differences have been flagged.
Comparison software
I have written a simple global editing program htmform.c, originally intended for converting
ASCII represented Usercode into HTML so that it looks like KDF9 Flexowriter output.
I have a file of transformations which converts the various typists’ preferences for things such as multiply
and divide into a canonical form, and convert to upper case.
I have a .BAT file / shell script that uses this program on each of the candidates, and then runs it through:
diff -wiB ttt1.txt ttt2.txt > t.txt
The file t.txt flags up all significant differences.
I’ve only had about 2 false positives — caused by a defect in my file of transformations.
On my Win32 PC, I use the Cygnus version of diff.