Olaf,
> Could you please round this thread up by stating which solution you found works best for your original question, how to find missing records/misspelled words?
The approach I will be suggesting to Craig Boyd is to pre-parse the text being spellchecked into a cursor of words and then take multiple passes against this cursor of words, deleting words that match an ignore option (ignore words with digits, ignore all uppercase words, etc) and finally, deleting all words that match a dictionary table. The words that remain in this cursor are unmatched words that can be displayed to a user for review (correction, add, ignore, etc).
In the code snippet that follows, curWord is a cursor containing all words parsed from the source being spellchecked and curDictionary is the dictionary table being used to identify known words.
delete from curWord
where exists ( select * from curDictionary where curDictionary.cWord = curWord.cWord )
For those following this thread, I recognize that this approach is counter to the approach taken by many 3rd party and open-source spellcheckers which parse on sentence or word basis and then spellcheck each word individually. One of the main advantages of my approach is that it is EXTREMELY fast with plain text source (leverages ALINES parsing and RUSHMORE).
I will be sharing any enhancements via Craig Boyd's spellchecking code found on his blog at
www.sweetpotatosoftware.com/spsblog.
Malcolm