To comment on the following update, log in, then open the issue:
http://www.openoffice.org/issues/show_bug.cgi?id=64400





------- Additional comments from [EMAIL PROTECTED] Wed Sep  3 11:31:15 +0000 
2008 -------
About Lazlo Nemeth's comments from May, 13th 2008:

It is not because of the different spell checker implementation in StarOffice.
Especially not now since from OOo 3.0 SO uses the very same linguistic as OOo.

I see 3 possible solutions to this problem:

a) It can be handled by a grammar checker since that one gets to see the whole
   sentence and can mark only sub-strings of that text as wrong and provide
   new suggestions for those parts.
   But implementing a grammar checker just to solve this problem seems a 
   bit oversized.

Solutions b) and c) both require the patch by khong to pass the dash (and
probably it's unicode variants) along to the spell checker as a single word.
However because those words will than be handled by the spell checker and that
one does not know the context, words starting or ending with a dash (which is
not uncommon in German see SBA's examples) will not always be handled correctly.
Either they can be accepted in general or always reported as wrong...

In the above setting with the spell checker there are two variants:

a) the text with all the dashes gets passed on to the spell checker and the 
   spell checker is allowed to report only a substring of that text as 
   wrong and the suggestions are only replacements for that sub-string.
   
   Pros:
   The advantage would be that for a text like aa-xxx-bbbb where only xxx is
   wrong only that xxx part will get underlined and thus it is easy to see
   where sth. is wrong.
   
   Cons:
   That solution will require an new interface for spell checking results
   and all the main applications cores needs to be changed to handle the
   fact that spell checking may now only change a sub-string of the text.

c) the text with all the dashes gets passed on to the spell checker and the
   spell checker is only allowed to make new suggestions for that very same
   complete text. That is if the word is aa-xxx-bbbb and the xxx part
   should have been YYY than the suggestion has to be aa-YYY-bbbb.

   Pros: That can basically be done with the current API, and thus it needs
   no core changes in the applications.

   Cons: When the text gets longer it is not that easily to see where it 
   is incorrect, especially if no suggestion can be provided.
   We probably need to allow for only changing one misspelling even though
   there might be two different wrong parts in the text!!!

   Reasoning: if the text is xxx-aa-yy and xxx is wrong and yy as well.
   If you now usually would have 5 suggestions for xxx and 4 for yy, how
   are you going to display them??
   If you want to cover all of them you need to build 5*4 = 20 suggestions
   from them to cover all the cases.
   And the problem explodes with the number of errors found in the text!
   Thus in the end it will probably be reasonable that only one position
   gets handled/fixed and after applying the suggestion the text will
   still be incorrect and needs to be corrected once more. Once for each
   error in the text. And always the complete text will be marked as wrong.
   It may be a bit hard to explain to the user that applying a correct
   suggestion will still lead to a misspelled text, and that correcting a
   single 'word' may now be an iterative process.


Also in all of the three solutions the spell checker will be required to parse
the text for the dashes in it and if needs be break the text up and those
positions in order to see if each sub-string should be treated single and then
be recognized as correct.


I think the ways to go should be b) or c).
If we want a quick solution, that is one that can be implemented in time frame
with 3.01 or 3.1 I think the only choice will be c).

Thus if this one is to be addressed soon we need Karl's fix in order to get the
dashes passed on to the spell checker and at the same time the hunspell checker
is required to take care of them in proper way if needs be.


TL->Nemeth: What do you think about this?


---------------------------------------------------------------------
Please do not reply to this automatically generated notification from
Issue Tracker. Please log onto the website and enter your comments.
http://qa.openoffice.org/issue_handling/project_issues.html#notification

---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]


---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]

Reply via email to