Hi all,
> Have an indicator somewhere, to tell the user which grammar checker is > being used. > This can affect the confidence that the user has, in the accuracy of > the error report. When considering chained grammar checkers that would be the appropriate mean to find out for which grammar checker the rules have to modified for other results (in case that it reports errors that are no errors that is). This also implies that grammar checkers should have some (hopefully unique) way to have them identified in the UI. ->Bruno: The respective thing for current spell checkers is the XServiceDisplayName that was introduced for that purpose. > Ok, it could work, but believe me, in some place the errors should be > cached, now we got these problems: > - cache problem, it is not good to store several objects in memory > since it can compromise the performance > - redrawing wavy lines in oo, after grammar checker change it will > redraw all over again and it can compromise the performance Shouldn't the change be only internally? And thus the wavy lines only drawn once. After all the grammar checker itself should not draw the lines but the instance using it. And that instance may decide to call several grammar checkers before it draws any underline. Also I think there probably has to be some super grammar checker that knows of the actual implementations and dispatches the calls to them depending on the language it is called with. It will also be the instance that takes care of chaining grammar checkers for the same language. The corresponding object for spellchecking etc. in the current linguistc would be the LinguServiceManager. ->Bruno: Please read the IDL description for the service com.sun.star.linguistic2.LinguServiceManager and it's interfaces if you have not yet done so. > - ok several checkers are installed but they are not running in the > same time since the user has to activate them. If we use the split model > and send text to it proper grammar checker it can run all of em in the > same time. This is wrong. Consider two different documents with different languages both of them may get checked at the same time (at least when one of them is checked automatically). Thus to the user those two grammar checkers will virtually run parallel. And if you actually have a multi-processor environment this may actually happen as well. And if you consider the API since it is designed to run simultaneously to whatever the user does or some other entity using it, it is easily possible. This is a serious problem when the grammar checker is required to have a consisting state over several calls. For example if it's state is effected by the previous paragraph or sentence that was handled in a different call. > Let the user set the grammar checker to be interactive, and throw all > text to that grammar checker. When the user selects "check entire > document", then sequentially throw the paragraphs at the grammar > checkers. > > > same problems as above, i think if we want a check entire document it > shoult start checking from beginning, sentence by sentence, what do you > think? There are some possible optimizations similar to the ones applied for spell checking. E.g. it would be possible to keep a flag that indicates the paragraph was already fully checked and nothing has been changed since then. Such kind of optimizations seems desirable since it prevents the checking of already corrected paragraphs and thus speeds up things. On the other hand: If we apply such optimizations the order of paragraphs being checked will not necessarily be the same as when one would read them. Thomas --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
