On Tue, 2004-03-02 at 06:43, Paul Kinnucan wrote: > Because no one has yet bothered to implement completion of class > names, which is a fundamentally different problem from completion of > class methods and fields.
I can understand that. I didn't want to sound offensive or demanding. > > I even think that the () shouldn't be inserted into this case, because > > most often what I try to do is something like FooBarBaz fbz = new > > FooBarBaz(...) with completion on the available classes. > > So supposing somebody took the trouble to implement completion of > class names, how do you think the user should convey to the JDEE that > they are seeking completion of the constructor as opposed to > completion of a class name? hmm? if the string before point is not already a full classname, but just a prefix to a class, do only class completion. If it is a complete class, and no prefix to another class, do constructor completion. If it is both, merge the two result sets. All of this of course assumes that the case where the string before point is a loval variable is already handled. I think the merging idea could possibly extended to taht, too: if the string before point is a prefix to a class and a locally defined variable, complete on both. if it is the whole variable, complete only methods, if it's a whole variable and part of a classname, merge the results again. But maybe this is simply too expensive to do. > > > > This is basically the only important thing left for me that makes > > somthing like eclipse "stronger" in the completion area. > > People have occasionally asked for completion of class names. All we > need now is someone willing to do the implementation.
signature.asc
Description: This is a digitally signed message part
