well... I cannot. But for me, most of the time OCompletion suggestions are accurate (because I tend to use always the same methods, probably). btw, I'm not saying that OCompletion is perfect, I think it can be improved a lot... but I like it (what you did is already better, but it can be improved more) On Aug 6, 2012, at 1:02 PM, Camillo Bruni <camillobruni@gmail.com> wrote:

Maybe I don't really see when it gets useful... Can you give me an example where you rely on the history information of OCompletion?

On 2012-08-06, at 12:34, Esteban Lorenzano <estebanlm@gmail.com> wrote:

...

OCompletion works fine for me. And is much more valuable for me than EC... yes we can (and need) improve it, but I don't want to drop OC. Is the same case of nautilus many-methods and tabs. Just because you don't find the function valuable, that does not means that it is not worthy for anybody :)

Esteban

On Aug 6, 2012, at 12:15 AM, Camillo Bruni <camillobruni@gmail.com> wrote:

...

the more time I spend with OCompletion the more I think it's the wrong model:

NECompletion provides in many cases way better / quicker results

NOCompletion does tons of magic to increase the chances of a correct hit, neglecting - self sends - super sends - ClassSends

which is basically the main source for getting nice input...

I think NOCompletion should be removed. The only thing that might maintain is a Database which will give back a date for a given selector so we can sort them accordingly. But IMO that can be solved muuuuch easier.

Things against NOCompletion: - not documentation whatsoever (no class comments no comments no categories... nada) - totally breaks type completion (self sends usw...) - duplication of NECompletion...

Since I don't see anybody maintaining this dead body we should remove it from the system!