Better autocomplete?

Is the autocomplete parcel included with VW state-of-the-art for
VisualWorks or is there something better?  I remember a type inferencing
system for VW (can't remember the name)...did that include an
autocomplete-r?

Not a rant, just hoping for feedback: I'm using VW in a class in the
Fall and these students are used to eclipse and Java.  They won't be
happy to have to go browse a class just to find a "constructor" for it,
for example.  Other things that the existing package could do better
include (please correct me if it already does these things):

  1) ranking/prioritizing selectors (if receiver is known, work from its
methodDict up hierarchy, if not know show recently added matching
selectors first etc) -- configurable with settings of course :-)
  2) providing class/method documentation (Squeak's autocompleter does a
very nice job of this)
  3) any little improvements giving the illusion of type inferencing
will be helpful (dog := Dog new, completing a message send to dog can
prioritize selectors in Dog higher than others) -- this one looks tough
given that the method isn't complete but even an ugly hack is better
than an alphabetical list of matching selectors
  4) I can only think of 3 things off the top of my head but when I set
my students loose on this I'm sure I'll have more

I see Travis published some iterations on this.  Any plans Travis?

David

_______________________________________________
vwnc mailing list
[hidden email]
http://lists.cs.uiuc.edu/mailman/listinfo/vwnc

As addition to the first point, the following things come to mind:

1. self and super: the receiver is known, so it is clear which selectors can be sent to it.
2. Classes: same applies here. So class methods only apply.


-- 
Claus Kick

"Wenn Sie mich suchen: Ich halte mich in der Nähe des Wahnsinns auf. 
Genauer gesagt auf der schmalen Linie zwischen Wahnsinn und Panik. 
Gleich um die Ecke von Todesangst, nicht weit weg von Irrwitz und Idiotie."

"If you are looking for me: I am somewhere near to lunacy. 
More clearly, on the narrow path between lunacy and panic. 
Right around the corner of  fear of death, 
not far away from idiocy and insanity."

-----Ursprüngliche Nachricht-----
Von: "C. David Shaffer" <[hidden email]>
Gesendet: 24.07.2010 18:18:24
An: vwnc <[hidden email]>
Betreff: [vwnc] Better autocomplete?

_______________________________________________
vwnc mailing list
[hidden email]
http://lists.cs.uiuc.edu/mailman/listinfo/vwnc

  I come thinking of a package called RBBlueMagic that is available in
Cincom StORE, but I neither know the status of it or haven't used it
much (I just tested it once two years ago).

Björn

On 2010-07-24 18:18, C. David Shaffer wrote: _______________________________________________
vwnc mailing list
[hidden email]
http://lists.cs.uiuc.edu/mailman/listinfo/vwnc

On 07/25/10 06:19, Claus Kick wrote:
> As addition to the first point, the following things come to mind:
>
> 1. self and super: the receiver is known, so it is clear which selectors can be sent to it.
> 2. Classes: same applies here. So class methods only apply.
>
>
>  
Yes, definitely.  The current package does a bit of this (if the
receiver is self, super or something that Compiler can compile) which is
why I didn't list them explicitly.  One problem with its current
implementation, though, is that it doesn't complete "blank" (no
characters type).  This is easily fixed but might cause problems for
people who want to type a tab.  ;-)  This might be difficult to sort out
(ie when "tab" means complete and when it just means tab) but it is
worth another try since the current code just assumes that if the
character to the left of the insertion point is blank.  So long as a
normal key is used for completion there will be some problems with
triggering autocompletion versus triggering insertion of that character
but we could:

Trigger insertion (not autocomplete) if we're on the left edge of the
code view or if all of the characters to the left (on the same line) are
the auto-complete character.  This would make it hard to complete an
expression that ended on the previous line:

    self (press enter)
tab tab tab(now I want to autocomplete)

You'd have to type a space then press tab to trigger autocompletion,
yuk.  But at least you /can/ trigger it if you are in the inner-circle
of autocomplete developers :-)

An alternative is to provide enhanced functionality (completing empty
selectors) to developers who are willing to use a non-printable as their
autocomplete trigger.  This would me permitting configuration of the
autocomplete trigger character.

David

_______________________________________________
vwnc mailing list
[hidden email]
http://lists.cs.uiuc.edu/mailman/listinfo/vwnc