Re: [Help-smalltalk] autocomplete in gst REPL

[Stefan Schmiedl]

On Thu, 02 Jul 2009 22:15:32 +0200
Paolo Bonzini <address@hidden> wrote:

> On 07/02/2009 08:34 PM, Stefan Schmiedl wrote:
> > While I tried out Nico's sample code, I noticed that
> > the tab-auto-complete in gst's REPL is not as all-knowing
> > as it could be:
> >
> > st>  error sign<TAB>
> > signal:               signalOn:             signalWithArguments:
> > signalClass:          signalWith:           signed:
> >
> > although error signalingContext is defined in my gst.
> >
> > Bug or feature?
> 
> Half-half.  The completion list is built from the symbol table, which 
> includes both variables and methods.  Including all methods including 
> unary methods would have meant having variables too in the completion 
> list.  I can change it if people prefer so.

Why do I use autocompletion?
- because I'm spoiled by bash's and vim's completion features
- because I can't remember how exactly something is spelled

In these cases, including all would make sense. Also as a non-intrusive
learning aid ... to this day I do not know whether ObjectMemory has
a method snapshot (for overwriting the image file), because whenever
I typed, snapshot: appeared and I obviously can't be bothered to look
for it. Were all symbols included I'd be much smarter .-)

I also noted that each part of multi-keyword messages is suggested
regardless of whether it is available.

st> Object extend [ abc: abc def: def ghi: ghi [] ]
st> 9 gh<TAB>
st> 9 ghi:

While catching that seems to be non-trivial for me, as the line editor
would have to learn about gst, would it be possible (and desirable)
to suggest multi-keyword messages as one item?

st> 9 abc:<TAB>
st> 9 abc:def:ghi:

Filling out that "template" would be easy enough, since Alt-b ends up
in exactly the right place to start typing.

Thanks,
s.

-- 
Stefan Schmiedl
EDV-Beratung Schmiedl, Berghangstr. 5, D-93413 Cham
im Büro: 09971 9966 989, am Handy: 0160 9981 6278