Hi,
To implement this I have used a Cell ("FoundWord", marked *** below),
assigned initially to "false", which is set to "true", if the first
branch succeeds, hoping that it could be used as a guard, to prevent the
second choice.
Trouble is, it doesn't work. All the rest of the parser works fine, but
"FoundWord" always seems to be "false" when accessed in the second
choice, even if it certain that it has been previously assigned true.
Which makes me realise that I have completely failed to understand
something, probably to do with the scoping, but my logic might be wrong,
too, who knows...
Yes, the cell's value is logically "backtracked" as well as everything
else when you reach the second choice. This is because "choice" creates
computational spaces, try reading chapter 12 of the language tutorial.
Here's the basic code:
proc {Encode NumberStr Words PrevIsNum}
First Rest Word WList CurrIsNum
FoundWord = {NewCell false} % <---- ***
in
choice
NumberStr = nil
Words = nil
[]
Words = Word|WList
{Splitter First Rest NumberStr} % NonDet - Works like a kind of
backward append/3
choice
%% NonDet - "Corresp" is based on member/2 and checks a dict
{Corresp First Word}
CurrIsNum = false
{Assign FoundWord true} % <---- ***
[]
% there was no word, use the single digit
{Length First 1}
% check that we are not following a digit
PrevIsNum = false
{Access FoundWord false} % <---- ***
First = Word
CurrIsNum = true
end
{Encode Rest WList CurrIsNum} % Recursive Call
end
end
What about this:
proc {Encode NumberStr Words PrevIsNum}
First Rest Word WList CurrIsNum
FoundWord
in
choice
NumberStr = nil
Words = nil
[]
Words = Word|WList
{Splitter First Rest NumberStr}
choice
{Corresp First Word}
CurrIsNum = false
FoundWord=true
[]
FoundWord=false
end
if {Not FoundWord} then
% there was no word, use the single digit
{Length First 1}
% check that we are not following a digit
PrevIsNum = false
First = Word
CurrIsNum = true
end
{Encode Rest WList CurrIsNum} % Recursive Call
end
end
I hope I didn't screw it up :-)
Cheers,
Filip
_________________________________________________________________________________
mozart-users mailing list
[email protected]
http://www.mozart-oz.org/mailman/listinfo/mozart-users
