Thread: [Seed7-users] Dynamic dispatch strange behaviour
Interpreter and compiler for the Seed7 programming language.
Brought to you by:
thomas_mertes
Menu
▾
▴
seed7-users
|
From: Leonardo C. <leo...@gm...> - 2008-02-11 08:41:27
|
Hi.
I was reading the new manual chapter (it's the chapter I was looking
for) I discovered a strange issue.
I don't know if it's a bug but I can't explain to myself the behaviour
of the following program:
$ include "seed7_05.s7i";
const type: strangeInterface is new interface;
const func string: getNome(in strangeInterface param) is DYNAMIC;
const type: astronave is new struct
var string: nome is "";
end struct;
type_implements_interface(astronave, strangeInterface);
const func astronave: astronave(in string: nome) is func
result
var astronave: result is astronave.value;
begin
result.nome := nome;
end func;
const func string: getNome(in astronave: a1) is func
result
var string: result is "";
begin
result:=a1.nome;
end func;
const proc: main is func
local
var strangeInterface: enterprise is astronave("Enterprise galactica");
begin
writeln("one");
writeln("the name:" & getNome(enterprise));
writeln("two");
end func;
The result is:
C:\msys\1.0\home\lcecchi\seed7\src>hi provadue.sd7
HI INTERPRETER Version 4.5.3082 Copyright (c) 1990-2008 Thomas Mertes
259 C:/msys/1.0/home/lcecchi/seed7/lib/syntax.s7i
3395 C:/msys/1.0/home/lcecchi/seed7/lib/seed7_05.s7i
31 provadue.sd7
3685 lines total
118870 lines per second
1628430 bytes
one
So the execution is blocked at the line with "the name" and the line
who writes "two" it's never executed.
Where I'm wrong?
Thanks,
Leonardo
|
|
From: Thomas M. <tho...@gm...> - 2008-02-11 10:59:47
|
"Leonardo Cecchi" <leo...@gm...> wrote:
> I was reading the new manual chapter (it's the chapter I was looking
> for) I discovered a strange issue.
> I don't know if it's a bug but I can't explain to myself the behaviour
> of the following program:
>
> $ include "seed7_05.s7i";
>
> const type: strangeInterface is new interface;
> const func string: getNome(in strangeInterface param) is DYNAMIC;
>
> const type: astronave is new struct
> var string: nome is "";
> end struct;
> type_implements_interface(astronave, strangeInterface);
>
> const func astronave: astronave(in string: nome) is func
> result
> var astronave: result is astronave.value;
> begin
> result.nome := nome;
> end func;
>
> const func string: getNome(in astronave: a1) is func
> result
> var string: result is "";
> begin
> result:=a1.nome;
> end func;
>
> const proc: main is func
> local
> var strangeInterface: enterprise is astronave("Enterprise galactica");
> begin
> writeln("one");
> writeln("the name:" & getNome(enterprise));
> writeln("two");
> end func;
As far as I can see you discovered a bug.
It has to do with the initialisation of interface
variables. When you change the 'main' program to
const proc: main is func
local
var strangeInterface: enterprise is astronave("Enterprise galactica");
begin
enterprise := astronave("Enterprise galactica");
writeln("one");
writeln("the name:" & getNome(enterprise));
writeln("two");
end func;
it works (at least for me). Does it work for you?
I have not investigated more about the bug than that.
It is obvious that I need to revise the initialisation
of interface variables. In the meantime it is
necessary to assign a value with := to an interface
variable.
There is more to revise in this area:
It can happen that several interface variables
point to one implementation value. Since this is
sometimes desired and sometimes not I may move
to the use of an explicit 'new' function.
What do you think?
Greetings Thomas Mertes
Seed7 Homepage: http://seed7.sourceforge.net
Seed7 - The extensible programming language: User defined statements
and operators, abstract data types, templates without special
syntax, OO with interfaces and multiple dispatch.
--
Der GMX SmartSurfer hilft bis zu 70% Ihrer Onlinekosten zu sparen!
Ideal für Modem und ISDN: http://www.gmx.net/de/go/smartsurfer
|
|
From: Thomas M. <tho...@gm...> - 2008-02-11 21:49:24
|
"Leonardo Cecchi" <leo...@gm...> wrote (off-list):
> I've another question about dynamic dispatching.
> In the new chapter you say:
>
> "Although there is a 'draw' function declared for 'circle' it is not
> called directly. Instead the 'draw' interface function for 'shape' is
> called. When the program is analyzed (in the interpreter or compiler)
> *interface functions take precedence over normal functions* when both
> are to be considered."
>
> I've made a little test:
>
> $ include "seed7_05.s7i";
>
> const type: strangeInterface is new interface;
> const func string: getNome(in strangeInterface param) is DYNAMIC;
>
> const type: astronave is new struct
> var string: nome is "";
> end struct;
> type_implements_interface(astronave, strangeInterface);
>
> const type: astronaveBetter is sub astronave struct
> var integer: power is 0;
> end struct;
> type_implements_interface(astronaveBetter, strangeInterface);
>
> const func astronave: astronave(in string: nome) is func
> result
> var astronave: result is astronave.value;
> begin
> result.nome := nome;
> end func;
>
> const func string: getNome(in astronave: a1) is func
> result
> var string: result is "";
> begin
> result:=a1.nome;
> end func;
>
> const func astronaveBetter: astronaveBetter(in string: nome, in
> integer:power) is func
> result
> var astronaveBetter: result is astronaveBetter.value;
> begin
> result.nome := nome;
> result.power := power;
> end func;
>
> const func string: getNome(in astronaveBetter: a1) is func
> result
> var string: result is "";
> begin
> result:=a1.nome & " better";
> end func;
>
> const proc: main is func
> local
> var astronaveBetter: myBetter is astronaveBetter("Enterprise
> galactica ", 1500);
> var strangeInterface: enterprise is astronave.value;
> var astronave: myAstro is astronave.value;
> begin
> enterprise:=myBetter;
> myAstro:=myBetter;
>
> writeln(getNome(enterprise));
> writeln(getNome(myAstro));
> end func;
>
> The output of this program is the following:
>
> C:\msys\1.0\home\lcecchi\seed7\src>hi provatre.sd7
> HI INTERPRETER Version 4.5.3082 Copyright (c) 1990-2008 Thomas Mertes
> 259 C:/msys/1.0/home/lcecchi/seed7/lib/syntax.s7i
> 3395 C:/msys/1.0/home/lcecchi/seed7/lib/seed7_05.s7i
> 55 provatre.sd7
> 3709 lines total
> 119645 lines per second
> 1673829 bytes
> Enterprise galactica better
> Enterprise galactica
>
> I don't understand why it doesn't do a dynamic dispatching on the
> second instruction.
> The functions to be considered are:
>
> const func string: getNome(in strangeInterface param) is DYNAMIC;
> const func string: getNome(in astronave: a1);
> const func string: getNome(in astronaveBetter: a1);
>
> So the interface should have precedence on the other functions
> considering that either "astronave" or "astronaveBetter" implements
> the interface "strangeInterface".
> I haven't understood something... Where I'm wrong?
> Thanks,
>
> Leonardo
The local variable 'myAstro' has the type 'astronave' which is
a structure. Assignments of structures do a (deep) copy. The struct 'astronaveBetter' has more elements than 'astronave', but only the
elements present in both structures are copied. The assignment
'myAstro:=myBetter' does a copy of the structure (The element 'nome'
is copied). With that copy the information that 'myBetter' had the
type 'astronaveBetter' is lost.
My description of the process:
"Although there is a 'draw' function declared for 'circle' it is not
called directly. Instead the 'draw' interface function for 'shape' is
called. When the program is analyzed (in the interpreter or compiler)
*interface functions take precedence over normal functions* when both
are to be considered."
is right, but does not have the desired effect here.
The variable 'myAstro' is not a reference to a value, but the
value itself. The process described above makes sense for parameters
(except for a value parameter where the value is copied).
For structs the parameters 'in', 'ref' and 'inout' are reference
parameters. In this cases a parameter of type 'astronave' can point to
a value of type 'astronaveBetter'. In this cases a parameter of type
'astronave' behaves iside a function (method) like a parameter of type
'strangeInterface'. If you add to your example:
const func integer: getPower (in strangeInterface param) is DYNAMIC;
const func integer: getPower (in astronave: a1) is
return 0;
const func integer: getPower (in astronaveBetter: a2) is
return a2.power;
const func string: getPowerName (in astronave: a1) is
return getNome(a1) & "-" & str(getPower(a1));
and add to the main program:
writeln(getPowerName(myBetter));
you will get:
Enterprise galactica better-1500
which is exactly what you expect.
Something is clear now: My description of the process
of dynamic dispatch is not precise enough and needs improvement.
May be my idea of having value copy for implementation objects
and reference copy for interface objects is too confusing.
May be I should describe the role of assignments in more detail.
What do you think about my concepts and their description?
BTW.: If possible I would prefer the discussion to be on
see...@li...
That way others can also follow the discussion.
Greetings Thomas Mertes
Seed7 Homepage: http://seed7.sourceforge.net
Seed7 - The extensible programming language: User defined statements
and operators, abstract data types, templates without special
syntax, OO with interfaces and multiple dispatch.
--
Psssst! Schon vom neuen GMX MultiMessenger gehört?
Der kann`s mit allen: http://www.gmx.net/de/go/multimessenger
|
|
From: Leonardo C. <leo...@gm...> - 2008-02-12 11:02:47
|
2008/2/11, Thomas Mertes <tho...@gm...>: > The variable 'myAstro' is not a reference to a value, but the > value itself. The process described above makes sense for parameters > (except for a value parameter where the value is copied). > For structs the parameters 'in', 'ref' and 'inout' are reference > parameters. You have been really clear. In the manual chapter about parameters (chapter 6) I haven't found anything about structures but only about arrays. "For arrays 'in' parameters are equal to 'ref' parameters" Probably the structs behaviour may be inferred from what you said about arrays and from the sources but I think that specifying it would be better. > Something is clear now: My description of the process > of dynamic dispatch is not precise enough and needs improvement. I think the description is good: I simply wasn't aware that implementation objects have value-copy semantics. > May be my idea of having value copy for implementation objects > and reference copy for interface objects is too confusing. > May be I should describe the role of assignments in more detail. > What do you think about my concepts and their description? I think the idea it's not confusing but can be better documented. Thanks, Leonardo |
×
Want the latest updates on software, tech news, and AI?
Get latest updates about software, tech news, and AI from SourceForge directly in your inbox once a month.
Thanks for helping keep SourceForge clean.
X