Hi Anthony,

Sorry for the late reply (trying to take care of backlog of emails today =) ).

Regarding functions, I don't know what is ideal in this case.  We
could add an 0xffff entry for functions to make them callable as
opcodes, but to expose the different implementations of a polymorphic
opcode seems tricky.

What to allow as a function is a tricky question.  It becomes syntax
sugar really, as it all gets compiled down to opcodes regardless.  So
imagine being able to call table and tablew as a function:

kcounter = 0
loop_start:
    tablew(table(kcounter, imytable) * 5, kcounter, imytable)
loop_lt kcounter, 1, 10, loop_start

versus:

kcounter = 0
loop_start:
  kval table kcounter, imytable
  tablew kval * 5, kcounter, imytable
loop_lt kcounter, 1, 10, loop_start

It's not perhaps the best example as in this case it doesn't save that
much, but it could be useful.

As for the fprintks example, would it work to call:

fprintks    "CAstates", "%f %f %f %f \\n", \
           tb0(k(0)), tb0(k(1)), tb0(k(2)), tb0(k(3))

Perhaps it would suffice to have some general coding guidelines,
perhaps put together a tips page in the manual?

steven

On 10/12/06, Anthony Kozar <anthonykozar@sbcglobal.net> wrote:
> I would prefer that we be able to call any function (or operator) using
> opcode syntax because it is more efficient.  There are times when you cannot
> use --expression-opt and might prefer to write
>
> aa  add  ab, ac
>
> instead of
>
> aa = ab + ac
>
> If I understand correctly, all that would be needed to allow this would be
> to add the 0xffff entries for all functions and operators.
>
> By the way, it WAS possible under Csound 4 to write the following:
>
> aa  add_aa  ab, ac
>
> in the orchestra code.  This would imply to me that when add_aa was changed
> to add.aa that the parser should have been changed to allow periods in the
> opcode name to keep "combatibility".  (Yes, I realize that it was an
> undocumented side effect of the way that things worked ... would be better
> to just add the 0xffff entries ...)
>
> I am not sure whether any opcode (with[out] a single argument) should be
> allowed to behave as a function.  I guess it really depends on whether you
> think that a function should always return the same value for the same input
> or whether it is OK for opcodes like readclock to be functions.
>
> Another point that I think should be considered changed is the way that the
> parser chooses between k-rate and i-rate versions of a function.  The
> current method of looking at the input argument's rate leads in my opinion
> to some non-intuitive and even moronic situations.
>
> For example, when using the tbN() family of functions with "vectorial
> signals", I have wanted to print the values of a function table like this:
>
> fprintks    "CAstates", "%f %f %f %f \\n", \
>             tb0(0), tb0(1), tb0(2), tb0(3)
>
> expecting that since the arguments to fprintks are k-rate, that I would get
> the k-rate version of tb0().  Instead you get the i-rate version and have to
> instead initialize a bunch of dummy k-rate variables:
>
> k0    init    0
> k1    init    1
> k2    init    2
> k3    init    3
> fprintks    "CAstates", "%f %f %f %f \\n", \
>             tb0(k0), tb0(k1), tb0(k2), tb0(k3)
>
> This is really annoying when you are trying to access many more than four
> table slots!  I really see no value in interpreting tb0() this way, but
> perhaps there are other functions or situations where this makes sense ...
> ??
>
> Anthony Kozar
> anthonykozar AT sbcglobal DOT net
>
>
> Michael Gogins wrote on 10/12/06 9:31 AM:
>
> > Going entirely function would break backwards compatibility, would it not?
> >
> > That is why I prefer having all opcodes be callable as functions and all
> > functions be callable as opcodes. Then those who prefer functions, like
> > myself, could go entirely functional without impeding the work of others who
> > do not.
>
>
> -------------------------------------------------------------------------
> Using Tomcat but need to do more? Need to support web services, security?
> Get stuff done quickly with pre-integrated technology to make your job easier
> Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo
> http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642
> _______________________________________________
> Csound-devel mailing list
> Csound-devel@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/csound-devel
>

-------------------------------------------------------------------------
Using Tomcat but need to do more? Need to support web services, security?
Get stuff done quickly with pre-integrated technology to make your job easier
Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo
http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642
_______________________________________________
Csound-devel mailing list
Csound-devel@lists.sourceforge.net