Julia

So you propose an inplace=TRUE/FALSE entry for each
argument to each function which may may want to avoid
allocating memory?  The major problem is that the function
writer has no idea what the value of inplace should be,
as it depends on how the function gets called.  This makes
writing reusable functions (hence packages) difficult.

Bill Dunlap
Spotfire, TIBCO Software
wdunlap tibco.com

-----Original Message-----
From: oliver [mailto:oliver at first.in-berlin.de]
Sent: Thursday, March 08, 2012 7:40 AM
To: William Dunlap
Cc: R-devel
Subject: Re: [Rd] Julia

Ah, and you mean if it's an anonymous array it could be reused directly from the
args.

OK, now I see why you insist on the anonymous data thing.
I didn't grasped it even in my last mail.



But that somehow also relates to what I wrote about reusing an already
existing, named vector.

Just the moment of in-place-modification is different.

From
  x  <- runif(n)
  cx <- cos(x)

instead of

    cx <- cos(x=runif(n)) # no allocation needed, use the input
space for the return value

to something like

  cx  <- runif(n)
  cos( cx, inplace=TRUE)

or

  cos( runif(n), inplace=TRUE)




This way it would be possible to specify the reusage of the input *explicitly*
(without  implicit rules like anonymous vs. named values).



In Pseudo-Code something like that:

   if (in_place == TRUE )
   {
     input_val[idx] = cos( input_val[idx] );
     return input_val;
   }
   else
   {
     result_val = alloc_vec( LENGTH(input_val), ... );
     result_val[idx] = cos( input_val[idx] );
     return result_val;
   }



Is this matching, what you were looking for?


Ciao,
   Oliver


On Thu, Mar 08, 2012 at 02:56:24PM +0100, oliver wrote:

Hi,

ok, thank you for clarifiying what you meant.
You only referred to the reusage of the args, not of an already
existing vector.
So I overgenerealized your example.

But when looking at your example,
and how I would implement the cos()
I doubt I would use copying the args
before calculating the result.

Just allocate a result-vector, and then place the cos() of the
input-vector into the result vector.

I didn't looked at how it is done in R, but I would guess it's like
that.


  In pseudo-Code something like that:
    cos_val[idx] = cos( input_val[idx] );

But R also handles complex data with cos() so it will look a bit more
laborious.

What I have seen so far from implementing C-extensions for R is rather
C-ish, and so you have the control on many details. Copying the input
just to read it would not make sense here.

I doubt that R internally is doing that.
Or did you found that in the R-code?

The other problem, someone mentioned, was *changing* the contents of a
matrix... and that this is NO>T done in-place, when using a function
for it.
But the namespace-name / variable-name as "references" to the matrix
might solve that problem.


Ciao,
  Oliver



On Wed, Mar 07, 2012 at 07:10:43PM +0000, William Dunlap wrote:

No my examples are what I meant.  My point was that a function, say
cos(), can act like it does call-by-value but conserve memory when
it can  if it can distinguish between the case
    cx <- cos(x=runif(n)) # no allocation needed, use the input
space for the return value and and the case
   x <- runif(n)
   cx <- cos(x=x) # return value cannot reuse the argument's memory, so

allocate space for return value

   sum(x)              # Otherwise sum(x) would return sum(cx)
The function needs to know if a memory block is referred to by a
name in any environment in order to do that.

Bill Dunlap
Spotfire, TIBCO Software
wdunlap tibco.com

-----Original Message-----
From: oliver [mailto:oliver at first.in-berlin.de]
Sent: Wednesday, March 07, 2012 10:22 AM
To: Dominick Samperi
Cc: William Dunlap; R-devel
Subject: Re: [Rd] Julia

On Tue, Mar 06, 2012 at 12:49:32PM -0500, Dominick Samperi wrote:

On Tue, Mar 6, 2012 at 11:44 AM, William Dunlap
<wdunlap at tibco.com>

wrote:

S (and its derivatives and successors) promises that functions
will not change their arguments, so in an expression like
? val <- func(arg)
you know that arg will not be changed. ?You can do that by
having func copy arg before doing anything, but that uses
space and time that you want to conserve.
If arg is not a named item in any environment then it should
be fine to write over the original because there is no way the
caller can detect that shortcut. ?E.g., in
? ?cx <- cos(runif(n))
the cos function does not need to allocate new space for its
output, it can just write over its input because, without a
name attached to it, the caller has no way of looking at what
runif(n) returned. ?If you did
? ?x <- runif(n)
? ?cx <- cos(x)

You have two names here, x and cx, hence your example does not fit
into what you want to explain.

A better example would be:
x <- runif(n)
x <- cos(x)

then cos would have to allocate new space for its output
because overwriting its input would affect a subsequent
? ?sum(x)
I suppose that end-users and function-writers could learn to
live with having to decide when to copy, but not having to
make that decision makes S more pleasant (and safer) to use.
I think that is a major reason that people are able to share S
code so easily.

But don't forget the "Holy Grail" that Doug mentioned at the
start of this thread: finding a flexible language that is also
fast. Currently many R packages employ C/C++ components to
compensate for the fact that the R interpreter can be slow, and
the pass-by-value semantics of S provides no protection here.

[...]

The distinction imperative vs. functional has nothing to do with
the distinction interpreted vs. directly executed.




Thinking again on the problem that was mentioned here, I think it
might be circumvented.

Looking again at R's properties, looking again into U.Ligges
"Programmieren in R", I saw there was mentioned that in R anything
(?!) is an object... so then it's OOP; but also it was mentioned,
R is a functional language. But this does not mean it's purely functional or

has no imperative data structures.

As R relies heavily on vectors, here we have an imperative datastructure.

So, it rather looks to me that "<-" does work in-place on the vectors, even

"<-"

itself is a function (which does not matter for the problem).

If thats true (I assume here, it is; correct me, if it's wrong),
then I think, assigning with "<<-" and assign() also would do an
imperative
(in-place) change of the contents.

Then the copying-of-big-objects-when-passed-as-args problem can be
circumvented by working on either a variable in the GlobalEnv (and
using "<<-", or using a certain environment for the big data and
passing it's name (and the
variable) as value to the function which then uses assign() and
get() to work on that data.
Then in-place modification should be possible.

In 2008 Ross Ihaka and Duncan Temple Lang published the paper
"Back to the Future: Lisp as a base for a statistical computing
system" where they propose Common Lisp as a new foundation for
R. They suggest that this could be done while maintaining the same

familiar R syntax.

A key requirement of any strategy is to maintain easy access to
the huge universe of existing C/C++/Fortran numerical and
graphics libraries, as these libraries are not likely to be rewritten.

Thus there will always be a need for a foreign function
interface, and the problem is to provide a flexible and
type-safe language that does not force developers to use another
unfamiliar, less flexible, and error-prone language to optimize the hot

spots.

If I here "type safe" I rather would think about OCaml or maybe
Ada, but not LISP.

Also, LISP has so many "("'s and ")"'s, that it's making people
going crazy ;-)

Ciao,
   Oliver

-----Original Message-----
From: oliver [mailto:oliver at first.in-berlin.de]
Sent: Thursday, March 08, 2012 2:23 PM
To: William Dunlap
Cc: R-devel
Subject: Re: [Rd] Julia

I don't think that using in-place modification as a general property would make
sense.

In-place modification brings in side-effects and that would mean that the order
of evaluation can change the result.

To get reliable results, the order of evaluation should not be the reason for
different results, and thats the reason, why the functional approach is much
better for reliable programs.

So, in general I would say, this feature is a no-no.
In general I would rather discourage in-place modification.

For some certain cases it might help...
but for such certain cases either such a boolean flag or programming a sparate
module in C would make sense.

There could also be a global in-place-flag that might be used (via options
maybe) but if such a thing would be implemented, the default value should be
FALSE.



Ciao,
   Oliver


On Thu, Mar 08, 2012 at 04:21:42PM +0000, William Dunlap wrote:

So you propose an inplace=TRUE/FALSE entry for each argument to each
function which may may want to avoid allocating memory?  The major
problem is that the function writer has no idea what the value of
inplace should be, as it depends on how the function gets called.
This makes writing reusable functions (hence packages) difficult.

Bill Dunlap
Spotfire, TIBCO Software
wdunlap tibco.com

-----Original Message-----
From: oliver [mailto:oliver at first.in-berlin.de]
Sent: Thursday, March 08, 2012 7:40 AM
To: William Dunlap
Cc: R-devel
Subject: Re: [Rd] Julia

Ah, and you mean if it's an anonymous array it could be reused
directly from the args.

OK, now I see why you insist on the anonymous data thing.
I didn't grasped it even in my last mail.



But that somehow also relates to what I wrote about reusing an
already existing, named vector.

Just the moment of in-place-modification is different.

From
  x  <- runif(n)
  cx <- cos(x)

instead of

    cx <- cos(x=runif(n)) # no allocation needed, use the input
space for the return value

to something like

  cx  <- runif(n)
  cos( cx, inplace=TRUE)

or

  cos( runif(n), inplace=TRUE)




This way it would be possible to specify the reusage of the input
*explicitly* (without  implicit rules like anonymous vs. named values).



In Pseudo-Code something like that:

   if (in_place == TRUE )
   {
     input_val[idx] = cos( input_val[idx] );
     return input_val;
   }
   else
   {
     result_val = alloc_vec( LENGTH(input_val), ... );
     result_val[idx] = cos( input_val[idx] );
     return result_val;
   }



Is this matching, what you were looking for?


Ciao,
   Oliver


On Thu, Mar 08, 2012 at 02:56:24PM +0100, oliver wrote:

Hi,

ok, thank you for clarifiying what you meant.
You only referred to the reusage of the args, not of an already
existing vector.
So I overgenerealized your example.

But when looking at your example,
and how I would implement the cos() I doubt I would use copying
the args before calculating the result.

Just allocate a result-vector, and then place the cos() of the
input-vector into the result vector.

I didn't looked at how it is done in R, but I would guess it's
like that.


  In pseudo-Code something like that:
    cos_val[idx] = cos( input_val[idx] );

But R also handles complex data with cos() so it will look a bit
more laborious.

What I have seen so far from implementing C-extensions for R is
rather C-ish, and so you have the control on many details. Copying
the input just to read it would not make sense here.

I doubt that R internally is doing that.
Or did you found that in the R-code?

The other problem, someone mentioned, was *changing* the contents
of a matrix... and that this is NO>T done in-place, when using a
function for it.
But the namespace-name / variable-name as "references" to the
matrix might solve that problem.


Ciao,
  Oliver



On Wed, Mar 07, 2012 at 07:10:43PM +0000, William Dunlap wrote:

No my examples are what I meant.  My point was that a function,
say cos(), can act like it does call-by-value but conserve
memory when it can  if it can distinguish between the case
    cx <- cos(x=runif(n)) # no allocation needed, use the input
space for the return value and and the case
   x <- runif(n)
   cx <- cos(x=x) # return value cannot reuse the argument's
memory, so

allocate space for return value

   sum(x)              # Otherwise sum(x) would return sum(cx)
The function needs to know if a memory block is referred to by a
name in any environment in order to do that.

Bill Dunlap
Spotfire, TIBCO Software
wdunlap tibco.com

-----Original Message-----
From: oliver [mailto:oliver at first.in-berlin.de]
Sent: Wednesday, March 07, 2012 10:22 AM
To: Dominick Samperi
Cc: William Dunlap; R-devel
Subject: Re: [Rd] Julia

On Tue, Mar 06, 2012 at 12:49:32PM -0500, Dominick Samperi wrote:

On Tue, Mar 6, 2012 at 11:44 AM, William Dunlap
<wdunlap at tibco.com>

wrote:

S (and its derivatives and successors) promises that
functions will not change their arguments, so in an
expression like
? val <- func(arg)
you know that arg will not be changed. ?You can do that by
having func copy arg before doing anything, but that uses
space and time that you want to conserve.
If arg is not a named item in any environment then it
should be fine to write over the original because there is
no way the caller can detect that shortcut. ?E.g., in
? ?cx <- cos(runif(n))
the cos function does not need to allocate new space for
its output, it can just write over its input because,
without a name attached to it, the caller has no way of
looking at what
runif(n) returned. ?If you did
? ?x <- runif(n)
? ?cx <- cos(x)

You have two names here, x and cx, hence your example does not
fit into what you want to explain.

A better example would be:
x <- runif(n)
x <- cos(x)

then cos would have to allocate new space for its output
because overwriting its input would affect a subsequent
? ?sum(x)
I suppose that end-users and function-writers could learn
to live with having to decide when to copy, but not having
to make that decision makes S more pleasant (and safer) to use.
I think that is a major reason that people are able to
share S code so easily.

But don't forget the "Holy Grail" that Doug mentioned at the
start of this thread: finding a flexible language that is
also fast. Currently many R packages employ C/C++ components
to compensate for the fact that the R interpreter can be
slow, and the pass-by-value semantics of S provides no protection

here.

[...]

The distinction imperative vs. functional has nothing to do
with the distinction interpreted vs. directly executed.




Thinking again on the problem that was mentioned here, I think
it might be circumvented.

Looking again at R's properties, looking again into U.Ligges
"Programmieren in R", I saw there was mentioned that in R
anything
(?!) is an object... so then it's OOP; but also it was
mentioned, R is a functional language. But this does not mean
it's purely functional or

has no imperative data structures.

As R relies heavily on vectors, here we have an imperative

datastructure.

So, it rather looks to me that "<-" does work in-place on the
vectors, even

"<-"

itself is a function (which does not matter for the problem).

If thats true (I assume here, it is; correct me, if it's
wrong), then I think, assigning with "<<-" and assign() also
would do an imperative
(in-place) change of the contents.

Then the copying-of-big-objects-when-passed-as-args problem
can be circumvented by working on either a variable in the
GlobalEnv (and using "<<-", or using a certain environment for
the big data and passing it's name (and the
variable) as value to the function which then uses assign()
and
get() to work on that data.
Then in-place modification should be possible.

In 2008 Ross Ihaka and Duncan Temple Lang published the
paper "Back to the Future: Lisp as a base for a statistical
computing system" where they propose Common Lisp as a new
foundation for R. They suggest that this could be done while
maintaining the same

familiar R syntax.

A key requirement of any strategy is to maintain easy access
to the huge universe of existing C/C++/Fortran numerical and
graphics libraries, as these libraries are not likely to be rewritten.

Thus there will always be a need for a foreign function
interface, and the problem is to provide a flexible and
type-safe language that does not force developers to use
another unfamiliar, less flexible, and error-prone language
to optimize the hot

spots.

If I here "type safe" I rather would think about OCaml or
maybe Ada, but not LISP.

Also, LISP has so many "("'s and ")"'s, that it's making
people going crazy ;-)

Ciao,
   Oliver

I guess my point is not getting across.  The user should see
the functional programming style but under the hood the
evaluator should be able to use whatever memory and time
saving tricks it can.  Julia seems to want to be a nonfunctional
language, which I think makes it harder to write the sort of
easily reusable functions that S allows.

Bill Dunlap
Spotfire, TIBCO Software
wdunlap tibco.com

-----Original Message-----
From: oliver [mailto:oliver at first.in-berlin.de]
Sent: Thursday, March 08, 2012 2:23 PM
To: William Dunlap
Cc: R-devel
Subject: Re: [Rd] Julia

I don't think that using in-place modification as a general property would make
sense.

In-place modification brings in side-effects and that would mean that the order
of evaluation can change the result.

To get reliable results, the order of evaluation should not be the reason for
different results, and thats the reason, why the functional approach is much
better for reliable programs.

So, in general I would say, this feature is a no-no.
In general I would rather discourage in-place modification.

For some certain cases it might help...
but for such certain cases either such a boolean flag or programming a sparate
module in C would make sense.

There could also be a global in-place-flag that might be used (via options
maybe) but if such a thing would be implemented, the default value should be
FALSE.



Ciao,
   Oliver


On Thu, Mar 08, 2012 at 04:21:42PM +0000, William Dunlap wrote:

So you propose an inplace=TRUE/FALSE entry for each argument to each
function which may may want to avoid allocating memory?  The major
problem is that the function writer has no idea what the value of
inplace should be, as it depends on how the function gets called.
This makes writing reusable functions (hence packages) difficult.

Bill Dunlap
Spotfire, TIBCO Software
wdunlap tibco.com

-----Original Message-----
From: oliver [mailto:oliver at first.in-berlin.de]
Sent: Thursday, March 08, 2012 7:40 AM
To: William Dunlap
Cc: R-devel
Subject: Re: [Rd] Julia

Ah, and you mean if it's an anonymous array it could be reused
directly from the args.

OK, now I see why you insist on the anonymous data thing.
I didn't grasped it even in my last mail.



But that somehow also relates to what I wrote about reusing an
already existing, named vector.

Just the moment of in-place-modification is different.

From
  x  <- runif(n)
  cx <- cos(x)

instead of

    cx <- cos(x=runif(n)) # no allocation needed, use the input
space for the return value

to something like

  cx  <- runif(n)
  cos( cx, inplace=TRUE)

or

  cos( runif(n), inplace=TRUE)




This way it would be possible to specify the reusage of the input
*explicitly* (without  implicit rules like anonymous vs. named values).



In Pseudo-Code something like that:

   if (in_place == TRUE )
   {
     input_val[idx] = cos( input_val[idx] );
     return input_val;
   }
   else
   {
     result_val = alloc_vec( LENGTH(input_val), ... );
     result_val[idx] = cos( input_val[idx] );
     return result_val;
   }



Is this matching, what you were looking for?


Ciao,
   Oliver


On Thu, Mar 08, 2012 at 02:56:24PM +0100, oliver wrote:

Hi,

ok, thank you for clarifiying what you meant.
You only referred to the reusage of the args, not of an already
existing vector.
So I overgenerealized your example.

But when looking at your example,
and how I would implement the cos() I doubt I would use copying
the args before calculating the result.

Just allocate a result-vector, and then place the cos() of the
input-vector into the result vector.

I didn't looked at how it is done in R, but I would guess it's
like that.


  In pseudo-Code something like that:
    cos_val[idx] = cos( input_val[idx] );

But R also handles complex data with cos() so it will look a bit
more laborious.

What I have seen so far from implementing C-extensions for R is
rather C-ish, and so you have the control on many details. Copying
the input just to read it would not make sense here.

I doubt that R internally is doing that.
Or did you found that in the R-code?

The other problem, someone mentioned, was *changing* the contents
of a matrix... and that this is NO>T done in-place, when using a
function for it.
But the namespace-name / variable-name as "references" to the
matrix might solve that problem.


Ciao,
  Oliver



On Wed, Mar 07, 2012 at 07:10:43PM +0000, William Dunlap wrote:

No my examples are what I meant.  My point was that a function,
say cos(), can act like it does call-by-value but conserve
memory when it can  if it can distinguish between the case
    cx <- cos(x=runif(n)) # no allocation needed, use the input
space for the return value and and the case
   x <- runif(n)
   cx <- cos(x=x) # return value cannot reuse the argument's
memory, so

allocate space for return value

   sum(x)              # Otherwise sum(x) would return sum(cx)
The function needs to know if a memory block is referred to by a
name in any environment in order to do that.

Bill Dunlap
Spotfire, TIBCO Software
wdunlap tibco.com

-----Original Message-----
From: oliver [mailto:oliver at first.in-berlin.de]
Sent: Wednesday, March 07, 2012 10:22 AM
To: Dominick Samperi
Cc: William Dunlap; R-devel
Subject: Re: [Rd] Julia

On Tue, Mar 06, 2012 at 12:49:32PM -0500, Dominick Samperi wrote:

On Tue, Mar 6, 2012 at 11:44 AM, William Dunlap
<wdunlap at tibco.com>

wrote:

S (and its derivatives and successors) promises that
functions will not change their arguments, so in an
expression like
? val <- func(arg)
you know that arg will not be changed. ?You can do that by
having func copy arg before doing anything, but that uses
space and time that you want to conserve.
If arg is not a named item in any environment then it
should be fine to write over the original because there is
no way the caller can detect that shortcut. ?E.g., in
? ?cx <- cos(runif(n))
the cos function does not need to allocate new space for
its output, it can just write over its input because,
without a name attached to it, the caller has no way of
looking at what
runif(n) returned. ?If you did
? ?x <- runif(n)
? ?cx <- cos(x)

You have two names here, x and cx, hence your example does not
fit into what you want to explain.

A better example would be:
x <- runif(n)
x <- cos(x)

then cos would have to allocate new space for its output
because overwriting its input would affect a subsequent
? ?sum(x)
I suppose that end-users and function-writers could learn
to live with having to decide when to copy, but not having
to make that decision makes S more pleasant (and safer) to use.
I think that is a major reason that people are able to
share S code so easily.

But don't forget the "Holy Grail" that Doug mentioned at the
start of this thread: finding a flexible language that is
also fast. Currently many R packages employ C/C++ components
to compensate for the fact that the R interpreter can be
slow, and the pass-by-value semantics of S provides no protection

here.

[...]

The distinction imperative vs. functional has nothing to do
with the distinction interpreted vs. directly executed.




Thinking again on the problem that was mentioned here, I think
it might be circumvented.

Looking again at R's properties, looking again into U.Ligges
"Programmieren in R", I saw there was mentioned that in R
anything
(?!) is an object... so then it's OOP; but also it was
mentioned, R is a functional language. But this does not mean
it's purely functional or

has no imperative data structures.

As R relies heavily on vectors, here we have an imperative

datastructure.

So, it rather looks to me that "<-" does work in-place on the
vectors, even

"<-"

itself is a function (which does not matter for the problem).

If thats true (I assume here, it is; correct me, if it's
wrong), then I think, assigning with "<<-" and assign() also
would do an imperative
(in-place) change of the contents.

Then the copying-of-big-objects-when-passed-as-args problem
can be circumvented by working on either a variable in the
GlobalEnv (and using "<<-", or using a certain environment for
the big data and passing it's name (and the
variable) as value to the function which then uses assign()
and
get() to work on that data.
Then in-place modification should be possible.

In 2008 Ross Ihaka and Duncan Temple Lang published the
paper "Back to the Future: Lisp as a base for a statistical
computing system" where they propose Common Lisp as a new
foundation for R. They suggest that this could be done while
maintaining the same

familiar R syntax.

A key requirement of any strategy is to maintain easy access
to the huge universe of existing C/C++/Fortran numerical and
graphics libraries, as these libraries are not likely to be rewritten.

Thus there will always be a need for a foreign function
interface, and the problem is to provide a flexible and
type-safe language that does not force developers to use
another unfamiliar, less flexible, and error-prone language
to optimize the hot

spots.

If I here "type safe" I rather would think about OCaml or
maybe Ada, but not LISP.

Also, LISP has so many "("'s and ")"'s, that it's making
people going crazy ;-)

Ciao,
   Oliver

S (and its derivatives and successors) promises that functions
will not change their arguments, so in an expression like
   val <- func(arg)
you know that arg will not be changed.  You can
do that by having func copy arg before doing anything,
but that uses space and time that you want to conserve.
If arg is not a named item in any environment then it
should be fine to write over the original because there
is no way the caller can detect that shortcut.  E.g., in
    cx <- cos(runif(n))
the cos function does not need to allocate new space for
its output, it can just write over its input because, without
a name attached to it, the caller has no way of looking
at what runif(n) returned.  If you did
    x <- runif(n)
    cx <- cos(x)
then cos would have to allocate new space for its output
because overwriting its input would affect a subsequent
    sum(x)
I suppose that end-users and function-writers could learn
to live with having to decide when to copy, but not having
to make that decision makes S more pleasant (and safer) to use.
I think that is a major reason that people are able to
share S code so easily.

Bill Dunlap
Spotfire, TIBCO Software
wdunlap tibco.com 

-----Original Message-----
From: oliver [mailto:oliver at first.in-berlin.de]
Sent: Tuesday, March 06, 2012 1:12 AM
To: William Dunlap
Cc: Herv? Pag?s; R-devel
Subject: Re: [Rd] Julia

On Tue, Mar 06, 2012 at 12:35:32AM +0000, William Dunlap wrote:
[...]

I find R's (& S+'s & S's) copy-on-write-if-not-copying-would-be-discoverable-
by-the-uer machanism for giving the allusion of pass-by-value a good way
to structure the contract between the function writer and the function user.

[...]


Can you elaborate more on this,
especially on the ...-...-...-if-not-copying-would-be-discoverable-by-the-uer
stuff?

What do you mean with discoverability of not-copying?

Ciao,
   Oliver

Hello,

regarding the copying issue,
I would like to point to the

"Writing R-Extensions" documentation.

There it is mentio9ned, that functions of extensions
that use the .C interface normally do get their arguments
pre-copied...


In section 5.2:

   "There can be up to 65 further arguments giving R objects to be
   passed to compiled code. Normally these are copied before being
   passed in, and copied again to an R list object when the compiled
   code returns."

But for the .Call and .Extension interfaces this is NOT the case.



In section 5.9:
   "The .Call and .External interfaces allow much more control, but
   they also impose much greater responsibilities so need to be used
   with care. Neither .Call nor .External copy their arguments. You
   should treat arguments you receive through these interfaces as
   read-only."


Why is read-only preferred?

Please, see the discussion in section 5.9.10.

It's mentioned there, that a copy of an object in the R-language
not necessarily doies a real copy of that object, but instead of
this, just a "rerference" to the real data is created (two names
referring to one bulk of data). That's typical functional
programming: not a variable, but a name (and possibly more than one
name) bound to an object.


Of course, if yo change the orgiginal named value, when there
would be no copy of it, before changing it, then both names
would refer to the changed data.
of course that is not, what is wanted.

But what you also can see in section 5.9.10 is, that
there already is a mechanism (reference counting) that allows
to distinguish between unnamed and named object.

So, this is directly adressing the points you have mentioned in your
examples.

So, at least in principial, R allows to do in-place modifications
of object with the .Call interface.

You seem to refer to the .C interface, and I had explored the .Call
interface. That's the reason why you may insist on "it's copyied
always" and I wondered, what you were talking about, because the
.Call interface allowed me rather C-like raw style of programming
(and the user of it to decide, if copying will be done or not).

The mechanism to descide, if copying should be done or not,
also is mentioined in section 5.9.10: NAMED and SET_NAMED macros.
with NAMED you can get the number of references.

But later in that section it is mentioned, that - at least for now -
NAMED always returns the value 2.


   "Currently all arguments to a .Call call will have NAMED set to 2,
   and so users must assume that they need to be duplicated before
   alteration."
                (section 5.9.10, last sentence)


So, the in-place modification can be done already with the .Call
intefcae for example. But the decision if it is safe or not
is not supported at the moment.

So the situation is somewhere between: "it is possible" and
"R does not support a safe decision if, what is possible, also
can be recommended".
At the moment R rather deprecates in-place modification by default
(the save way, and I agree with this default).

But it's not true, that R in general copies arguments.

But this seems to be true for the .C interface.

Maybe a lot of performance-/memory-problems can be solved
by rewriting already existing packages, by providing them
via .Call instead of .C.

Ciao,
    Oliver




On Tue, Mar 06, 2012 at 04:44:49PM +0000, William Dunlap wrote:

S (and its derivatives and successors) promises that functions
will not change their arguments, so in an expression like
    val<- func(arg)
you know that arg will not be changed.  You can
do that by having func copy arg before doing anything,
but that uses space and time that you want to conserve.
If arg is not a named item in any environment then it
should be fine to write over the original because there
is no way the caller can detect that shortcut.  E.g., in
     cx<- cos(runif(n))
the cos function does not need to allocate new space for
its output, it can just write over its input because, without
a name attached to it, the caller has no way of looking
at what runif(n) returned.  If you did
     x<- runif(n)
     cx<- cos(x)
then cos would have to allocate new space for its output
because overwriting its input would affect a subsequent
     sum(x)
I suppose that end-users and function-writers could learn
to live with having to decide when to copy, but not having
to make that decision makes S more pleasant (and safer) to use.
I think that is a major reason that people are able to
share S code so easily.

Bill Dunlap
Spotfire, TIBCO Software
wdunlap tibco.com

-----Original Message-----
From: oliver [mailto:oliver at first.in-berlin.de]
Sent: Tuesday, March 06, 2012 1:12 AM
To: William Dunlap
Cc: Herv? Pag?s; R-devel
Subject: Re: [Rd] Julia

On Tue, Mar 06, 2012 at 12:35:32AM +0000, William Dunlap wrote:
[...]

I find R's (&  S+'s&  S's) copy-on-write-if-not-copying-would-be-discoverable-
by-the-uer machanism for giving the allusion of pass-by-value a good way
to structure the contract between the function writer and the function user.

[...]


Can you elaborate more on this,
especially on the ...-...-...-if-not-copying-would-be-discoverable-by-the-uer
stuff?

What do you mean with discoverability of not-copying?

Ciao,
    Oliver

So the situation is somewhere between: "it is possible" and
"R does not support a safe decision if, what is possible, also
can be recommended".
At the moment R rather deprecates in-place modification by default
(the save way, and I agree with this default).

But it's not true, that R in general copies arguments.

But this seems to be true for the .C interface.

Maybe a lot of performance-/memory-problems can be solved
by rewriting already existing packages, by providing them
via .Call instead of .C.

My understanding is that most packages use the .C interface
because it's simpler to deal with and because they don't need
to pass complicated objects at the C level, just atomic vectors.
My guess is that it's probably rarely the case that the cost
of copying the arguments passed to .C is significant, but,
if that was the case, then they could always call .C() with
DUP=FALSE. However, using DUP=FALSE is dangerous (see Warning
section in the man page).

No need to switch to .Call

Hi Oliver,

On 03/17/2012 08:35 AM, oliver wrote:

Hello,

regarding the copying issue,
I would like to point to the

"Writing R-Extensions" documentation.

There it is mentio9ned, that functions of extensions
that use the .C interface normally do get their arguments
pre-copied...


In section 5.2:

  "There can be up to 65 further arguments giving R objects to be
  passed to compiled code. Normally these are copied before being
  passed in, and copied again to an R list object when the compiled
  code returns."

But for the .Call and .Extension interfaces this is NOT the case.



In section 5.9:
  "The .Call and .External interfaces allow much more control, but
  they also impose much greater responsibilities so need to be used
  with care. Neither .Call nor .External copy their arguments. You
  should treat arguments you receive through these interfaces as
  read-only."


Why is read-only preferred?

Please, see the discussion in section 5.9.10.

It's mentioned there, that a copy of an object in the R-language
not necessarily doies a real copy of that object, but instead of
this, just a "rerference" to the real data is created (two names
referring to one bulk of data). That's typical functional
programming: not a variable, but a name (and possibly more than one
name) bound to an object.


Of course, if yo change the orgiginal named value, when there
would be no copy of it, before changing it, then both names
would refer to the changed data.
of course that is not, what is wanted.

But what you also can see in section 5.9.10 is, that
there already is a mechanism (reference counting) that allows
to distinguish between unnamed and named object.

So, this is directly adressing the points you have mentioned in your
examples.

So, at least in principial, R allows to do in-place modifications
of object with the .Call interface.

You seem to refer to the .C interface, and I had explored the .Call
interface. That's the reason why you may insist on "it's copyied
always" and I wondered, what you were talking about, because the
.Call interface allowed me rather C-like raw style of programming
(and the user of it to decide, if copying will be done or not).

The mechanism to descide, if copying should be done or not,
also is mentioined in section 5.9.10: NAMED and SET_NAMED macros.
with NAMED you can get the number of references.

But later in that section it is mentioned, that - at least for now -
NAMED always returns the value 2.


  "Currently all arguments to a .Call call will have NAMED set to 2,
  and so users must assume that they need to be duplicated before
  alteration."
               (section 5.9.10, last sentence)


So, the in-place modification can be done already with the .Call
intefcae for example. But the decision if it is safe or not
is not supported at the moment.

So the situation is somewhere between: "it is possible" and
"R does not support a safe decision if, what is possible, also
can be recommended".
At the moment R rather deprecates in-place modification by default
(the save way, and I agree with this default).

But it's not true, that R in general copies arguments.

But this seems to be true for the .C interface.

Maybe a lot of performance-/memory-problems can be solved
by rewriting already existing packages, by providing them
via .Call instead of .C.

My understanding is that most packages use the .C interface
because it's simpler to deal with and because they don't need
to pass complicated objects at the C level, just atomic vectors.
My guess is that it's probably rarely the case that the cost
of copying the arguments passed to .C is significant, but,
if that was the case, then they could always call .C() with
DUP=FALSE. However, using DUP=FALSE is dangerous (see Warning
section in the man page).

No need to switch to .Call

Cheers,
H.

Ciao,
   Oliver




On Tue, Mar 06, 2012 at 04:44:49PM +0000, William Dunlap wrote:

S (and its derivatives and successors) promises that functions
will not change their arguments, so in an expression like
   val<- func(arg)
you know that arg will not be changed.  You can
do that by having func copy arg before doing anything,
but that uses space and time that you want to conserve.
If arg is not a named item in any environment then it
should be fine to write over the original because there
is no way the caller can detect that shortcut.  E.g., in
    cx<- cos(runif(n))
the cos function does not need to allocate new space for
its output, it can just write over its input because, without
a name attached to it, the caller has no way of looking
at what runif(n) returned.  If you did
    x<- runif(n)
    cx<- cos(x)
then cos would have to allocate new space for its output
because overwriting its input would affect a subsequent
    sum(x)
I suppose that end-users and function-writers could learn
to live with having to decide when to copy, but not having
to make that decision makes S more pleasant (and safer) to use.
I think that is a major reason that people are able to
share S code so easily.

Bill Dunlap
Spotfire, TIBCO Software
wdunlap tibco.com

-----Original Message-----
From: oliver [mailto:oliver at first.in-berlin.de]
Sent: Tuesday, March 06, 2012 1:12 AM
To: William Dunlap
Cc: Herv? Pag?s; R-devel
Subject: Re: [Rd] Julia

On Tue, Mar 06, 2012 at 12:35:32AM +0000, William Dunlap wrote:
[...]

I find R's (&  S+'s&  S's) copy-on-write-if-not-copying-would-be-discoverable-
by-the-uer machanism for giving the allusion of pass-by-value a good way
to structure the contract between the function writer and the function user.

[...]


Can you elaborate more on this,
especially on the ...-...-...-if-not-copying-would-be-discoverable-by-the-uer
stuff?

What do you mean with discoverability of not-copying?

Ciao,
   Oliver

-- 
Herv? Pag?s

Program in Computational Biology
Division of Public Health Sciences
Fred Hutchinson Cancer Research Center
1100 Fairview Ave. N, M1-B514
P.O. Box 19024
Seattle, WA 98109-1024

E-mail: hpages at fhcrc.org
Phone:  (206) 667-5791
Fax:    (206) 667-1319

On Mar 20, 2012, at 3:08 PM, Herv? Pag?s wrote:

Hi Oliver,

On 03/17/2012 08:35 AM, oliver wrote:

Hello,

regarding the copying issue,
I would like to point to the

"Writing R-Extensions" documentation.

There it is mentio9ned, that functions of extensions
that use the .C interface normally do get their arguments
pre-copied...


In section 5.2:

   "There can be up to 65 further arguments giving R objects to be
   passed to compiled code. Normally these are copied before being
   passed in, and copied again to an R list object when the compiled
   code returns."

But for the .Call and .Extension interfaces this is NOT the case.



In section 5.9:
   "The .Call and .External interfaces allow much more control, but
   they also impose much greater responsibilities so need to be used
   with care. Neither .Call nor .External copy their arguments. You
   should treat arguments you receive through these interfaces as
   read-only."


Why is read-only preferred?

Please, see the discussion in section 5.9.10.

It's mentioned there, that a copy of an object in the R-language
not necessarily doies a real copy of that object, but instead of
this, just a "rerference" to the real data is created (two names
referring to one bulk of data). That's typical functional
programming: not a variable, but a name (and possibly more than one
name) bound to an object.


Of course, if yo change the orgiginal named value, when there
would be no copy of it, before changing it, then both names
would refer to the changed data.
of course that is not, what is wanted.

But what you also can see in section 5.9.10 is, that
there already is a mechanism (reference counting) that allows
to distinguish between unnamed and named object.

So, this is directly adressing the points you have mentioned in your
examples.

So, at least in principial, R allows to do in-place modifications
of object with the .Call interface.

You seem to refer to the .C interface, and I had explored the .Call
interface. That's the reason why you may insist on "it's copyied
always" and I wondered, what you were talking about, because the
.Call interface allowed me rather C-like raw style of programming
(and the user of it to decide, if copying will be done or not).

The mechanism to descide, if copying should be done or not,
also is mentioined in section 5.9.10: NAMED and SET_NAMED macros.
with NAMED you can get the number of references.

But later in that section it is mentioned, that - at least for now -
NAMED always returns the value 2.


   "Currently all arguments to a .Call call will have NAMED set to 2,
   and so users must assume that they need to be duplicated before
   alteration."
                (section 5.9.10, last sentence)


So, the in-place modification can be done already with the .Call
intefcae for example. But the decision if it is safe or not
is not supported at the moment.

So the situation is somewhere between: "it is possible" and
"R does not support a safe decision if, what is possible, also
can be recommended".
At the moment R rather deprecates in-place modification by default
(the save way, and I agree with this default).

But it's not true, that R in general copies arguments.

But this seems to be true for the .C interface.

Maybe a lot of performance-/memory-problems can be solved
by rewriting already existing packages, by providing them
via .Call instead of .C.

My understanding is that most packages use the .C interface
because it's simpler to deal with and because they don't need
to pass complicated objects at the C level, just atomic vectors.
My guess is that it's probably rarely the case that the cost
of copying the arguments passed to .C is significant, but,
if that was the case, then they could always call .C() with
DUP=FALSE. However, using DUP=FALSE is dangerous (see Warning
section in the man page).

No need to switch to .Call

I strongly disagree. I'm appalled to see that sentence here.

The overhead is significant for any large vector and it is in particular unnecessary since in .C you have to allocate *and copy* space even for results (twice!). Also it is very error-prone, because you have no information about the length of vectors so it's easy to run out of bounds and there is no way to check. IMHO .C should not be used for any code written in this century (the only exception may be if you are passing no data, e.g. if all you do is to pass a flag and expect no result, you can get away with it even if it is more dangerous). It is a legacy interface that dates way back and is essentially just re-named .Fortran interface. Again, I would strongly recommend the use of .Call in any recent code because it is safer and more efficient (if you don't care about either attribute, well, feel free ;)).

Cheers,
Simon

Cheers,
H.

Ciao,
    Oliver




On Tue, Mar 06, 2012 at 04:44:49PM +0000, William Dunlap wrote:

S (and its derivatives and successors) promises that functions
will not change their arguments, so in an expression like
    val<- func(arg)
you know that arg will not be changed.  You can
do that by having func copy arg before doing anything,
but that uses space and time that you want to conserve.
If arg is not a named item in any environment then it
should be fine to write over the original because there
is no way the caller can detect that shortcut.  E.g., in
     cx<- cos(runif(n))
the cos function does not need to allocate new space for
its output, it can just write over its input because, without
a name attached to it, the caller has no way of looking
at what runif(n) returned.  If you did
     x<- runif(n)
     cx<- cos(x)
then cos would have to allocate new space for its output
because overwriting its input would affect a subsequent
     sum(x)
I suppose that end-users and function-writers could learn
to live with having to decide when to copy, but not having
to make that decision makes S more pleasant (and safer) to use.
I think that is a major reason that people are able to
share S code so easily.

Bill Dunlap
Spotfire, TIBCO Software
wdunlap tibco.com

-----Original Message-----
From: oliver [mailto:oliver at first.in-berlin.de]
Sent: Tuesday, March 06, 2012 1:12 AM
To: William Dunlap
Cc: Herv? Pag?s; R-devel
Subject: Re: [Rd] Julia

On Tue, Mar 06, 2012 at 12:35:32AM +0000, William Dunlap wrote:
[...]

I find R's (&   S+'s&   S's) copy-on-write-if-not-copying-would-be-discoverable-
by-the-uer machanism for giving the allusion of pass-by-value a good way
to structure the contract between the function writer and the function user.

[...]


Can you elaborate more on this,
especially on the ...-...-...-if-not-copying-would-be-discoverable-by-the-uer
stuff?

What do you mean with discoverability of not-copying?

Ciao,
    Oliver

--
Herv? Pag?s

Program in Computational Biology
Division of Public Health Sciences
Fred Hutchinson Cancer Research Center
1100 Fairview Ave. N, M1-B514
P.O. Box 19024
Seattle, WA 98109-1024

E-mail: hpages at fhcrc.org
Phone:  (206) 667-5791
Fax:    (206) 667-1319

On 03/21/2012 06:23 PM, Simon Urbanek wrote:

On Mar 20, 2012, at 3:08 PM, Herv? Pag?s wrote:

Hi Oliver,

On 03/17/2012 08:35 AM, oliver wrote:

Hello,

regarding the copying issue,
I would like to point to the

"Writing R-Extensions" documentation.

There it is mentio9ned, that functions of extensions
that use the .C interface normally do get their arguments
pre-copied...


In section 5.2:

  "There can be up to 65 further arguments giving R objects to be
  passed to compiled code. Normally these are copied before being
  passed in, and copied again to an R list object when the compiled
  code returns."

But for the .Call and .Extension interfaces this is NOT the case.



In section 5.9:
  "The .Call and .External interfaces allow much more control, but
  they also impose much greater responsibilities so need to be used
  with care. Neither .Call nor .External copy their arguments. You
  should treat arguments you receive through these interfaces as
  read-only."


Why is read-only preferred?

Please, see the discussion in section 5.9.10.

It's mentioned there, that a copy of an object in the R-language
not necessarily doies a real copy of that object, but instead of
this, just a "rerference" to the real data is created (two names
referring to one bulk of data). That's typical functional
programming: not a variable, but a name (and possibly more than one
name) bound to an object.


Of course, if yo change the orgiginal named value, when there
would be no copy of it, before changing it, then both names
would refer to the changed data.
of course that is not, what is wanted.

But what you also can see in section 5.9.10 is, that
there already is a mechanism (reference counting) that allows
to distinguish between unnamed and named object.

So, this is directly adressing the points you have mentioned in your
examples.

So, at least in principial, R allows to do in-place modifications
of object with the .Call interface.

You seem to refer to the .C interface, and I had explored the .Call
interface. That's the reason why you may insist on "it's copyied
always" and I wondered, what you were talking about, because the
.Call interface allowed me rather C-like raw style of programming
(and the user of it to decide, if copying will be done or not).

The mechanism to descide, if copying should be done or not,
also is mentioined in section 5.9.10: NAMED and SET_NAMED macros.
with NAMED you can get the number of references.

But later in that section it is mentioned, that - at least for now -
NAMED always returns the value 2.


  "Currently all arguments to a .Call call will have NAMED set to 2,
  and so users must assume that they need to be duplicated before
  alteration."
               (section 5.9.10, last sentence)


So, the in-place modification can be done already with the .Call
intefcae for example. But the decision if it is safe or not
is not supported at the moment.

So the situation is somewhere between: "it is possible" and
"R does not support a safe decision if, what is possible, also
can be recommended".
At the moment R rather deprecates in-place modification by default
(the save way, and I agree with this default).

But it's not true, that R in general copies arguments.

But this seems to be true for the .C interface.

Maybe a lot of performance-/memory-problems can be solved
by rewriting already existing packages, by providing them
via .Call instead of .C.

My understanding is that most packages use the .C interface
because it's simpler to deal with and because they don't need
to pass complicated objects at the C level, just atomic vectors.
My guess is that it's probably rarely the case that the cost
of copying the arguments passed to .C is significant, but,
if that was the case, then they could always call .C() with
DUP=FALSE. However, using DUP=FALSE is dangerous (see Warning
section in the man page).

No need to switch to .Call

I strongly disagree. I'm appalled to see that sentence here.

Come on!

The overhead is significant for any large vector and it is in particular unnecessary since in .C you have to allocate *and copy* space even for results (twice!). Also it is very error-prone, because you have no information about the length of vectors so it's easy to run out of bounds and there is no way to check. IMHO .C should not be used for any code written in this century (the only exception may be if you are passing no data, e.g. if all you do is to pass a flag and expect no result, you can get away with it even if it is more dangerous). It is a legacy interface that dates way back and is essentially just re-named .Fortran interface. Again, I would strongly recommend the use of .Call in any recent code because it is safer and more efficient (if you don't care about either attribute, well, feel free ;)).

So aleph will not support the .C interface? ;-)

H.

Cheers,
Simon

Cheers,
H.

Ciao,
   Oliver




On Tue, Mar 06, 2012 at 04:44:49PM +0000, William Dunlap wrote:

S (and its derivatives and successors) promises that functions
will not change their arguments, so in an expression like
   val<- func(arg)
you know that arg will not be changed.  You can
do that by having func copy arg before doing anything,
but that uses space and time that you want to conserve.
If arg is not a named item in any environment then it
should be fine to write over the original because there
is no way the caller can detect that shortcut.  E.g., in
    cx<- cos(runif(n))
the cos function does not need to allocate new space for
its output, it can just write over its input because, without
a name attached to it, the caller has no way of looking
at what runif(n) returned.  If you did
    x<- runif(n)
    cx<- cos(x)
then cos would have to allocate new space for its output
because overwriting its input would affect a subsequent
    sum(x)
I suppose that end-users and function-writers could learn
to live with having to decide when to copy, but not having
to make that decision makes S more pleasant (and safer) to use.
I think that is a major reason that people are able to
share S code so easily.

Bill Dunlap
Spotfire, TIBCO Software
wdunlap tibco.com

-----Original Message-----
From: oliver [mailto:oliver at first.in-berlin.de]
Sent: Tuesday, March 06, 2012 1:12 AM
To: William Dunlap
Cc: Herv? Pag?s; R-devel
Subject: Re: [Rd] Julia

On Tue, Mar 06, 2012 at 12:35:32AM +0000, William Dunlap wrote:
[...]

I find R's (&   S+'s&   S's) copy-on-write-if-not-copying-would-be-discoverable-
by-the-uer machanism for giving the allusion of pass-by-value a good way
to structure the contract between the function writer and the function user.

[...]


Can you elaborate more on this,
especially on the ...-...-...-if-not-copying-would-be-discoverable-by-the-uer
stuff?

What do you mean with discoverability of not-copying?

Ciao,
   Oliver

--
Herv? Pag?s

Program in Computational Biology
Division of Public Health Sciences
Fred Hutchinson Cancer Research Center
1100 Fairview Ave. N, M1-B514
P.O. Box 19024
Seattle, WA 98109-1024

E-mail: hpages at fhcrc.org
Phone:  (206) 667-5791
Fax:    (206) 667-1319

-- 
Herv? Pag?s

Program in Computational Biology
Division of Public Health Sciences
Fred Hutchinson Cancer Research Center
1100 Fairview Ave. N, M1-B514
P.O. Box 19024
Seattle, WA 98109-1024

E-mail: hpages at fhcrc.org
Phone:  (206) 667-5791
Fax:    (206) 667-1319