Re: [Rd] clusterApply arguments

[Tomas Kalibera]

Fixed in R-devel (74418).
Tomas

On 03/15/2018 08:57 PM, Tomas Kalibera wrote:

On 03/15/2018 05:25 PM, Henrik Bengtsson wrote:

On Thu, Mar 15, 2018 at 3:39 AM, <florianschwendin...@gmx.at> wrote:

Thank you for your answer!
I agree with you except for the 3 (Error) example and
I realize now I should have started with that in the explanation.

 From my point of view
parLapply(cl = clu, X = 1:2, fun = fun, c = 1)
shouldn't give an error.

This could be easily avoided by using all the argument
names in the custerApply call of parLapply which means changing,

parLapply <- function(cl = NULL, X, fun, ...)  {
     cl <- defaultCluster(cl)
     do.call(c, clusterApply(cl, x = splitList(X, length(cl)),
             fun = lapply, fun, ...), quote = TRUE)
}

to

parLapply <- function (cl = NULL, X, fun, ...)  {
     cl <- defaultCluster(cl)
     do.call(c, clusterApply(cl = cl, x = splitList(X, length(cl)),
             fun = lapply, fun, ...), quote = TRUE)
}

Oh... sorry I missed that point.  Yes, I agree, this should be a
trivial fix to the 'parallel' package.

/Henrik

Yes, thanks for the report, I am testing a fix for this (and other missing argument names in calls involving ...) in parallel.

Tomas

.

Best regards,
Florian



Gesendet: Mittwoch, 14. März 2018 um 19:05 Uhr
Von: "Henrik Bengtsson" <henrik.bengts...@gmail.com>
An: "Florian Schwendinger" <florian_schwendin...@gmx.at>
Cc: fschw...@wu.ac.at, R-devel <r-devel@r-project.org>
Betreff: Re: [Rd] clusterApply arguments
This is nothing specific to parallel::clusterApply() per se. It is the
default behavior of R where it allows for partial argument names. I
don't think there's much that can be done here except always using
fully named arguments to the "apply" function itself as you show.

You can "alert" yourself when there's a mistake by using:

options(warnPartialMatchArgs = TRUE)

e.g.

clusterApply(clu, x = 1:2, fun = fun, c = 1) ## Error

Warning in clusterApply(clu, x = 1:2, fun = fun, c = 1) :
partial argument match of 'c' to 'cl'
Error in checkCluster(cl) : not a valid cluster

It's still only a warning, but an informative one.

/Henrik

On Wed, Mar 14, 2018 at 8:50 AM, Florian Schwendinger
<florian_schwendin...@gmx.at> wrote:

Hi!

I recognized that the argument matching of clusterApply (and therefore parLapply) goes wrong when one of the arguments of the function is called "c". In this case, the argument "c" is used as cluster and the functions give the following error message "Error in checkCluster(cl) : not a valid cluster".

Of course, "c" is for many reasons an unfortunate argument name and this can be easily fixed by the user side.

See below for a small example.

library(parallel)

clu <- makeCluster(2, "PSOCK")

fun <- function(x0, x1) (x0 + x1)
clusterApply(clu, x = 1:2, fun = fun, x1 = 1) ## OK
parLapply(cl = clu, X = 1:2, fun = fun, x1 = 1) #OK


fun <- function(b, c) (b + c)
clusterApply(clu, x = 1:2, fun = fun, c = 1) ## Error
clusterApply(cl = clu, x = 1:2, fun = fun, c = 1) ## OK
parLapply(cl = clu, X = 1:2, fun = fun, c = 1) ## Error

stopCluster(clu)


I used "R version 3.4.3 Patched (2018-01-07 r74099".


Best regards,
Florian

______________________________________________
R-devel@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-devel

______________________________________________
R-devel@r-project.org mailing list

https://stat.ethz.ch/mailman/listinfo/r-devel[https://stat.ethz.ch/mailman/listinfo/r-devel]

______________________________________________
R-devel@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-devel

______________________________________________
R-devel@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-devel