Re: [Rd] Print bug for matrix(list(NA_complex_, ...))
> William Dunlap > on Wed, 3 Jun 2009 11:49:00 -0700 writes: > Changing the uninitialized 'w' to 'R_print.na_width' for > the complex NA code fixes up the valgrind complaints on > Linux and the bad behavior on Windows: [.] Thanks a lot, Bill, both for the diagnosis and the fix! I'll commit (a slight variation of) it to R-devel and R-patched. Martin __ R-devel@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-devel
Re: [Rd] Print bug for matrix(list(NA_complex_, ...))
Quick action, thanks! Does this also fix the spacing problem (at the end of
my bug report)?
-s
On Wed, Jun 3, 2009 at 2:49 PM, William Dunlap wrote:
> Changing the uninitialized 'w' to 'R_print.na_width' for the complex NA
> code fixes up the valgrind complaints on Linux and the bad behavior
> on Windows:
>
> ===
> --- print.c (revision 48703)
> +++ print.c (working copy)
> @@ -357,7 +357,7 @@
> formatComplex(x, 1, &wr, &dr, &er, &wi, &di, &ei,
> 0);
> if (ISNA(x[0].r) || ISNA(x[0].i))
> snprintf(pbuf, 115, "%s",
> -EncodeReal(NA_REAL, w, 0, 0, OutDec));
> +EncodeReal(NA_REAL, R_print.na_width,
> 0, 0, OutDec)); /* formatReal would set w=R_print.na_width,d=0,e=0 for
> NA */
>else
> snprintf(pbuf, 115, "%s", EncodeComplex(x[0],
> wr, dr, er, wi, di, ei, OutDec));
>
> E.g., the following prints go very quickly and don't have lots of
> trailing
> blanks.
>
> > val <- matrix(list(NA_complex_, 3, "A string", NA_complex_), 2,2)
> > print(val)
> [,1] [,2]
> [1,] NA "A string"
> [2,] 3NA
> > print(val, na.print="")
> [,1]
> [1,]
> [2,] 3
> [,2]
> [1,] "A string"
> [2,]
>
> Bill Dunlap
> TIBCO Software Inc - Spotfire Division
> wdunlap tibco.com
>
> > -Original Message-----
> > From: r-devel-boun...@r-project.org
> > [mailto:r-devel-boun...@r-project.org] On Behalf Of William Dunlap
> > Sent: Wednesday, June 03, 2009 9:21 AM
> > To: Martin Maechler; Wacek Kusnierczyk
> > Cc: r-devel@r-project.org
> > Subject: Re: [Rd] Print bug for matrix(list(NA_complex_, ...))
> >
> >
> > > From: William Dunlap
> > > Sent: Wednesday, June 03, 2009 8:58 AM
> > > To: 'Martin Maechler'; Wacek Kusnierczyk
> > > Cc: r-devel@r-project.org
> > > Subject: RE: [Rd] Print bug for matrix(list(NA_complex_, ...))
> > ...
> > > > >> There is a bug in printing val <-
> > > > matrix(list(NA_complex_,NA_complex_),1).
> > ...
> > > > >> Note that a large number of spaces are printed
> > > instead of NA.
> > > >
> > > > vQ> on ubuntu 8.04 with r 2.10.0 r48703 there is almost
> > > > no problem (still
> > > > vQ> some unnecessary spaces):
> > > >
> > > > vQ> [,1] [,2]
> > > > vQ> [1,]NANA
> > > >
> > > > thank you, Waclav.
> > > >
> > > > I don't see any unnecessary space (or other CPU problems),
> > > > using another Linux variant (RHEL 5)
> > > > in the three R version
> > > > 2.9.0 ,
> > > > 2.9.0-patched (r48700)
> > > > 2.10.0-devel (r48700)
> > > >
> > > > Can other Windows (or ...) users confirm problems in R 2.9.0
> > > > (or newer) ?
> > >
> > > I just tried it on 2.9.0 on the Windows GUI (RGui.exe)
> > > and saw similar problems. When printing to the
> > > command window, with either print(val) or show(val) it took
> > > a very long time and put $'s at the ends of the lines
> > > (indicating that there was text beyond the edge of the window).
> > > I used sink() to put the output into a file and then print() and
> > > show() acted quickly. The resulting file had 119 character
> > > lines, consisting mainly of trailing blanks:
> > >
> > > > readLines("c:/temp/val.txt") # output of
> > > print(val);show(val);str(val)
> > > [1] " [,1]
> > >"
> > > [2] "[1,]
> > >"
> > > [3] " [,2]
> > >"
> > > [4] "[1,]
> > >"
> > > [5] " [,1]
> > >"
> > > [6] "[1,]
> > >"
> > > [7] " [,2]
> > >
Re: [Rd] Print bug for matrix(list(NA_complex_, ...))
Changing the uninitialized 'w' to 'R_print.na_width' for the complex NA
code fixes up the valgrind complaints on Linux and the bad behavior
on Windows:
===
--- print.c (revision 48703)
+++ print.c (working copy)
@@ -357,7 +357,7 @@
formatComplex(x, 1, &wr, &dr, &er, &wi, &di, &ei,
0);
if (ISNA(x[0].r) || ISNA(x[0].i))
snprintf(pbuf, 115, "%s",
-EncodeReal(NA_REAL, w, 0, 0, OutDec));
+EncodeReal(NA_REAL, R_print.na_width,
0, 0, OutDec)); /* formatReal would set w=R_print.na_width,d=0,e=0 for
NA */
else
snprintf(pbuf, 115, "%s", EncodeComplex(x[0],
wr, dr, er, wi, di, ei, OutDec));
E.g., the following prints go very quickly and don't have lots of
trailing
blanks.
> val <- matrix(list(NA_complex_, 3, "A string", NA_complex_), 2,2)
> print(val)
[,1] [,2]
[1,] NA "A string"
[2,] 3NA
> print(val, na.print="")
[,1]
[1,]
[2,] 3
[,2]
[1,] "A string"
[2,]
Bill Dunlap
TIBCO Software Inc - Spotfire Division
wdunlap tibco.com
> -Original Message-
> From: r-devel-boun...@r-project.org
> [mailto:r-devel-boun...@r-project.org] On Behalf Of William Dunlap
> Sent: Wednesday, June 03, 2009 9:21 AM
> To: Martin Maechler; Wacek Kusnierczyk
> Cc: r-devel@r-project.org
> Subject: Re: [Rd] Print bug for matrix(list(NA_complex_, ...))
>
>
> > From: William Dunlap
> > Sent: Wednesday, June 03, 2009 8:58 AM
> > To: 'Martin Maechler'; Wacek Kusnierczyk
> > Cc: r-devel@r-project.org
> > Subject: RE: [Rd] Print bug for matrix(list(NA_complex_, ...))
> ...
> > > >> There is a bug in printing val <-
> > > matrix(list(NA_complex_,NA_complex_),1).
> ...
> > > >> Note that a large number of spaces are printed
> > instead of NA.
> > >
> > > vQ> on ubuntu 8.04 with r 2.10.0 r48703 there is almost
> > > no problem (still
> > > vQ> some unnecessary spaces):
> > >
> > > vQ> [,1] [,2]
> > > vQ> [1,]NANA
> > >
> > > thank you, Waclav.
> > >
> > > I don't see any unnecessary space (or other CPU problems),
> > > using another Linux variant (RHEL 5)
> > > in the three R version
> > > 2.9.0 ,
> > > 2.9.0-patched (r48700)
> > > 2.10.0-devel (r48700)
> > >
> > > Can other Windows (or ...) users confirm problems in R 2.9.0
> > > (or newer) ?
> >
> > I just tried it on 2.9.0 on the Windows GUI (RGui.exe)
> > and saw similar problems. When printing to the
> > command window, with either print(val) or show(val) it took
> > a very long time and put $'s at the ends of the lines
> > (indicating that there was text beyond the edge of the window).
> > I used sink() to put the output into a file and then print() and
> > show() acted quickly. The resulting file had 119 character
> > lines, consisting mainly of trailing blanks:
> >
> > > readLines("c:/temp/val.txt") # output of
> > print(val);show(val);str(val)
> > [1] " [,1]
> >"
> > [2] "[1,]
> >"
> > [3] " [,2]
> >"
> > [4] "[1,]
> >"
> > [5] " [,1]
> >"
> > [6] "[1,]
> >"
> > [7] " [,2]
> >"
> > [8] "[1,]
> >
Re: [Rd] Print bug for matrix(list(NA_complex_, ...))
> From: William Dunlap
> Sent: Wednesday, June 03, 2009 8:58 AM
> To: 'Martin Maechler'; Wacek Kusnierczyk
> Cc: r-devel@r-project.org
> Subject: RE: [Rd] Print bug for matrix(list(NA_complex_, ...))
...
> > >> There is a bug in printing val <-
> > matrix(list(NA_complex_,NA_complex_),1).
...
> > >> Note that a large number of spaces are printed
> instead of NA.
> >
> > vQ> on ubuntu 8.04 with r 2.10.0 r48703 there is almost
> > no problem (still
> > vQ> some unnecessary spaces):
> >
> > vQ> [,1] [,2]
> > vQ> [1,]NANA
> >
> > thank you, Waclav.
> >
> > I don't see any unnecessary space (or other CPU problems),
> > using another Linux variant (RHEL 5)
> > in the three R version
> > 2.9.0 ,
> > 2.9.0-patched (r48700)
> > 2.10.0-devel (r48700)
> >
> > Can other Windows (or ...) users confirm problems in R 2.9.0
> > (or newer) ?
>
> I just tried it on 2.9.0 on the Windows GUI (RGui.exe)
> and saw similar problems. When printing to the
> command window, with either print(val) or show(val) it took
> a very long time and put $'s at the ends of the lines
> (indicating that there was text beyond the edge of the window).
> I used sink() to put the output into a file and then print() and
> show() acted quickly. The resulting file had 119 character
> lines, consisting mainly of trailing blanks:
>
> > readLines("c:/temp/val.txt") # output of
> print(val);show(val);str(val)
> [1] " [,1]
>"
> [2] "[1,]
>"
> [3] " [,2]
>"
> [4] "[1,]
>"
> [5] " [,1]
>"
> [6] "[1,]
>"
> [7] " [,2]
>"
> [8] "[1,]
>"
> [9] "List of 2"
>
> [10] " $ : cplx NA"
>
> [11] " $ : cplx NA"
>
> [12] " - attr(*, \"dim\")= int [1:2] 1 2"
>
> > nchar(readLines("c:/temp/val.txt"))
> [1] 119 119 119 119 119 119 119 119 9 12 12 32
>
> On Linux, using yesterday's 2.9.0, valgrind shows lots
> of 'depends on uninitialized value' problems that could
> explain the Linux-Windows difference in behavior:
>
> > val<-matrix(list(NA_complex_,NA_complex_),1)
> > val
> ==24859== Conditional jump or move depends on uninitialised value(s)
> ==24859==at 0x6F7E2F: vfprintf (in /lib/tls/libc-2.3.4.so)
> ==24859==by 0x7177A5: vsnprintf (in /lib/tls/libc-2.3.4.so)
> ==24859==by 0x6FF4E1: snprintf (in /lib/tls/libc-2.3.4.so)
> ==24859==by 0x809DDC1: Rf_EncodeReal (printutils.c:179)
> ==24859==by 0x8099AA9: Rf_PrintValueRec (print.c:359)
> ==24859==by 0x809A124: Rf_PrintValueEnv (print.c:956)
> ==24859==by 0x80584CA: Rf_ReplIteration (main.c:261)
> ==24859==by 0x8058572: R_ReplConsole (main.c:306)
> ==24859==by 0x8058BB8: run_Rmainloop (main.c:967)
> ==24859==by 0x8056A70: main (Rmain.c:33)
Here is the traceback when it enters Rf_EncodeReal:
(gdb) where
#0 Rf_EncodeReal (x=nan(0x007a2), w=168381312, d=0, e=0, cdec=46
'.')
at printutils.c:160
#1 0x08099aaa in Rf_PrintValueRec (s=0x9f4a4a8, env=0x9e4cc90) at
print.c:359
#2 0x0809a125 in Rf_PrintValueEnv (s=0x9f4a4a8, env=0x9e4cc90) at
print.c:956
#3 0x080584cb in Rf_ReplIteration (rho=0x9e4cc90, savestack=0,
b
Re: [Rd] Print bug for matrix(list(NA_complex_, ...))
> -Original Message-
> From: r-devel-boun...@r-project.org
> [mailto:r-devel-boun...@r-project.org] On Behalf Of Martin Maechler
> Sent: Wednesday, June 03, 2009 2:37 AM
> To: Wacek Kusnierczyk
> Cc: r-devel@r-project.org
> Subject: Re: [Rd] Print bug for matrix(list(NA_complex_, ...))
>
> >>>>> "vQ" == Wacek Kusnierczyk
>
> >>>>> on Wed, 03 Jun 2009 10:43:25 +0200 writes:
>
> vQ> Stavros Macrakis wrote:
> >> In R 2.8.0 on Windows (tested both under ESS and under
> R Console in case
> >> there was an I/O issue)
> >>
> >> There is a bug in printing val <-
> matrix(list(NA_complex_,NA_complex_),1).
> >>
> >>
> >>> dput(val)
> >>>
> >> structure(list(NA_complex_, NA_complex_), .Dim = 1:2)
> >>
> >>
> >>> print(val)
> >>>
> >>
> >> [,1]
> >>
> >> [1,]
> >>
> >>
> >> [,2]
> >>
> >> [1,]
> >>
> >>
> >> Note that a large number of spaces are printed instead of NA.
>
> vQ> on ubuntu 8.04 with r 2.10.0 r48703 there is almost
> no problem (still
> vQ> some unnecessary spaces):
>
> vQ> [,1] [,2]
> vQ> [1,]NANA
>
> thank you, Waclav.
>
> I don't see any unnecessary space (or other CPU problems),
> using another Linux variant (RHEL 5)
> in the three R version
> 2.9.0 ,
> 2.9.0-patched (r48700)
> 2.10.0-devel (r48700)
>
> Can other Windows (or ...) users confirm problems in R 2.9.0
> (or newer) ?
I just tried it on 2.9.0 on the Windows GUI (RGui.exe)
and saw similar problems. When printing to the
command window, with either print(val) or show(val) it took
a very long time and put $'s at the ends of the lines
(indicating that there was text beyond the edge of the window).
I used sink() to put the output into a file and then print() and
show() acted quickly. The resulting file had 119 character
lines, consisting mainly of trailing blanks:
> readLines("c:/temp/val.txt") # output of print(val);show(val);str(val)
[1] " [,1]
"
[2] "[1,]
"
[3] " [,2]
"
[4] "[1,]
"
[5] " [,1]
"
[6] "[1,]
"
[7] " [,2]
"
[8] "[1,]
"
[9] "List of 2"
[10] " $ : cplx NA"
[11] " $ : cplx NA"
[12] " - attr(*, \"dim\")= int [1:2] 1 2"
> nchar(readLines("c:/temp/val.txt"))
[1] 119 119 119 119 119 119 119 119 9 12 12 32
On Linux, using yesterday's 2.9.0, valgrind shows lots
of 'depends on uninitialized value' problems that could
explain the Linux-Windows difference in behavior:
> val<-matrix(list(NA_complex_,NA_complex_),1)
> val
==24859== Conditional jump or move depends on uninitialised value(s)
==24859==at 0x6F7E2F: vfprintf (in /lib/tls/libc-2.3.4.so)
==24859==by 0x7177A5: vsnprintf (in /lib/tls/libc-2.3.4.so)
==24859==by 0x6FF4E1: snprintf (in /lib/tls/libc-2.3.4.so)
==24859==by 0x809DDC1: Rf_EncodeReal (printutils.c:179)
==24859==by 0x8099AA9: Rf_PrintValueRec (print.c:359)
==24859==by 0x809A124: Rf_PrintValueEnv (print.c:956)
==24859==by 0x80584CA: Rf_ReplIteration (main.c:261)
==24859==by 0x8058572: R_ReplConsole (main.c:306)
==24859==by 0x8058BB8: run_Rmainloop (main.c:967)
==24859==by 0x8056A70: main (Rmain.c:33)
==24859==
==24859== Conditional jump or move depends on uninitialised value(s)
==24859==at 0x6F7E44: vfprintf (in /lib/tls/libc-2.3.4.so)
==24859==by 0x7177A5: vsnprintf (in /lib/tls/libc-2.3.4.so)
==24859==by 0x6FF4E1: snprintf (in /lib/tls/libc-2.3.4.so)
==24859==by 0x809DDC1: Rf_EncodeReal (printutils.c:179)
==24859==by 0x8099AA9: Rf_PrintValueRec (print.c:359)
==24859==by 0x809A124: Rf_PrintValueEnv (print.c:956)
==24859==by 0x80584CA: Rf_ReplIteration (main.c:261)
==24859==by 0x8058572: R_ReplConsole (main.c:306)
==24859==by 0x8058BB8: run_Rmainloop (main.c:967)
==24859==by 0x8056A70: main (Rmain.c:33)
==24859==
==24859== Conditional jump or move depends on uninitialised value(s)
==24859==at 0x6F7E4C: vfprintf (in /lib/tls/libc-2.3.4.so)
==24859==by 0x7177A5: vsnprintf (in /lib/tls/libc-2.3.4.so)
==24859==by 0x6FF4E1: snprintf (in /lib/tls/libc-2.3.4.so)
==24859==by 0x809DDC1: Rf_EncodeReal (printutils.c:179)
==24859==by 0x8099AA9: Rf_PrintValueRec (print.c:359)
==24859==by 0x809A124: Rf_PrintValueEnv (print.c:956)
==24859==by 0x80584CA: Rf_ReplIteration (main.c:261)
==24859==by 0x
Re: [Rd] Print bug for matrix(list(NA_complex_, ...))
> "vQ" == Wacek Kusnierczyk > on Wed, 03 Jun 2009 10:43:25 +0200 writes: vQ> Stavros Macrakis wrote: >> In R 2.8.0 on Windows (tested both under ESS and under R Console in case >> there was an I/O issue) >> >> There is a bug in printing val <- matrix(list(NA_complex_,NA_complex_),1). >> >> >>> dput(val) >>> >> structure(list(NA_complex_, NA_complex_), .Dim = 1:2) >> >> >>> print(val) >>> >> >> [,1] >> >> [1,] >> >> >> [,2] >> >> [1,] >> >> >> Note that a large number of spaces are printed instead of NA. vQ> on ubuntu 8.04 with r 2.10.0 r48703 there is almost no problem (still vQ> some unnecessary spaces): vQ> [,1] [,2] vQ> [1,]NANA thank you, Waclav. I don't see any unnecessary space (or other CPU problems), using another Linux variant (RHEL 5) in the three R version 2.9.0 , 2.9.0-patched (r48700) 2.10.0-devel (r48700) Can other Windows (or ...) users confirm problems in R 2.9.0 (or newer) ? Regards, Martin Maechler >> Compare the >> unproblematic real case: >> >> print(structure(list(NA_real_, NA_real_), .Dim = 1:2)) >> [,1] [,2] >> [1,] NA NA >> >> Also, when printed in the read-eval-print loop, printing takes a very very >> long time: >> >> >>> proc.time(); matrix(list(NA_complex_,NA_complex_),1); proc.time() >>> >> user system elapsed >> 74.350.09 329.45 >> >> [,1] >> >> [1,] >> >> >> [,2] >> >> [1,] >> >> user system elapsed >> 92.630.15 347.86 >> >> 18 seconds runtime! >> vQ> user system elapsed vQ> 0.648 0.056 155.843 vQ> [,1] [,2] vQ> [1,] NA NA vQ> user system elapsed vQ> 0.648 0.056 155.843 vQ> vQ vQ> __ vQ> R-devel@r-project.org mailing list vQ> https://stat.ethz.ch/mailman/listinfo/r-devel __ R-devel@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-devel
Re: [Rd] Print bug for matrix(list(NA_complex_, ...))
Stavros Macrakis wrote: > In R 2.8.0 on Windows (tested both under ESS and under R Console in case > there was an I/O issue) > > There is a bug in printing val <- matrix(list(NA_complex_,NA_complex_),1). > > >> dput(val) >> > structure(list(NA_complex_, NA_complex_), .Dim = 1:2) > > >> print(val) >> > > [,1] > > [1,] > > > [,2] > > [1,] > > > Note that a large number of spaces are printed instead of NA. on ubuntu 8.04 with r 2.10.0 r48703 there is almost no problem (still some unnecessary spaces): [,1] [,2] [1,]NANA > Compare the > unproblematic real case: > > print(structure(list(NA_real_, NA_real_), .Dim = 1:2)) > [,1] [,2] > [1,] NA NA > > Also, when printed in the read-eval-print loop, printing takes a very very > long time: > > >> proc.time(); matrix(list(NA_complex_,NA_complex_),1); proc.time() >> >user system elapsed > 74.350.09 329.45 > > [,1] > > [1,] > > > [,2] > > [1,] > >user system elapsed > 92.630.15 347.86 > > 18 seconds runtime! > user system elapsed 0.648 0.056 155.843 [,1] [,2] [1,] NA NA user system elapsed 0.648 0.056 155.843 vQ __ R-devel@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-devel
[Rd] Print bug for matrix(list(NA_complex_, ...))
In R 2.8.0 on Windows (tested both under ESS and under R Console in case there was an I/O issue) There is a bug in printing val <- matrix(list(NA_complex_,NA_complex_),1). > dput(val) structure(list(NA_complex_, NA_complex_), .Dim = 1:2) > print(val) [,1] [1,] [,2] [1,] Note that a large number of spaces are printed instead of NA. Compare the unproblematic real case: print(structure(list(NA_real_, NA_real_), .Dim = 1:2)) [,1] [,2] [1,] NA NA Also, when printed in the read-eval-print loop, printing takes a very very long time: > proc.time(); matrix(list(NA_complex_,NA_complex_),1); proc.time() user system elapsed 74.350.09 329.45 [,1] [1,] [,2] [1,] user system elapsed 92.630.15 347.86 18 seconds runtime! This does not happen with an explicit print: > proc.time(); print(structure(list(NA_complex_,NA_complex_), .Dim = 1:2)); proc.time() user system elapsed 92.630.15 355.90 [,1] [1,] [,2] [1,] user system elapsed 92.770.15 356.05 -- A perhaps related problem is that the NA's are not properly aligned in some cases: > structure(list(NaN, NA_complex_, NA_real_, NA_complex_, NA_real_, NA_complex_, NA_real_, NA_complex_, NA_complex_, NA_complex_, NaN, NA_complex_, NA_complex_, NA_complex_), .Dim = c(2L, 7L)) [,1] [,2] [,3] [,4] [,5] [,6] [,7] [1,] NaN NA NA NA NA NaN NA [2,] NA NA NA NA NANA NA It is perhaps hard to see this in email, but there is an extra space in front of the NAs at [1,7], [2,1], and [2,6]. Please let me know if you need additional information to troubleshoot this problem. -s [[alternative HTML version deleted]] __ R-devel@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-devel
