R devel - Aug 2019 - Producing different text formats in R

Dear all,

I am facing a strange problem with text files formats.

I am currently using a C++ code named voro++ (http://math.lbl.gov/voro++/). This
code accepts text files with four columns:

<id> <x> <y> <z>

where id is an identification number and x,y,z are the coordinates of a point in
a 3D space.

The input file, name it myfile_cpp.txt, is generated by another C++ code written
by myself (named quadscheme.cpp). So far, these calculations involve no R code
and it works just fine.

However, now I am trying to replace my C++ code quadscheme.cpp by a R function.
The four columns (id,x,y,z) are produced in a matrix or Data Frame and then
saved in a file myfile_r.txt using write.table(). For example using the
following function:

quadscheme <- function(window, ntile) {
  # Creates a grid of points in a 3D orthohedron. 
  # First point lies at the (mix,miny,miz) corner of the volume and the last one
at (maxx,maxy,maxz)
  # window: vector. Contains the minimum and maximum values of a 3D orthohedron
  #       window <- c(minx, maxx, miny, maxy, minz, maxz)
  # ntile: vector. Number of points per dimension minus one. We manually add a
extra row of points per dimension
  #       ntile <- c(nstepx, nstepy, nstepz)
  M <- prod(ntile+1) 
  mat <- matrix(NA,M,4)
  mat[,1] <- 1:M # id column
  
  # step length per dimension
  hi <- (window[2] - window[1])/ntile[1]
  hj <- (window[4] - window[3])/ntile[2]
  hk <- (window[6] - window[5])/ntile[3]
  
  c <- 1
  for(i in 0:ntile[1]) {
    x <- i*hi + window[1]
    for(j in 0:ntile[2]) {
      y <- hj*j + window[3]
      for(k in 0:ntile[3]) {
        z <- k*hk + window[5]
        mat[c,2:4] <- c(x,y,z) # Add a new point to the grid
        c <- c + 1
      }
    }
  }
  write.table(mat, file="myfile_r.txt", row.names=FALSE,
col.names=FALSE)
}

And then calling:
> window <- c(18, 43, 171, 196, 95, 102)
> ntile <- c(100,100,28)
> quadscheme(window, ntile)
I see no difference between both myfile_*.txt files, one is created with C++ and
the other one with R. However, function voro++ do not accept the later one
(created with R and saved with write.table). I've also noted that voro++
usually accepts R generated files when they are small enough, but it accepts all
C++ generated files, regardless the size.

I know this is more a C++ than a R question, but may be you can help me as well.
I suspect that even if both files are *.txt they have some differences that I
cannot see. Is there any way in R to produce different kinds of txt files? Like
binary instead of text files? Could this be the problem?

I attach two identical files, one produced by C++ (myfile_cpp.txt) and another
one by the previous R function (myfile_r.txt). I attach as well the C++ code
used to generate myfile_cpp.txt.

Thank you in advance,

Llu?s Hurtado-Gil

-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: myfile_cpp.txt
URL:
<https://stat.ethz.ch/pipermail/r-devel/attachments/20190808/465ee6d9/attachment-0002.txt>

-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: myfile_r.txt
URL:
<https://stat.ethz.ch/pipermail/r-devel/attachments/20190808/465ee6d9/attachment-0003.txt>

>>>>> "Llu?s" ==   <Lluis.Hurtado at uv.es>
writes:
    Llu?s> Dear all,
    Llu?s> I am facing a strange problem with text files formats.

    Llu?s> I am currently using a C++ code named voro++
(http://math.lbl.gov/voro++/). This code accepts text files with four columns:

    Llu?s> <id> <x> <y> <z>

    Llu?s> where id is an identification number and x,y,z are the coordinates
of a point in a 3D space.

    Llu?s> The input file, name it myfile_cpp.txt, is generated by another
C++ code written by myself (named quadscheme.cpp). So far, these calculations
involve no R code and it works just fine.

    Llu?s> However, now I am trying to replace my C++ code quadscheme.cpp by
a R function. The four columns (id,x,y,z) are produced in a matrix or Data Frame
and then saved in a file myfile_r.txt using write.table(). For example using the
following function:

    Llu?s> quadscheme <- function(window, ntile) {
    Llu?s>   # Creates a grid of points in a 3D orthohedron. 
    Llu?s>   # First point lies at the (mix,miny,miz) corner of the volume
and the last one at (maxx,maxy,maxz)
    Llu?s>   # window: vector. Contains the minimum and maximum values of a
3D orthohedron
    Llu?s>   #       window <- c(minx, maxx, miny, maxy, minz, maxz)
    Llu?s>   # ntile: vector. Number of points per dimension minus one. We
manually add a extra row of points per dimension
    Llu?s>   #       ntile <- c(nstepx, nstepy, nstepz)
    Llu?s>   M <- prod(ntile+1) 
    Llu?s>   mat <- matrix(NA,M,4)
    Llu?s>   mat[,1] <- 1:M # id column
  
    Llu?s>   # step length per dimension
    Llu?s>   hi <- (window[2] - window[1])/ntile[1]
    Llu?s>   hj <- (window[4] - window[3])/ntile[2]
    Llu?s>   hk <- (window[6] - window[5])/ntile[3]
  
    Llu?s>   c <- 1
    Llu?s>   for(i in 0:ntile[1]) {
    Llu?s>     x <- i*hi + window[1]
    Llu?s>     for(j in 0:ntile[2]) {
    Llu?s>       y <- hj*j + window[3]
    Llu?s>       for(k in 0:ntile[3]) {
    Llu?s>         z <- k*hk + window[5]
    Llu?s>         mat[c,2:4] <- c(x,y,z) # Add a new point to the grid
    Llu?s>         c <- c + 1
    Llu?s>       }
    Llu?s>     }
    Llu?s>   }
    Llu?s>   write.table(mat, file="myfile_r.txt", row.names=FALSE,
col.names=FALSE)
    Llu?s> }

    Llu?s> And then calling:

    >> window <- c(18, 43, 171, 196, 95, 102)
    >> ntile <- c(100,100,28)
    >> quadscheme(window, ntile)

    Llu?s> I see no difference between both myfile_*.txt files,
    Llu?s> one is created with C++ and the other one with
    Llu?s> R. However, function voro++ do not accept the later one
    Llu?s> (created with R and saved with write.table). I've also
    Llu?s> noted that voro++ usually accepts R generated files
    Llu?s> when they are small enough, but it accepts all C++
    Llu?s> generated files, regardless the size.

    Llu?s> I know this is more a C++ than a R question, but may be
    Llu?s> you can help me as well. I suspect that even if both
    Llu?s> files are *.txt they have some differences that I
    Llu?s> cannot see. Is there any way in R to produce different
    Llu?s> kinds of txt files? Like binary instead of text files?
    Llu?s> Could this be the problem?

    Llu?s> I attach two identical files, one produced by C++
    Llu?s> (myfile_cpp.txt) and another one by the previous R
    Llu?s> function (myfile_r.txt). I attach as well the C++ code
    Llu?s> used to generate myfile_cpp.txt.

    Llu?s> Thank you in advance,

    Llu?s> Llu?s Hurtado-Gil

In your R file, scientific notation is used:

R:
  99999 26.5 174.5 96.5
  1e+05 26.5 174.5 96.75
  100001 26.5 174.5 97

cpp:
  99999 26.5 174.5 96.5
  100000 26.5 174.5 96.75
  100001 26.5 174.5 97

Try setting options(scipen = 20) before you write the
file in R.



-- 
Enrico Schumann
Lucerne, Switzerland
http://enricoschumann.net

Llu?s,

You sent a 8+ mb file to a (I presume) 1000+ subscribers of this list. That
is not generally a good idea, and I am surprised this got past the mailman
software which (usually) filters at 100k (and for a reason).

In the future, please consider uploading the file somewhere (GitHub gists are
easy) and send a link.  We also often talk about "minimally complete
verifiable examples" (mcve) and the minimal is there for a reason.

Thanks for your consideration,  Dirk

-- 
http://dirk.eddelbuettel.com | @eddelbuettel | edd at debian.org

Thank you for your fast answer. It just works fine now.

And apologies for attaching heavy files.

Llu?s.
> >>>>> "Llu?s" ==   <Lluis.Hurtado at uv.es>
writes:
> 
>     Llu?s> Dear all,
>     Llu?s> I am facing a strange problem with text files formats.
> 
>     Llu?s> I am currently using a C++ code named voro++
(http://mathlbl.gov/voro++/). This code accepts text files with four columns:
> 
>     Llu?s> <id> <x> <y> <z>
> 
>     Llu?s> where id is an identification number and x,y,z are the
coordinates of a point in a 3D space.
> 
>     Llu?s> The input file, name it myfile_cpp.txt, is generated by
another C++ code written by myself (named quadscheme.cpp). So far, these
calculations involve no R code and it works just fine.
> 
>     Llu?s> However, now I am trying to replace my C++ code
quadscheme.cpp by a R function. The four columns (id,x,y,z) are produced in a
matrix or Data Frame and then saved in a file myfile_r.txt using write.table().
For example using the following function:
> 
>     Llu?s> quadscheme <- function(window, ntile) {
>     Llu?s>   # Creates a grid of points in a 3D orthohedron. 
>     Llu?s>   # First point lies at the (mix,miny,miz) corner of the
volume and the last one at (maxx,maxy,maxz)
>     Llu?s>   # window: vector. Contains the minimum and maximum values
of a 3D orthohedron
>     Llu?s>   #       window <- c(minx, maxx, miny, maxy, minz, maxz)
>     Llu?s>   # ntile: vector. Number of points per dimension minus one.
We manually add a extra row of points per dimension
>     Llu?s>   #       ntile <- c(nstepx, nstepy, nstepz)
>     Llu?s>   M <- prod(ntile+1) 
>     Llu?s>   mat <- matrix(NA,M,4)
>     Llu?s>   mat[,1] <- 1:M # id column
>   
>     Llu?s>   # step length per dimension
>     Llu?s>   hi <- (window[2] - window[1])/ntile[1]
>     Llu?s>   hj <- (window[4] - window[3])/ntile[2]
>     Llu?s>   hk <- (window[6] - window[5])/ntile[3]
>   
>     Llu?s>   c <- 1
>     Llu?s>   for(i in 0:ntile[1]) {
>     Llu?s>     x <- i*hi + window[1]
>     Llu?s>     for(j in 0:ntile[2]) {
>     Llu?s>       y <- hj*j + window[3]
>     Llu?s>       for(k in 0:ntile[3]) {
>     Llu?s>         z <- k*hk + window[5]
>     Llu?s>         mat[c,2:4] <- c(x,y,z) # Add a new point to the
grid
>     Llu?s>         c <- c + 1
>     Llu?s>       }
>     Llu?s>     }
>     Llu?s>   }
>     Llu?s>   write.table(mat, file="myfile_r.txt",
row.names=FALSE, col.names=FALSE)
>     Llu?s> }
> 
>     Llu?s> And then calling:
> 
>     >> window <- c(18, 43, 171, 196, 95, 102)
>     >> ntile <- c(100,100,28)
>     >> quadscheme(window, ntile)
> 
>     Llu?s> I see no difference between both myfile_*.txt files,
>     Llu?s> one is created with C++ and the other one with
>     Llu?s> R. However, function voro++ do not accept the later one
>     Llu?s> (created with R and saved with write.table). I've also
>     Llu?s> noted that voro++ usually accepts R generated files
>     Llu?s> when they are small enough, but it accepts all C++
>     Llu?s> generated files, regardless the size.
> 
>     Llu?s> I know this is more a C++ than a R question, but may be
>     Llu?s> you can help me as well. I suspect that even if both
>     Llu?s> files are *.txt they have some differences that I
>     Llu?s> cannot see. Is there any way in R to produce different
>     Llu?s> kinds of txt files? Like binary instead of text files?
>     Llu?s> Could this be the problem?
> 
>     Llu?s> I attach two identical files, one produced by C++
>     Llu?s> (myfile_cpp.txt) and another one by the previous R
>     Llu?s> function (myfile_r.txt). I attach as well the C++ code
>     Llu?s> used to generate myfile_cpp.txt.
> 
>     Llu?s> Thank you in advance,
> 
>     Llu?s> Llu?s Hurtado-Gil
> 
> In your R file, scientific notation is used:
> 
> R:
>   99999 26.5 174.5 96.5
>   1e+05 26.5 174.5 96.75
>   100001 26.5 174.5 97
> 
> cpp:
>   99999 26.5 174.5 96.5
>   100000 26.5 174.5 96.75
>   100001 26.5 174.5 97
> 
> Try setting options(scipen = 20) before you write the
> file in R.
> 
> 
> 
> -- 
> Enrico Schumann
> Lucerne, Switzerland
> http://enricoschumann.net
>