Skip to content

Split

13 messages · Rui Barradas, Val, Bill Dunlap +2 more

Val
# 
HI All,

I am trying to create   new columns based on another column string
content. First I want to identify rows that contain a particular
string.  If it contains, I want to split the string and create two
variables.

Here is my sample of data.
F1<-read.table(text="ID1  ID2  text
A1 B1   NONE
A1 B1   cf_12
A1 B1   NONE
A2 B2   X2_25
A2 B3   fd_15  ",header=TRUE,stringsAsFactors=F)
If the variable "text" contains this "_" I want to create an indicator
variable as shown below

F1$Y1 <- ifelse(grepl("_", F1$text),1,0)


Then I want to split that string in to two, before "_" and after "_"
and create two variables as shown below
x1= strsplit(as.character(F1$text),'_',2)

My problem is how to combine this with the original data frame. The
desired  output is shown   below,


ID1 ID2  Y1   X1    X2
A1  B1    0   NONE   .
A1  B1   1    cf        12
A1  B1   0  NONE   .
A2  B2   1    X2    25
A2  B3   1    fd    15

Any help?
Thank you.
Rui Barradas
# 
Hello,

Something like this?


F1$Y1 <- +grepl("_", F1$text)
F1 <- F1[c(1, 2, 4, 3)]
F1 <- tidyr::separate(F1, text, into = c("X1", "X2"), sep = "_", fill = 
"right")
F1


Hope this helps,

Rui Barradas

?s 19:55 de 22/09/20, Val escreveu:
Rui Barradas
# 
Hello,

A base R solution with strsplit, like in your code.

F1$Y1 <- +grepl("_", F1$text)

tmp <- strsplit(as.character(F1$text), "_")
tmp <- lapply(tmp, function(x) if(length(x) == 1) c(x, ".") else x)
tmp <- do.call(rbind, tmp)
colnames(tmp) <- c("X1", "X2")
F1 <- cbind(F1[-3], tmp)    # remove the original column
rm(tmp)

F1
#  ID1 ID2 Y1   X1 X2
#1  A1  B1  0 NONE  .
#2  A1  B1  1   cf 12
#3  A1  B1  0 NONE  .
#4  A2  B2  1   X2 25
#5  A2  B3  1   fd 15


Note that cbind dispatches on F1, an object of class "data.frame".
Therefore it's the method cbind.data.frame that is called and the result 
is also a df, though tmp is a "matrix".


Hope this helps,

Rui Barradas


?s 20:07 de 22/09/20, Rui Barradas escreveu:
LMH
# 
Sometimes it just makes more sense to pre-process your data and get it into the format you need. It
just depends on whether you are more comfortable programing in R or in some other text manipulation
language like bash/sed/awk/grep etc.

If you know how to do this with other tools, you could write a script and probably call the script
from R. I could post a sample if you are interested.

LMH
Val wrote:
Bert Gunter
# 
To be clear, I think Rui's solution is perfectly fine and probably better
than what I offer below. But just for fun, I wanted to do it without the
lapply().  Here is one way. I think my comments suffice to explain.
ID1 ID2   text    1  2
1  A1  B1 NONE_. NONE  .
2  A1  B1  cf_12   cf 12
3  A1  B1 NONE_. NONE  .
4  A2  B2  X2_25   X2 25
5  A2  B3  fd_15   fd 15
Cheers,
Bert

Bert Gunter

"The trouble with having an open mind is that people keep coming along and
sticking things into it."
-- Opus (aka Berkeley Breathed in his "Bloom County" comic strip )
On Tue, Sep 22, 2020 at 12:19 PM Rui Barradas <ruipbarradas at sapo.pt> wrote:

            

      
      
      
    

        

      
      
    

            

      
      
      
    

        

      
      
    

            

      
      
      
    

        

      
      
    

        

  
  


  


      

    

    

      
      

        


  

        

      Val
    

    

      
      #
    

  


  

      
Thank you all for the help!

LMH, Yes I would like to see the alternative.  I am using this for a
large data set and if the  alternative is more efficient than this
then I would be happy.

        
On Tue, Sep 22, 2020 at 6:25 PM Bert Gunter <bgunter.4567 at gmail.com> wrote:

            

      
      
      
    

        

      
      
    

            

      
      
      
    

        

      
      
    

            

      
      
      
    

        

      
      
    

  
  


  


      

    

    

      
      

        


  

        

      Bert Gunter
    

    

      
      #
    

  


  

      
Oh, if efficiency is a consideration, then my code is about 15 times as
fast as Rui's:

            

      
      
      
    

        
##Rui's

            

      
      
      
    

        
+     F2$Y1 <- +grepl("_", F2$text)
+     tmp <- strsplit(as.character(F2$text), "_")
+     tmp <- lapply(tmp, function(x) if(length(x) == 1) c(x, ".") else x)
+     tmp <- do.call(rbind, tmp)
+     colnames(tmp) <- c("X1", "X2")
+     F2 <- cbind(F2[-3], tmp)    # remove the original column
+ })
   user  system elapsed
 20.072   0.625  20.786

## my version

            

      
      
      
    

        
+     wh <- grep("_",F2$text, fixed = TRUE, invert = TRUE)
+     F2[wh,"text"] <- paste(F2[wh,"text"],".",sep = "_")
+     z <- unlist(strsplit(F1$text,"_"))
+     F2 <- cbind(F2, matrix(z, ncol = 2, byrow = TRUE))
+     F2
+ })
   user  system elapsed
  1.256   0.019   1.281

Cheers,
Bert

Bert Gunter

"The trouble with having an open mind is that people keep coming along and
sticking things into it."
-- Opus (aka Berkeley Breathed in his "Bloom County" comic strip )

        
On Tue, Sep 22, 2020 at 5:04 PM Val <valkremk at gmail.com> wrote:

        

            

      
      
      
    

        

      
      
    

            

      
      
      
    

        

      
      
    

            

      
      
      
    

        

      
      
    

        

      
      
    

        

  
  


  


      

    

    

      
      

        


  

        

      Bill Dunlap
    

    

      
      #
    

  


  

      
Another way to make columns out of the stuff before and after the
underscore, with NAs if there is no underscore, is

utils::strcapture("([^_]*)_(.*)", F1$text,
proto=data.frame(Before_=character(), After_=character()))

-Bill

        
On Tue, Sep 22, 2020 at 4:25 PM Bert Gunter <bgunter.4567 at gmail.com> wrote:

        

            

      
      
      
    

        

      
      
    

            

      
      
      
    

        

      
      
    

            

      
      
      
    

        

      
      
    

            

      
      
      
    

        

      
      
    

        

  
  


  


      

    

    

      
      

        


  

        

      Bert Gunter
    

    

      
      #
    

  


  

      
That was still slower and doesn't quite give what was requested:

            

      
      
      
    

        
proto=data.frame(Before_=character(), After_=character())))
  ID1 ID2  text Before_ After_
1  A1  B1  NONE    <NA>   <NA>
2  A1  B1 cf_12      cf     12
3  A1  B1  NONE    <NA>   <NA>
4  A2  B2 X2_25      X2     25
5  A2  B3 fd_15      fd     15

            

      
      
      
    

        
+ cbind(F2,utils::strcapture("([^_]*)_(.*)", F2$text,
proto=data.frame(Before_=character(), After_=character())))
+ }
+ )
   user  system elapsed
 32.712   0.736  33.587

Cheers,
Bert




On Tue, Sep 22, 2020 at 5:45 PM Bill Dunlap <williamwdunlap at gmail.com>
wrote:

            

      
      
      
    

        

      
      
    

            

      
      
      
    

        

      
      
    

            

      
      
      
    

        

      
      
    

            

      
      
      
    

        

      
      
    

            

      
      
      
    

        

  
  


  


      

    

    

      
      

        


  

        

      Rui Barradas
    

    

      
      #
    

  


  

      
Hello,

If speed is important, and following the previous discussion and Bert's 
tests, here are two other alternatives, both faster.

1. Bert2 is Bert's original but with scan(., sep = "_") substituted for 
unlist/strsplit.
2. A package data.table solution. These are always fast, many times the 
fastest. But have the inconvenience of coercing the data to class 
"data.table" and the rest of the code needs to be adapted to handle 
data.tables. Namely, the second index in dt[i, j] is no longer a column 
index.

Unlike Bert, I time my first code, the one with package tidyr and its 
performance clearly beats the second one.
I define a test function, running several input sizes. It doesn't take 
much time to complete, only several minutes. The times' differences are 
not as impressive as Bert's, probably due to be on a different OS. I'm 
running R 4.0.2 on Ubuntu 20.04, sessionInfo at the end.

Also, I find X$Y1 <- as.integer(grepl("_", X$text)) more readable than 
coercion to numeric with +grepl(.).



library(data.table)
library(microbenchmark)
library(ggplot2)

Rui1 <- function(X){
   #X$Y1 <- as.integer(grepl("_", X$text))
   tidyr::separate(X, text, into = c("X1", "X2"), sep = "_", fill = "right")
}
Bert <- function(X){
   ## which are the  non "_" indices?
   wh <- grep("_",X$text, fixed = TRUE, invert = TRUE)
   ## paste "_." to these
   X[wh,"text"] <- paste(X[wh,"text"],".",sep = "_")
   ## Now strsplit() and unlist() them to get a vector
   z <- unlist(strsplit(X$text, "_"))
   ## now cbind() to the data frame
   cbind(X, matrix(z, ncol = 2, byrow = TRUE))
}
Bert2 <- function(X){
   wh <- grep("_",X$text, fixed = TRUE, invert = TRUE)
   X[wh,"text"] <- paste(X[wh,"text"],".",sep = "_")
   z <- scan(what = character(), text = X$text, sep = "_")
   cbind(X, matrix(z, ncol = 2, byrow = TRUE))
}
DT <- function(X){
   Y <- as.data.table(X)
   Y[, c("X1", "X2") := tstrsplit(text, "_", fixed = TRUE)]
}

testSeparate <- function(X, size = 0:6, times = 10){
   row_nums <- seq_len(nrow(X))
   res <- lapply(size, function(s){
     Y <- X[rep(row_nums, 10^s), ]
     mb <- microbenchmark(
       Rui = Rui1(Y),
       Bert = Bert(Y),
       Bert2 = Bert2(Y),
       DT = DT(Y),
       times = times
     )
     mb$size <- s
     mb
   })
   # return median times
   res <- do.call(rbind, res)
   aggregate(time ~ size + expr, res, median)
}

F1 <- read.table(text="ID1  ID2  text
A1 B1   NONE
A1 B1   cf_12
A1 B1   NONE
A2 B2   X2_25
A2 B3   fd_15  ",header=TRUE,stringsAsFactors=F)

agg <- testSeparate(F1, times = 5)

ggplot(agg, aes(size, time, color = expr)) +
   geom_line() + geom_point() +
   scale_y_continuous(trans = "log10") +
   xlab(expression(log[10] ~ "(size)")) +
   ylab(expression(log[10] ~ "(time)"))


sessionInfo()
R version 4.0.2 (2020-06-22)
Platform: x86_64-pc-linux-gnu (64-bit)
Running under: Ubuntu 20.04.1 LTS

Matrix products: default
BLAS:   /usr/lib/x86_64-linux-gnu/blas/libblas.so.3.9.0
LAPACK: /usr/lib/x86_64-linux-gnu/lapack/liblapack.so.3.9.0

locale:
  [1] LC_CTYPE=pt_PT.UTF-8       LC_NUMERIC=C
  [3] LC_TIME=pt_PT.UTF-8        LC_COLLATE=pt_PT.UTF-8
  [5] LC_MONETARY=pt_PT.UTF-8    LC_MESSAGES=pt_PT.UTF-8
  [7] LC_PAPER=pt_PT.UTF-8       LC_NAME=C
  [9] LC_ADDRESS=C               LC_TELEPHONE=C
[11] LC_MEASUREMENT=pt_PT.UTF-8 LC_IDENTIFICATION=C

attached base packages:
[1] stats     graphics  grDevices utils     datasets  methods   base

other attached packages:
[1] ggplot2_3.3.2        microbenchmark_1.4-7 data.table_1.12.8

loaded via a namespace (and not attached):
  [1] Rcpp_1.0.5       magrittr_1.5     tidyselect_1.1.0 munsell_0.5.0
  [5] colorspace_1.4-1 R6_2.4.1         rlang_0.4.7      dplyr_1.0.2
  [9] tools_4.0.2      grid_4.0.2       gtable_0.3.0     withr_2.2.0
[13] ellipsis_0.3.1   digest_0.6.25    tibble_3.0.3     lifecycle_0.2.0
[17] crayon_1.3.4     purrr_0.3.4      farver_2.0.3     tidyr_1.0.2
[21] vctrs_0.3.4      glue_1.4.2       labeling_0.3     stringi_1.4.6
[25] compiler_4.0.2   pillar_1.4.6     generics_0.0.2   scales_1.1.0
[29] pkgconfig_2.0.3


Hope this helps,

Rui Barradas


?s 02:47 de 23/09/20, Bert Gunter escreveu:

            

      
      
      
    

        

      
      
    

            

      
      
      
    

        

      
      
    

            

      
      
      
    

        

      
      
    

            

      
      
      
    

        

      
      
    

  
  


  


      

    

    

      
      

        


  

        

      LMH
    

    

      
      #
    

  


  

      
What is the delimiter is in the input data? Is it tab, space, etc?

Is this going to be the same for the output data that you will use for R input?

LMH

        
Val wrote:

            

      
      
      
    

        

      
      
    

            

      
      
      
    

        

      
      
    

            

      
      
      
    

        

      
      
    

            

      
      
      
    

        

      
      
    

  
  


  


      

    

    

      
      

        


  

        

      LMH
    

    

      
      #
    

  


  

      
Below is a script in bash the uses the awk tokenizer to do the work.

This assumes that your input and output delimiter is space. The number of consecutive delimiters in
the input is not important. This also assumes that the input file does not have a header row. That
is easy to modify if you want. I always keep header rows in my data files as I think that removing
them is asking for trouble down the road.

I added a NULL for cases where there is no value for the last field. You could use "." if you want.

You should be able to find how to run this from inside R if you want. You will, of course, need a
bash environment to run this, so if you are not in linux you will need cygwin or something similar.

This should be very fast, but let me know if needs to be faster. If the X1_X2 variant occurs less
frequently than not then we should switch the order in which the logic evaluates the options.

LMH


#! /bin/bash

# input filename
input_file=$1

# output filename
output_file=$2

# make sure the input file exists
if [ ! -f $input_file ]; then
   echo $input_file "  cannot be found"
   exit 0
fi

# create the output file
touch $output_file

# make sure the output was created
if [ ! -f $output_file ]; then
   echo $output_file "  was not created"
   exit 0
fi

# write the header row
echo "ID1 ID2 Y1 X1 X2" >> $output_file

# character to find in the third token
look_for='_'

# process with awk
# if the 3rd token contains '_'
#   split the third token on '_' into F[1] and F[2]
#   print the first two tokens, the indicator value of 1, and the split fields F[1] and F[2]
# otherwise,
#   print the first two tokens, the indicator value of 0, the 3rd token, and NULL

cat $input_file | \
awk -v find_char=$look_for '{ if($3 ~ find_char) { { split ($3, F, "_") }
                                                   { print $1, $2, "1", F[1], F[2] }
                                                 }
                              else { print $1, $2, "0", $3, "NULL" }
                            }' >> $output_file

        
Val wrote:

            

      
      
      
    

        

      
      
    

            

      
      
      
    

        

      
      
    

            

      
      
      
    

        

      
      
    

            

      
      
      
    

        

      
      
    

  
  


  


      

    

    

      
      

        


  

        

      Val
    

    

      
      #
    

  


  

      
Thank you again for your help  and giving me the opportunity to choose
the efficient method.  For a small data set there is no discernable
difference between the different approaches.  I will carry out a
comparison using  the large data set.

        
On Wed, Sep 23, 2020 at 11:52 AM LMH <lmh_users-groups at molconn.com> wrote: