R: how to save this special file as csv file?

My input file start from an ordinary csv table.

x <- read.table(textConnection(
+ ' models cores  time 
+ aa c1 xxx|yyy 
+ aa c2 xxx|zzz 
+ aa c3 www 
+ aa c4 xxx|vvv 
+ bb c1 vvv|www 
+ bb c2 www|qqq 
+ bb c3 xxx|uuu 
+ bb c4 uuu' ), header=TRUE)

It is a file with factor as all entry, as shown in following:

> str(x)
'data.frame':   8 obs. of  3 variables:
 $ models: Factor w/ 2 levels "aa","bb": 1 1 1 1 2 2 2 2
 $ cores : Factor w/ 4 levels "c1","c2","c3",..: 1 2 3 4 1 2 3 4
 $ time  : Factor w/ 8 levels "uuu","vvv|www",..: 7 8 3 6 2 4 5 1

In order to split the last column with the command "strsplit", I have done the following steps with reference to previous questions posted.

> write.csv(x, file="x.csv")
> y <- read.csv(file="x.csv",header=TRUE,stringsAsFactors=FALSE)
> str(y)

'data.frame':   8 obs. of  4 variables:
 $ X     : int开发者_高级运维  1 2 3 4 5 6 7 8
 $ models: chr  "aa" "aa" "aa" "aa" ...
 $ cores : chr  "c1" "c2" "c3" "c4" ...
 $ time  : chr  "xxx|yyy" "xxx|zzz" "www" "xxx|vvv" ...
Warning messages:
1: closing unused connection 4 (" models cores  time \naa c1 xxx|yyy \naa c2 xxx|zzz \naa c3 www \naa c4 xxx|vvv \nbb c1 vvv|www \nbb c2 www|qqq \nbb c3 xxx|uuu \nbb c4 uuu") 
2: closing unused connection 3 (" models cores  time \n4 1 0.000365 \n4 2 0.000259 \n4 3 0.000239 \n4 4 0.000220 \n8 1 0.000259 \n8 2 0.000249 \n8 3 0.000251 \n8 4 0.000258") 

> df2 <- as.data.frame(
+ t(
+ do.call(cbind,
+ lapply(1:nrow(y),function(x){
+ sapply(unlist(strsplit(y[x,4],"\\|")),c,y[x,2:3],USE.NAMES=FALSE)
+ })     )   ) )
> str(df2)

The result is what I needed.

> df2
    V1 models cores
1  xxx     aa    c1
2  yyy     aa    c1
3  xxx     aa    c2
4  zzz     aa    c2
5  www     aa    c3
6  xxx     aa    c4
7  vvv     aa    c4
8  vvv     bb    c1
9  www     bb    c1
10 www     bb    c2
11 qqq     bb    c2
12 xxx     bb    c3
13 uuu     bb    c3
14 uuu     bb    c4

When I type str(df2), I found all entry is a list of chr:

'data.frame':   14 obs. of  3 variables:
 $ V1    :List of 14
  ..$ : chr "xxx"...
$ models:List of 14
  ..$ : chr "aa"
  ..$ : chr "aa"
$ models:List of 14
  ..$ : chr "aa"
  ..$ : chr "aa"

However, I have difficulty to save this final results as csv table again.

> write.csv(df2, file="df2.csv")
Error in write.table(x, file, nrow(x), p, rnames, sep, eol, na, dec, as.integer(quote),  : 
  unimplemented type 'list' in 'EncodeElement'

How can I save the df2 file again in csv format? Pls help.

What you are doing seems epically silly - why write something out to CSV just to read back in again? - but given that df2 is roughly how you want it, you need to unlist() the three components in df2 and cast back as a data frame.

out <- data.frame(lapply(df2, function(x) factor(unlist(x))))

That gives us:

> out
    V1 models cores
1  xxx     aa    c1
2  yyy     aa    c1
3  xxx     aa    c2
4  zzz     aa    c2
5  www     aa    c3
6  xxx     aa    c4
7  vvv     aa    c4
8  vvv     bb    c1
9  www     bb    c1
10 www     bb    c2
11 qqq     bb    c2
12 xxx     bb    c3
13 uuu     bb    c3
14 uuu     bb    c4
> str(out)
'data.frame':   14 obs. of  3 variables:
 $ V1    : Factor w/ 7 levels "qqq","uuu","vvv",..: 5 6 5 7 4 5 3 3 4 4 ...
 $ models: Factor w/ 2 levels "aa","bb": 1 1 1 1 1 1 1 2 2 2 ...
 $ cores : Factor w/ 4 levels "c1","c2","c3",..: 1 1 2 2 3 4 4 1 1 2 ...

Which can be read out and in again:

> write.csv(out, file="out.csv", row.names = FALSE)
> read.csv("out.csv")
    V1 models cores
1  xxx     aa    c1
2  yyy     aa    c1
3  xxx     aa    c2
4  zzz     aa    c2
5  www     aa    c3
6  xxx     aa    c4
7  vvv     aa    c4
8  vvv     bb    c1
9  www     bb    c1
10 www     bb    c2
11 qqq     bb    c2
12 xxx     bb    c3
13 uuu     bb    c3
14 uuu     bb    c4

Update: It would be simpler to go straight from x to the desired output instead of reading it out to CSV and back in again and then processing y. For example this goes from x directly to the same result as out from above:

V1 <- with(x, strsplit(as.character(time), "\\|"))
lens <- lapply(V1, length)

out2 <- data.frame(V1 = factor(unlist(V1)),
                   models = with(x, rep(models, times = lens)),
                   cores = with(x, rep(cores, times = lens)))

Which gives:

> out2
    V1 models cores
1  xxx     aa    c1
2  yyy     aa    c1
3  xxx     aa    c2
4  zzz     aa    c2
5  www     aa    c3
6  xxx     aa    c4
7  vvv     aa    c4
8  vvv     bb    c1
9  www     bb    c1
10 www     bb    c2
11 qqq     bb    c2
12 xxx     bb    c3
13 uuu     bb    c3
14 uuu     bb    c4
> str(out2)
'data.frame':   14 obs. of  3 variables:
 $ V1    : Factor w/ 7 levels "qqq","uuu","vvv",..: 5 6 5 7 4 5 3 3 4 4 ...
 $ models: Factor w/ 2 levels "aa","bb": 1 1 1 1 1 1 1 2 2 2 ...
 $ cores : Factor w/ 4 levels "c1","c2","c3",..: 1 1 2 2 3 4 4 1 1 2 ...
> all.equal(out, out2)
[1] TRUE

Aside: As an aside, you don't make it easy for us to paste in your code as you just copied it from the R Console so it includes the prompts (+). Instead you could have done dput(x) and pasted that into your Q:

structure(list(models = structure(c(1L, 1L, 1L, 1L, 2L, 2L, 2L, 
2L), .Label = c("aa", "bb"), class = "factor"), cores = structure(c(1L, 
2L, 3L, 4L, 1L, 2L, 3L, 4L), .Label = c("c1", "c2", "c3", "c4"
), class = "factor"), time = structure(c(7L, 8L, 3L, 6L, 2L, 
4L, 5L, 1L), .Label = c("uuu", "vvv|www", "www", "www|qqq", "xxx|uuu", 
"xxx|vvv", "xxx|yyy", "xxx|zzz"), class = "factor")), .Names = c("models", 
"cores", "time"), class = "data.frame", row.names = c(NA, -8L
))

then we could all have simply done:

x <- structure(list(models = structure(c(1L, 1L, 1L, 1L, 2L, 2L, 2L, 
2L), .Label = c("aa", "bb"), class = "factor"), cores = structure(c(1L, 
2L, 3L, 4L, 1L, 2L, 3L, 4L), .Label = c("c1", "c2", "c3", "c4"
), class = "factor"), time = structure(c(7L, 8L, 3L, 6L, 2L, 
4L, 5L, 1L), .Label = c("uuu", "vvv|www", "www", "www|qqq", "xxx|uuu", 
"xxx|vvv", "xxx|yyy", "xxx|zzz"), class = "factor")), .Names = c("models", 
"cores", "time"), class = "data.frame", row.names = c(NA, -8L
))

Same with the call to create df2. This would have been preferable:

write.csv(x, file="x.csv")
y <- read.csv(file="x.csv", header=TRUE, stringsAsFactors=FALSE)
df2 <- data.frame(
 t(do.call(cbind,
       lapply(1:nrow(y),function(x){
            sapply(unlist(strsplit(y[x,4],"\\|")),c,y[x,2:3],
                     USE.NAMES=FALSE)
            }))))

That way it is a simple matter for us to reconstruct what objects you have and what you tried.

fun_transform <- function(.x){
    time_split <- strsplit(.x$time,split="\\|")
    n_rec <- sapply(time_split,length)
    ind <- rep(seq(nrow(.x)),n_rec)
    cbind(.x[ind,1:2],time=unlist(time_split,use.names=FALSE))
}

df2 <- fun_transform(y)

EDIT - example data

txt <- textConnection(
        ' models cores  time 
         aa c1 xxx|yyy 
         aa c2 xxx|zzz 
         aa c3 www 
         aa c4 xxx|vvv 
         bb c1 vvv|www 
         bb c2 www|qqq 
         bb c3 xxx|uuu 
         bb c4 uuu' )

y <- read.table(txt, header=TRUE,as.is=TRUE)
close(txt)