CODEWITHSUNDEEP
rlistmemorylapply
Max_sof
Max_sof

Reputation: 1

use a function on several columns of dataframes stored in a list (memory issues)

I have a list of 51 dataframes alltogether 90 MB with the mV measured by some sensors (column 2:7):

str(BoxlistF)

List of 51
$ :'data.frame':  33507 obs. of  7 variables:
  ..$ Date.Box                            : POSIXct[1:33507], format: "2013-01-01 00:15:00" ...
..$ Slnt.grn.00x.Box.001.SPADE.1: num [1:33507] 1811 1811 1810 1811 1810 ...
..$ Slnt.grn.00x.Box.001.SPADE.2: num [1:33507] 1739 1739 1737 1737 1736 ...
..$ Slnt.grn.00x.Box.001.SPADE.3: num [1:33507] 1634 1635 1634 1634 1637 ...
..$ Slnt.grn.00x.Box.001.SPADE.4: num [1:33507] 1572 1576 1576 1575 1576 ...
..$ Slnt.grn.00x.Box.001.SPADE.5: num [1:33507] 1660 1660 1659 1660 1659 ...
..$ Slnt.grn.00x.Box.001.SPADE.6: num [1:33507] 1454 1450 1453 1450 1451 ...

To remove measurement errors, I would like to use this ifelse function:

tt<-30
tb=-10
t<-2100
b<-800

l.new1<-lapply(BoxlistF, function(x){ x[,2:7]<-lapply(x[,2:7],function(x) ifelse(x<b|x>t,x<-NA,x))})

another Try was to exclude the Date Column of the function:

lapply(BoxlistF, function(x) ifelse(x[,-1]<b|x[,-1]>t,x[,-1]<-NA,x[,-1]))

So since my column names are varying and complicated I want to address the columns by indexes and not by names.

I dont know if this function works proper since R aborts it complaining about memory issues..:

Error: cannot allocate vector of size 131 Kb
In addition: Warning messages:
1: In ifelse(x[, -1] < b | x[, -1] > t, x[, -1] <- NA, x[, -1]) :
  Reached total allocation of 8147Mb: see help(memory.size)
2: In ifelse(x[, -1] < b | x[, -1] > t, x[, -1] <- NA, x[, -1]) :
  Reached total allocation of 8147Mb: see help(memory.size)
Called from: top level 
Error during wrapup: cannot allocate vector of size 512 Kb
Error during wrapup: target context is not on the stack

In the early stages I had a for loop like this:

for(i in 1:(length(BoxlistF)))
{for(j in 1:6)
{for(e in 1:(nrow(BoxlistF[[i]])))
  {
  ifelse((!is.na(BoxlistF[[i]][e,1+j])<b|!is.na(BoxlistF[[i]][e,1+j])>t),BoxlistF[[i]][e,1+j]<-NA,BoxlistF[[i]][e,1+j])
  ifelse((!is.na(BoxlistT[[i]][e,1+j])<tb|!is.na(BoxlistT[[i]][e,1+j])>tt),BoxlistT[[i]][e,1+j]<-NA,BoxlistT[[i]][e,1+j])
  }
}
}

This worked well on monthly data, but on this big data set (one year) I guess its not an option.

About solving memory issues in loops I found this post: Speed up the loop operation in R I tried to follow the instructions and came up with this:

l.new1<-lapply(BoxlistF,function(x)
{ res <- numeric(nrow(x))
  for(ff in 2:7)
  {
  for(cc in 1:nrow(BoxlistF[[1]]))
  {ifelse(x[cc,ff]<b|x[cc,ff]>t,res[cc]<-NA,res[cc]<-x[cc,ff])
   x[,ff]<-res
   return(x)
  }}})

that function worked quick, but well dont understand the outcome:

Date.Box    Slnt.grn.00x.Box.001.SPADE.1    Slnt.grn.00x.Box.001.SPADE.2    Slnt.grn.00x.Box.001.SPADE.3    Slnt.grn.00x.Box.001.SPADE.4    Slnt.grn.00x.Box.001.SPADE.5    Slnt.grn.00x.Box.001.SPADE.6
1   2013-01-01 00:15:00 1811    1739    1634    1572    1660    1454
2   2013-01-01 01:09:00 0   1739    1635    1576    1660    1450
3   2013-01-01 02:03:00 0   1737    1634    1576    1659    1453
4   2013-01-01 02:57:00 0   1737    1634    1575    1660    1450
5   2013-01-01 03:51:00 0   1736    1637    1576    1659    1451
6   2013-01-01 04:46:00 0   1739    1634    1575    1659    1451
7   2013-01-01 05:40:00 0   1734    1643    1576    1660    1450
8   2013-01-01 06:34:00 0   1734    1634    1576    1660    1449
9   2013-01-01 07:28:00 0   1734    1643    1572    1660    1447
10  2013-01-01 08:22:00 0   1734    1634    1576    1657    1448

I hope my information provided is sufficient, if not tell me and I try to add up missing information.

Upvotes: 0

Views: 58

Answers (2)

hedgedandlevered
hedgedandlevered

Reputation: 2384

Don't loop with R. Its not C. This is an example of how to replace certain elements in a data frame effectively.

> testdat <- data.frame(matrix(1:16,ncol=4))
> testdat == 6 | testdat == 9
        X1    X2    X3    X4
[1,] FALSE FALSE  TRUE FALSE
[2,] FALSE  TRUE FALSE FALSE
[3,] FALSE FALSE FALSE FALSE
[4,] FALSE FALSE FALSE FALSE
> testdat[testdat == 6 | testdat == 9] <- 999
> testdat
  X1  X2  X3 X4
1  1   5 999 13
2  2 999  10 14
3  3   7  11 15
4  4   8  12 16

Upvotes: 0

Max_sof
Max_sof

Reputation: 1

I cant say that I understand in detail, why this one works but luckily it does:

BoxlistF<-lapply(BoxlistF, function(x) x<-cbind(x["Date.Box"], apply(x[,-1],c(1,2),function(x)ifelse(x<b|x>t,x<-NA,x))))

So if want to check every column and row with an apply loop I have to use apply? or are things like this as well possible with lapply?

Upvotes: 0

Related Questions