cbind a dataframe with an empty dataframe - cbind.fill?

Question

I think I'm looking for an analog of rbind.fill (in Hadley's plyr package) for cbind. I looked, but there is no cbind.fill.

What I want to do is the following:

#set these just for this example
one_option <- TRUE
diff_option <- TRUE

return_df <- data.frame()

if (one_option) {
    #do a bunch of calculations, produce a data.frame, for simplicity the following small_df
    small_df <- data.frame(a=1, b=2)
    return_df <- cbind(return_df,small_df)
}

if (diff_option) {
    #do a bunch of calculations, produce a data.frame, for simplicity the following small2_df
    small2_df <- data.frame(l="hi there", m=44)
    return_df <- cbind(return_df,small2_df)
}

return_df

Understandably, this produces an error:

Error in data.frame(..., check.names = FALSE) : 
arguments imply differing number of rows: 0, 1

My current fix is to replace the line return_df <- data.frame() with return_df <- data.frame(dummy=1) and then the code works. I then just remove dummy from the return_df at the end. After adding the dummy and running the above code, I get

      dummy a b        l  m
1     1 1 2 hi there 44

I then just need to get rid of the dummy, e.g.:

> return_df[,2:ncol(return_df)]
  a b        l  m
1 1 2 hi there 44

I'm sure I'm missing an easier way to do this.

edit: I guess I'm not looking for a cbind.fill because that would mean that an NA value would be created after the cbind, which is not what I want.

Holna

Without a data set and an expected output it's difficult to know exactly what you want. – Tegan

@TylerRinker, you're right. I did describe my current fix, but I did not say explicitly what my desired result was. I have added that information in now. – Holna

Answer 1

Here's a cbind fill:

cbind.fill <- function(...){
    nm <- list(...) 
    nm <- lapply(nm, as.matrix)
    n <- max(sapply(nm, nrow)) 
    do.call(cbind, lapply(nm, function (x) 
        rbind(x, matrix(, n-nrow(x), ncol(x))))) 
}

Let's try it:

x<-matrix(1:10,5,2)
y<-matrix(1:16, 4,4)
z<-matrix(1:12, 2,6)

cbind.fill(x,y)
cbind.fill(x,y,z)
cbind.fill(mtcars, mtcars[1:10,])

I think I stole this from somewhere.

EDIT STOLE FROM HERE: LINK

Tegan

Thanks Tyler. But I have a data frame and not a matrix, and I think that does disrupt this implementation. Also, I've edited my post because I think I was mistaken -- I don't think I want a cbind.fill, because that would create an NA, whereas I want nothing created. – Holna

I edited my post to make the code work on both dataframes and matrices. I don't think it's possible to output a dataframe of unequal rows. By definition a dataframe is a list of equal length n's, so a dataframe not in a rectangular shape would not be possible. – Tegan

@Xu Wang, there are functions as.data.frame() and as.matrix(), so it is not problem that you have data.frame and not matrix. – Triton

Just a quick sitenote: The code is also available on Gist. – Seng

Answer 2

While, I think Tyler's solution is direct and the best here, I just provide the other way, using rbind.fill() that we already have.

require(plyr) # requires plyr for rbind.fill()
cbind.fill <- function(...) {                                                                                                                                                       
  transposed <- lapply(list(...),t)                                                                                                                                                 
  transposed_dataframe <- lapply(transposed, as.data.frame)                                                                                                                         
  return (data.frame(t(rbind.fill(transposed_dataframe))))                                                                                                                          
} 

Triton

Thanks Max! I like your solution. I will point out (for future readers) that this solution requires loading the package plyr, where Tyler's does not depend on any extra packages (I think). – Holna

@Xu Wang, added require(plyr) statement. – Triton

One note is that Max's solution returns a data frame where as mine returns a matrix. If you really wanted the function to return a matrix you could wrap the last line with as.data.frame() – Tegan

Answer 3

Using rowr::cbind.fill

rowr::cbind.fill(df1,df2,fill = NA)
   A B
1  1 1
2  2 2
3  3 3
4  4 4
5  5 5
6 NA 6

Chinaman

Looks like rowr was deprecated? – Avowed

Answer 4

cbind.na from the qpcR package can do that.

    install.packages("qpcR")
    library(qpcR)
    qpcR:::cbind.na(1, 1:7)

Curriculum

the function cbind.na no longer appears to be a part of the package qpcR – Nicolais

Answer 5

When a and b are data frames, following should work just fine:

ab <- merge(a, b, by="row.names", all=TRUE)[,-1]

or another possibility:

rows <- unique(c(rownames(a), rownames(b)))
ab <- cbind(a[rows ,], b[rows ,])

Durrace

Answer 6

I suggest a modification of Tyler's answer. My function allows cbind-ing of data.frames and/or matrices with vectors without loosing column names as it happens in Tyler's solution

cbind.fill <- function(...){
  nm <- list(...) 
  dfdetect <- grepl("data.frame|matrix", unlist(lapply(nm, function(cl) paste(class(cl), collapse = " ") )))
  # first cbind vectors together 
  vec <- data.frame(nm[!dfdetect])
  n <- max(sapply(nm[dfdetect], nrow)) 
  vec <- data.frame(lapply(vec, function(x) rep(x, n)))
  if (nrow(vec) > 0) nm <- c(nm[dfdetect], list(vec))
  nm <- lapply(nm, as.data.frame)

  do.call(cbind, lapply(nm, function (df1) 
    rbind(df1, as.data.frame(matrix(NA, ncol = ncol(df1), nrow = n-nrow(df1), dimnames = list(NULL, names(df1))))) )) 
}

cbind.fill(data.frame(idx = numeric()), matrix(0, ncol = 2), 
           data.frame(qwe = 1:3, rty = letters[1:3]), type = "GOOD", mark = "K-5")
#       idx V1 V2 qwe rty type mark
#     1  NA  0  0   1   a GOOD  K-5
#     2  NA NA NA   2   b GOOD  K-5
#     3  NA NA NA   3   c GOOD  K-5

Kweiyang

Answer 7

I just find a trick that when we want to add columns into an empty dataframe, just rbind it at first time, than cbind it later.

    newdf <- data.frame()
    # add the first column
    newdf <- rbind(newdf,data.frame("col1"=c("row1"=1,"row2"=2)))
    # add the second column
    newdf <- cbind(newdf,data.frame("col2"=c("row1"=3,"row2"=4)))
    # add more columns
    newdf <- cbind(newdf,data.frame("col3"=c("row1"=5,"row2"=6)))
    # result
    #     col1 col2 col3
    #row1    1    3    5
    #row2    2    4    6

I don't know why, but it works for me.

Finke

Answer 8

We could add id column then use merge:

df1 <- mtcars[1:5, 1:2]
#                    mpg cyl id
# Mazda RX4         21.0   6  1
# Mazda RX4 Wag     21.0   6  2
# Datsun 710        22.8   4  3
# Hornet 4 Drive    21.4   6  4
# Hornet Sportabout 18.7   8  5

df2 <- mtcars[6:7, 3:4]
#            disp  hp
# Valiant     225 105
# Duster 360  360 245

#Add id column then merge
df1$id <- seq(nrow(df1)) 
df2$id <- seq(nrow(df2)) 

merge(df1, df2, by = "id", all.x = TRUE, check.names = FALSE)
#   id  mpg cyl disp  hp
# 1  1 21.0   6  225 105
# 2  2 21.0   6  360 245
# 3  3 22.8   4   NA  NA
# 4  4 21.4   6   NA  NA
# 5  5 18.7   8   NA  NA

Feign

Answer 9

We can use a list instead of data.frame and convert it to a data.frame at the end. For instance:

df = list()
df2 = data.frame(col1 = 1:3, col2 = c('a','b','c'))
df = as.data.frame(cbind(df, as.matrix(df2)))
df

#   col1 col2
# 1    1    a
# 2    2    b
# 3    3    c

Zhang

Answer 10

To cbind.fill named lists, where the names overlap and you wish to cbind by names, Tyler's answer may be modified to the following:

cbind.fill <- function(...){
    nm <- list(...) 
    nm <- lapply(nm, as.matrix)
    names <- unique(do.call(c, lapply(nm, rownames)))
    res <- matrix(nrow = length(names), ncol = length(nm))
    rownames(res) <- names
    for(i in 1:length(nm))
      res[rownames(nm[[i]]),i] <- nm[[i]][,1]
    res
}

Priapic