CODEWITHSUNDEEP
rmergena
Looper
Looper

Reputation: 295

Merge rows with similar information

I have data frame with several rows and I need to merge the rows with same ID. 

a=read.csv("a.csv")
view(a)

ID  Value1  Value2  Value3  Value4  Value5  Value6
1076    2940    NA  NA  2   NA  NA
1076    2940    1   A-  NA  302 549
1109    2940    NA  NA  3   NA  NA
1109    2940    NA  A-  NA  700 150

I need the results like 

ID  Value1  Value2  Value3  Value4  Value5  Value6
1076    2940    1   A-  2   302 549
1109    2940    NA  A-  3   700 150

I have already reviewed the answer to a similar problem (Merging rows with shared information). But I am getting an error in the results.

library(dplyr)
f <- function(x) {
  x <- na.omit(x)
  if (length(x) > 0) paste(x,collapse='-') else NA
}
a_merge <- a %>% group_by(ID)%>%summarise_all(list(f))

But I am getting the following error

Error: Column `Value2` can't promote group 1 to character

Please help.

Upvotes: 3

Views: 160

Answers (3)

IceCreamToucan
IceCreamToucan

Reputation: 28675

If you use data.table you can avoid converting all the columns to lists and only convert the ones where it's required.

library(data.table)
setDT(df)

df[, lapply(.SD, function(x)
          if(length(vals <- unique(x[!is.na(x)])) > 1)
            list(vals)
          else vals), 
  by = ID]

#      ID Value1 Value2 Value3 Value4 Value5 Value6
# 1: 1076   2940    2,1     A-      2    302    549
# 2: 1109   2940            A-      3    700    150

If you're using toString you can remove the if and simplify things. This should apply to dplyr also.

df[, lapply(.SD, function(x) toString(unique(x[!is.na(x)]))),
  by = ID]
# 1: 1076   2940   2, 1     A-      2    302    549
# 2: 1109   2940            A-      3    700    150

Modified example data (added a case with >1 distinct value)

df <- fread('
ID  Value1  Value2  Value3  Value4  Value5  Value6
1076    2940    2  NA  2   NA  NA
1076    2940    1   A-  NA  302 549
1109    2940    NA  NA  3   NA  NA
1109    2940    NA  A-  NA  700 150
')

Upvotes: 1

Sotos
Sotos

Reputation: 51582

Here is a base R approach,

setNames(do.call(rbind.data.frame, lapply(split(df, df$ID), function(i) 
                                       sapply(i, function(j) j[!is.na(j)][1]))), names(df))

#    ID Value1 Value2 Value3 Value4 Value5 Value6
#1 1076   2940      1     A-      2    302    549
#2 1109   2940   <NA>     A-      3    700    150

Upvotes: 2

akrun
akrun

Reputation: 886948

An option would be to create a condition with if/else to return NA when all the values in the column is NA or else get the unique non-NA elements in a list

library(dplyr)
a %>% 
   group_by(ID) %>%
   summarise_all(list(~ list(if(all(is.na(.))) NA else unique(.[!is.na(.)]))))
# A tibble: 2 x 7
#     ID Value1    Value2    Value3    Value4    Value5    Value6   
#  <int> <list>    <list>    <list>    <list>    <list>    <list>   
#1  1076 <int [1]> <int [1]> <chr [1]> <int [1]> <int [1]> <int [1]>
#2  1109 <int [1]> <lgl [1]> <chr [1]> <int [1]> <int [1]> <int [1]>

EDIT:

1) Wrapped in a list

2) @Gregor's comment - get only the unique non-NA elements

data

a <- structure(list(ID = c(1076L, 1076L, 1109L, 1109L), Value1 = c(2940L, 
2940L, 2940L, 2940L), Value2 = c(NA, 1L, NA, NA), Value3 = c(NA, 
"A-", NA, "A-"), Value4 = c(2L, NA, 3L, NA), Value5 = c(NA, 302L, 
NA, 700L), Value6 = c(NA, 549L, NA, 150L)), class = "data.frame", row.names = c(NA, 
-4L))

Upvotes: 3

Related Questions