Collapse all columns by an ID column [duplicate]

Answer 1

Here's an option using summarise_each (which makes it easy to apply the changes to all columns except the grouping variables) and toString:

require(dplyr)

have %>%
  group_by(ID) %>%
  summarise_each(funs(toString))

#Source: local data frame [5 x 3]
#
#     ID          info1                          info2
#1 id101            one                          first
#2 id102     twoA, twoB second alias A, second alias B
#3 id103 threeA, threeB   third alias A, third alias B
#4 id104           four                         fourth
#5 id105           five                          fifth

Or, if you want it separated by semicolons, you can use:

have %>%
  group_by(ID) %>%
  summarise_each(funs(paste(., collapse = "; ")))

Answer 2

Good old aggregate does this just fine

aggregate(have[,2:3], by=list(have$ID), paste, collapse=";")

Question is: does it scale?

Answer 3

library(stringr)
library(dplyr)
have %>% tbl_df %>% group_by(ID) %>% summarise_each(funs(str_c(., collapse="; ")))

Edit 1: So tbl_df may not needed and instead of the str_c of the stringr package you could use paste (in base). And what the above does is to group by the ID column and then apply the str_c (or paste) function to each remaining column for each group.

Edit 2: Another solution using the data.table package:

library(data.table)
dtbl <- as.data.table(have)
dtbl[,lapply(.SD, function(x) paste(x,collapse=";")), by=ID]

The above may be faster, especially if you set the key:

setkey(dtbl, ID)

"Hybrid" solution: You can use the dplyr syntax for data.tables! For example:

dtbl %>% tbl_dt %>%
     group_by(ID) %>% 
     summarise_each(funs(paste(., collapse="; ")))

Answer 4

Here is SQL solution^1:

library(sqldf)
#Static solution
sqldf("
SELECT ID,
       GROUP_CONCAT(info1,';') as info1,
       GROUP_CONCAT(info2,';') as info2
FROM have
GROUP BY ID")

#Dynamic solution
concat_cols <- colnames(have)[2:ncol(have)]
group_concat <-
  paste(paste0("GROUP_CONCAT(",concat_cols,",';') as ", concat_cols),
        collapse = ",")
sqldf(
  paste("
      SELECT ID,",
      group_concat,"
      FROM have
      GROUP BY ID"))

# Same output for both static and dynamic solutions
#      ID         info1                         info2
# 1 id101           one                         first
# 2 id102     twoA;twoB second alias A;second alias B
# 3 id103 threeA;threeB   third alias A;third alias B
# 4 id104          four                        fourth
# 5 id105          five                         fifth

^1 - probably data.table solution would perform better with millions of rows, luckily we don't have that many genes yet :)

Answer 5

Here's a data.table solution.

library(data.table)
setDT(have)[, lapply(.SD, paste, collapse = "; "), by = ID]
#       ID          info1                          info2
# 1: id101            one                          first
# 2: id102     twoA; twoB second alias A; second alias B
# 3: id103 threeA; threeB   third alias A; third alias B
# 4: id104           four                         fourth
# 5: id105           five                          fifth