merge almost identical rows filtering NAs and shor

Question

I have some almost identical rows in a dataframe, see ex., the criteria to establish they are related are some variables "sel1,sel2" in this example, the other variables, var1 and var2, must be integrated by the following criteria: 1. discarding NA, or 2. discarding the shorter string (in var2 in the example). So, until now I have discarded the NA, but not find a way to at the same time discard the shorter string. The strings are complex and might have commas, spaces and several types of characters.

df <- read.table(text = 
            "  sel1 sel2 var1    var2
1   pseudorepeated1   x    NA    \"longer string\"   # keep longer string instead of shortstring
2   pseudorepeated1   x    2     \"short string\"    # keep 2 instead of NA
3   pseudorepeated2   y    NA    \"longer string 2\" # keep longer string 2
4   pseudorepeated2   y    4     \"short string2\"   # keep 4
5                 3   x    gs    as
6                 4   y    fg    df
7                 5   x    eg    af
8                 6   y    df    fd", header = TRUE, stringsAsFactors=F)
df
df[is.na(df)] <- ""
df2<-aggregate(. ~ sel1 + sel2,data=df,FUN=function(X)paste(unique((X))) )
paste_noNA <- function(x,sep=", ") 
  gsub(", " ,sep, toString(x[!is.na(x) & x!="" & x!="NA"] ) )
df3<-as.data.frame(lapply(df2, function(X) unlist(lapply(X, function(x) paste_noNA(x)) ) ), 
                           stringsAsFactors=F )

The expected output does not have the ", short string" text in this table.

df3
               sel1 sel2 var1                        var2
1.1               3    x   gs                          as
1.3               5    x   eg                          af
1.5 pseudorepeated1    x    2 longer string, short string# only longer string desired
2.2               4    y   fg                          df
2.4               6    y   df                          fd
2.6 pseudorepeated2    y    4 longer string 2, short string2# only longer string 2 desired

Answer 1

group by sel1 and sel2 and remove NA in var1, and replace shorter string with longer string in var2. Finally, remove the duplicates in it.

library('data.table')
setDT(df)
df[, `:=` ( var2 = { temp <- nchar(var2); var2[ temp == max(temp) ] },
            var1 = na.omit(var1)),
   by = .(sel1, sel2)]
df[ !duplicated( df ), ]

#               sel1 sel2 var1         var2
# 1: pseudorepeated1    x    2 longerstring
# 2: pseudorepeated2    y    4 longerstring
# 3:               3    x   gs           as
# 4:               4    y   fg           df
# 5:               5    x   eg           af
# 6:               6    y   df           fd

EDIT: with many columns

Data:

df <- read.table(text = 
                   "  sel1 sel2 var1    var2
                 1   pseudorepeated1   x    NA    longerstring   # keep longerstring instead of shortstring
                 2   pseudorepeated1   x    2     shortstring    # keep 2 instead of NA
                 3   pseudorepeated2   y    NA    longerstring   # same as above
                 4   pseudorepeated2   y    4     shortstring    # same as above
                 5                 3   x    gs    as
                 6                 4   y    fg    df
                 7                 5   x    eg    af
                 8                 6   y    df    fd", header = TRUE, stringsAsFactors=F)

library('data.table')
setDT(df)
df$var3 <- df$var2
df$var4 <- df$var2

Code:

for( nm in c( "var1", "var2", "var3", "var4") ){
  df[,  paste0(nm) := { temp <- na.omit(get(nm)); temp[ nchar(temp) == max(nchar(temp)) ] },
     by = .(sel1, sel2)]
}
df[ !duplicated( df ), ]

Output:

#               sel1 sel2 var1         var2         var3         var4
# 1: pseudorepeated1    x    2 longerstring longerstring longerstring
# 2: pseudorepeated2    y    4 longerstring longerstring longerstring
# 3:               3    x   gs           as           as           as
# 4:               4    y   fg           df           df           df
# 5:               5    x   eg           af           af           af
# 6:               6    y   df           fd           fd           fd

EDIT 2: avoiding for loop, and using .SDcols and a column names variable

col_nm <- c( "var1", "var2", "var3", "var4")

df[,  paste0(col_nm) := lapply( .SD, function(x) { 
  temp <- na.omit(x)
  temp[ nchar(temp) == max(nchar(temp)) ] } ),
  by = .(sel1, sel2), 
  .SDcols = col_nm ]  

df[ !duplicated( df ), ]