arranging strings from one data frame based on another one - r

Question

I have a data frame like this one
df1<- structure(list(V1 = structure(c(8L, 4L, 5L, 7L, 6L, 3L, 9L, 1L,
2L), .Label = c("A0A061AKW6;Q19219;A0A061AJ82;Q7JLR4", "A0A061AL89;A0A061AJK8;Q21920-2;Q21920-7;Q21920",
"C1P641;C1P640;A0A061AD21;G5EEV6", "O16276", "O16520-2", "O17323-2",
"O17395", "O17403", "Q22501;A0A061AE05"), class = "factor")), .Names = "V1", class = "data.frame", row.names = c(NA,
-9L))
My second data from looks like this
df2<- structure(list(From = structure(c(12L, 10L, 11L, 8L, 7L, 1L,
9L, 15L, 2L, 5L, 13L, 3L, 16L, 6L, 4L, 14L), .Label = c("A0A061AD21",
"A0A061AE05", "A0A061AJ82", "A0A061AJK8", "A0A061AKW6", "A0A061AL89",
"C1P640", "C1P641", "G5EEV6", "O16276", "O17395", "O17403", "Q19219",
"Q21920", "Q22501", "Q7JLR4"), class = "factor"), To = structure(c(4L,
8L, 1L, 5L, 5L, 5L, 5L, 6L, 6L, 2L, 2L, 2L, 2L, 3L, 3L, 7L), .Label = c("aat-3",
"CELE_F08G5.3", "CELE_R11A8.7", "cpsf-2", "epi-1", "pps-1", "R11A8.7",
"ugt-61"), class = "factor")), .Names = c("From", "To"), class = "data.frame", row.names = c(NA,
-16L))
df2 is taken from df1 but some information are added and some are removed . I want to reconstruct the df2 like df1 and arrange the column named To based on that
So the output should look like this
From To
O17403 cpsf-2
O16276 ugt-61
O16520-2 -
O17395 aat-3
O17323-2 -
C1P641;C1P640;A0A061AD21;G5EEV6 epi-1
Q22501;A0A061AE05 pps-1
A0A061AKW6;Q19219;A0A061AJ82;Q7JLR4 CELE_F08G5.3
A0A061AL89;A0A061AJK8;Q21920-2;Q21920-7;Q21920 CELE_R11A8.7; R11AB.7
It means we have O17403 in df2 and was only one string in df1, so it stays the same. O16276 was only one string in a raw in df1 so it also stays the same
O16520-2 was in df1 was not in df2 so in column named to a hyphen
the same for the rest until C1P641;C1P640;A0A061AD21;G5EEV6 are all in the same row of df1 and their To is the same, so we put them the same as df1 and just add one epi-1
Probably the best is to put df1 as template and then parse the To to it , those that are in df2, parse their To , those that are not only a hyphen
It is very complicated, I even could not think how to do it.I will appreciate any help

Answer 1

To solve this I split the semicolon delimited strings and created a nested for-for-if-if loop.
Here's the logic behind the loop which runs against the split string's data.frame (tmp):
Fix data classes (i.e. change factor to character to avoid conflicting level sets) and append a temporary To column to tmp
For each column and row of tmp start by seeing if a cell contains a valid string for matching and a matched value in df2$To, if not, go to the next iteration
If it does then look at the matching value in To from df2, checking to see if we already have the matched value in tmp$To (if so, go to next iteration)
If there's a new matched value in df2$To then put it in the correspond cell of tmp$To, prepending it with any preceeding matches and semicolons if it is not the first match for that row
df1$V1 <- as.character(df1$V1)
df2$From <- as.character(df2$From)
df2$To <- as.character(df2$To)
library(stringr)
tmp <- as.data.frame(str_split_fixed(df1$V1, ";",n=5), stringsAsFactors = F)
tmp$To <- as.character(NA)
for(j in 1:nrow(tmp)){
for(i in 1:ncol(tmp)){
if(length(df2$To[df2$From == tmp[j,i]]) == 0 | is.null(tmp[j,i])){
next
} else if(length(df2$To[df2$From == tmp[j,i]] ) == 1 & !is.na(tmp[j,i])){
if(is.na(tmp$To[j]) | tmp$To[j] == df2$To[df2$From == tmp[j,i]]){
tmp$To[j] <- df2$To[df2$From == tmp[j,i] ]
} else{
tmp$To[j] <- paste(tmp$To[j],";",df2$To[df2$From == tmp[j,i] ], sep="")
}
} else{
next
}
}
}
df1 <- data.frame(From=df1$V1, To=tmp$To)
df1
From To
1 O17403 cpsf-2
2 O16276 ugt-61
3 O16520-2 <NA>
4 O17395 aat-3
5 O17323-2 <NA>
6 C1P641;C1P640;A0A061AD21;G5EEV6 epi-1
7 Q22501;A0A061AE05 pps-1
8 A0A061AKW6;Q19219;A0A061AJ82;Q7JLR4 CELE_F08G5.3
9 A0A061AL89;A0A061AJK8;Q21920-2;Q21920-7;Q21920 CELE_R11A8.7;R11A8.7

Answer 2

One way of doing this is to use the splitstackshape package (use cSplit). I converted the factors to character strings to simplify (and get rid of warnings).
library(dplyr)
library(data.table) # cSplit from 'splitstackshape' returns a 'data.table'.
library(splitstackshape)
### Remove the factors for convenience of manipulation
df1 <- df1 %>% mutate(From = as.character(V1))
df2 <- df2 %>% mutate(From = as.character(From), To = as.character(To))
### 'cSplit' will split on ';' and create a new row for each item. The
### original 'From' column is kept around as cSplit removes the split column.
### 'rn' (row number) is used for ordering later.
cSplit(df1 %>% mutate(rn = row_number(), From_temp = From),
"From_temp", sep = ";", direction = "long", drop = FALSE, type.convert = FALSE) %>%
left_join(df2, by = c(From_temp = 'From')) %>% # Join to 'df2' to get the 'To' column
group_by(From, rn) %>% # Group by original 'From' column.
summarise(To = paste(sort(unique(na.omit(To))), collapse = ';'), # Create 'To' by joining 'To' Values
To = ifelse(To=='', '-', To)) %>% # Set empty values to '-'
ungroup %>%
arrange(rn) %>% # Sort by original row number and
select(-rn) # remove 'rn' column.
## From To
## <chr> <chr>
## 1 O17403 cpsf-2
## 2 O16276 ugt-61
## 3 O16520-2 -
## 4 O17395 aat-3
## 5 O17323-2 -
## 6 C1P641;C1P640;A0A061AD21;G5EEV6 epi-1
## 7 Q22501;A0A061AE05 pps-1
## 8 A0A061AKW6;Q19219;A0A061AJ82;Q7JLR4 CELE_F08G5.3
## 9 A0A061AL89;A0A061AJK8;Q21920-2;Q21920-7;Q21920 CELE_R11A8.7;R11A8.7
There may be a cleaner way to do with dplyr that doesn't require the splitstackshape.