How to print fields for repeated key column in one line - awk

Question

I'd like to transform a table in such a way that for duplicated
values in column #2 it would have corresponding values from column #1.
I.e. something like that...
MZ00024296 AC148152.3_FG005
MZ00047079 AC148152.3_FG006
MZ00028122 AC148152.3_FG008
MZ00032922 AC148152.3_FG008
MZ00048218 AC148152.3_FG008
MZ00024680 AC148167.6_FG001
MZ00013456 AC149475.2_FG003
to
AC148152.3_FG005 MZ00024296
AC148152.3_FG006 MZ00047079
AC148152.3_FG008 MZ00028122|MZ00032922|MZ00048218
AC148167.6_FG001 MZ00024680
AC149475.2_FG003 MZ00013456
As I need it to computations in R I tried to use:
x=aggregate(mz_grmz,by=list(mz_grmz[,2]),FUN=paste(mz_grmz[,1],sep="|"))
but it don't work (wrong function)
Error in match.fun(FUN) :
'paste(mz_grmz[, 1], sep = "|")' is not a function, character or symbol
I also remind myself about unstack() function, but it isn't what I need.
I tried to do it using awk, based on my base knowledge I reworked code given here:
site1
#! /bin/sh
for y do
awk -v FS="\t" '{
for (x=1;x<=NR;x++) {
if (NR>2 && x=x+1) {
print $2"\t"x
}
else {print NR}
}
}' $y > $y.2
done
unfortunately it doesn't work, it's only produce enormous file with field #2 and some numbers.
I suppose it is easy task, but it is above my skills right now.
Could somebody give me a hint? Maybe just function to use in aggregate in R.
Thanks

Answer 1

You could do it in awk like this:
awk '
{
if ($2 in a)
a[$2] = a[$2] "|" $1
else
a[$2] = $1
}
END {
for (i in a)
print i, a[i]
}' INFILE > OUTFILE

Answer 2

to keep the output as same as the text in your question (empty lines etc..):
awk '{if($0 &&($2 in a))a[$2]=a[$2]"|"$1;else if ($0) a[$2]=$1;}\
END{for(x in a){print x,a[x];print ""}}' inputFile
test:
kent$ echo "MZ00024296 AC148152.3_FG005
MZ00047079 AC148152.3_FG006
MZ00028122 AC148152.3_FG008
MZ00032922 AC148152.3_FG008
MZ00048218 AC148152.3_FG008
MZ00024680 AC148167.6_FG001
MZ00013456 AC149475.2_FG003"|awk '{if($0 &&($2 in a))a[$2]=a[$2]"|"$1;else if ($0) a[$2]=$1;}END{for(x in a){print x,a[x];print ""}}'
AC149475.2_FG003 MZ00013456
AC148152.3_FG005 MZ00024296
AC148152.3_FG006 MZ00047079
AC148152.3_FG008 MZ00028122|MZ00032922|MZ00048218
AC148167.6_FG001 MZ00024680

Answer 3

This GNU sed solution might work for you:
sed -r '1{h;d};H;${x;s/(\S+)\s+(\S+)/\2\t\1/g;:a;s/(\S+\t)([^\n]*)(\n+)\1([^\n]*)\n*/\1\2|\4\3/;ta;p};d' input_file
Explanation: Use the extended regex option-r to make regex's more readable. Read the whole file into the hold space (HS). Then on end-of-file, switch to the HS and firstly swap and tab separate fields. Then compare the first fields in adjacent lines and if they match, tag the second field from the second record to the first line separated by a |. Repeated until no further adjacent lines have duplicate first fields then print the file out.