awk totally separate duplicate and non-duplicates - awk

Question

If we have an input:
TargetIDs,CPD,Value,SMILES
95,CPD-1111111,-2,c1ccccc1
95,CPD-2222222,-3,c1ccccc1
95,CPD-2222222,-4,c1ccccc1
95,CPD-3333333,-1,c1ccccc1N
Now we would like to separate the duplicates and non-duplicates based on the fourth column (smiles)
duplicate:
95,CPD-1111111,-2,c1ccccc1
95,CPD-2222222,-3,c1ccccc1
95,CPD-2222222,-4,c1ccccc1
non-duplicate
95,CPD-3333333,-1,c1ccccc1N
Now the following attempt could do separate the duplicate without any problem. However, the first occurrence of the duplicate will still be included into the non-duplicate file.
BEGIN { FS = ","; f1="a"; f2="b"}
{
# Keep count of the fields in fourth column
count[$4]++;
# Save the line the first time we encounter a unique field
if (count[$4] == 1)
first[$4] = $0;
# If we encounter the field for the second time, print the
# previously saved line
if (count[$4] == 2)
print first[$4] > f1 ;
# From the second time onward. always print because the field is
# duplicated
if (count[$4] > 1)
print > f1;
if (count[$4] == 1) #if (count[$4] - count[$4] == 0) <= change to this doesn't work
print first[$4] > f2;
duplicate output results from the attempt:
95,CPD-1111111,-2,c1ccccc1
95,CPD-2222222,-3,c1ccccc1
95,CPD-2222222,-4,c1ccccc1
non-duplicate output results from the attempt
TargetIDs,CPD,Value,SMILES
95,CPD-3333333,-1,c1ccccc1N
95,CPD-1111111,-2,c1ccccc1
May I know if any guru might have comments/solutions? Thanks.

Answer 1

I would do this:
awk '
NR==FNR {count[$2] = $1; next}
FNR==1 {FS=","; next}
{
output = (count[$NF] == 1 ? "nondup" : "dup")
print > output
}
' <(cut -d, -f4 input | sort | uniq -c) input
The process substitution will pre-process the file and perform a count on the 4th column. Then, you can process the file and decide if that line is "duplicated".
All in awk: Ed Morton shows a way to collect the data in a single pass. Here's a 2 pass solution that's virtually identical to my example above
awk -F, '
NR==FNR {count[$NF]++; next}
FNR==1 {next}
{
output = (count[$NF] == 1 ? "nondup" : "dup")
print > output
}
' input input
Yes, the input file is given twice.

Answer 2

$ cat tst.awk
BEGIN{ FS="," }
NR>1 {
if (cnt[$4]++) {
dups[$4] = nonDups[$4] dups[$4] $0 ORS
delete nonDups[$4]
}
else {
nonDups[$4] = $0 ORS
}
}
END {
print "Duplicates:"
for (key in dups) {
printf "%s", dups[key]
}
print "\nNon Duplicates:"
for (key in nonDups) {
printf "%s", nonDups[key]
}
}
$ awk -f tst.awk file
Duplicates:
95,CPD-1111111,-2,c1ccccc1
95,CPD-2222222,-3,c1ccccc1
95,CPD-2222222,-4,c1ccccc1
Non Duplicates:
95,CPD-3333333,-1,c1ccccc1N

Answer 3

This solution only works if the duplicates are grouped together.
awk -F, '
function fout( f, i) {
f = (cnt > 1) ? "dups" : "nondups"
for (i = 1; i <= cnt; ++i)
print lines[i] > f
}
NR > 1 && $4 != lastkey { fout(); cnt = 0 }
{ lastkey = $4; lines[++cnt] = $0 }
END { fout() }
' file

Answer 4

Little late
My version in awk
awk -F, 'NR>1{a[$0":"$4];b[$4]++}
END{d="\n\nnondupe";e="dupe"
for(i in a){split(i,c,":");b[c[2]]==1?d=d"\n"i:e=e"\n"i} print e d}' file
Another built similar to glenn jackmans but all in awk
awk -F, 'function r(f) {while((getline <f)>0)a[$4]++;close(f)}
BEGIN{r(ARGV[1])}{output=(a[$4] == 1 ? "nondup" : "dup");print >output} ' file