Column Maniputation - etl

Question

I would like to group my rows of inputs according to the first string of many comma-separated.
Basically, there will be 3 groups, which are "Motifs", "Chromatin_Structure" and "Protein_Binding"
The important output is the third one after 2 "|".
There might be some duplicates too like K562. The duplicates are not needed.
If the strings are not present, simply put a ". (dot)"
Input:
Motifs|PWM|Sox17|,Motifs|PWM|Sox8|,Chromatin_Structure|DNase-seq|K562|Znf4g7d3,Chromatin_Structure|DNase-seq|K562|,Chromatin_Structure|DNase-seq|TCF7L2|Znfe103c6,Protein_Binding|ChIP-seq|CTCF|HeLa-S3|,Protein_Binding|ChIP-seq|CTCF|HeLa-S3|
.
Motifs|PWM|TCF11|
Protein_Binding|ChIP-seq|MAFF|HepG2|
Desired output:
Sox17,Sox8 K562,TCF7L2 CTCF
. . .
TCF11 . .
. . MAFF
Codes that I tried.
sed 's/Motifs|PWM|//'
Appreciate your helps!

Answer 1

Perl one-liner (Using the term loosely):
$ perl -F, -lane '
my (%groups, #output);
for my $grp (#F) {
my #x = split /\|/, $grp;
$groups{$x[0]}{$x[2]} = 1;
}
for my $n (qw/Motifs Chromatin_Structure Protein_Binding/) {
if (exists $groups{$n}) {
push #output, join(",", sort keys %{$groups{$n}});
} else {
push #output, ".";
}
}
print join("\t", #output);' input.csv
Sox17,Sox8 K562,TCF7L2 CTCF
. . .
TCF11 . .
. . MAFF
And because I think it's woefully underappreciated as a scripting language, a tcl version:
#!/usr/bin/env tclsh
proc main {} {
while {[gets stdin line] >= 0} {
foreach grp [split $line ,] {
set x [split $grp |]
dict set groups [lindex $x 0] [lindex $x 2] 1
}
foreach n {Motifs Chromatin_Structure Protein_Binding} {
if {[dict exists $groups $n]} {
lappend output [join [dict keys [dict get $groups $n]] ,]
} else {
lappend output .
}
}
puts [join $output \t]
unset groups output
}
}
main
$ ./example.tcl < input.csv
Sox17,Sox8 K562,TCF7L2 CTCF
. . .
TCF11 . .
. . MAFF