Bio::SeqIO::fastq can't handle FASTQ with empty sequence and quality lines

[ntakebay]

After removing primers, there ends up with no sequence, which end up with FASTQ with empty sequence and quality lines like this:

@M0 M0:1

+

@M5 M5:1
TTTTTAGAG
+
GGGGGGGGG

When I try to read this FASTQ file withsomething like this:

my $seqio_obj = Bio::SeqIO->new(-file => $file, -format => 'fastq');
while (my $seq = $seqio_obj->next_seq()) {
    $seq->desc( $seq->id . " ". $seq->desc);
    push(@seqArr, $seq);
}

It stops processing with error:
------------- EXCEPTION: Bio::Root::Exception ------------- MSG: Missing sequence and/or quality data; line: 30 STACK: Error::throw STACK: Bio::Root::Root::throw /usr/local/share/perl5/5.40/Bio/Root/Root.pm:449 STACK: Bio::SeqIO::fastq::next_dataset /usr/local/share/perl5/5.40/Bio/SeqIO/fastq.pm:121 STACK: Bio::SeqIO::fastq::next_seq /usr/local/share/perl5/5.40/Bio/SeqIO/fastq.pm:52 STACK: /home/lab/bin/fastqSortByName.pl:44 -----------------------------------------------------------

In next_seq() of Bio/SeqIO/fastq.pm, there is a comment:

        # Are FASTQ sequences w/o any sequence valid?  Removing for now
        # -cjfields 6.22.09

and he chose to throw an error, but since empty fastq sequence does happen during analyses, it shouldn't quit when empty line is encountered, and it should read it in as empty sequence.

The version of Bio::SeqIO is the current ctan version 1.7.8.

[cjfields]

Yep, this was way back in 2009, around the time we published the FASTQ paper. We probably should aim to be somewhat consistent with other Bio*. @peterjc how does Biopython FASTQ parsing handle empty (0-length) sequence records? Error, ignore, or valid empty sequence?

[peterjc]

As I recall, treated as valid zero length sequence. The main usecase was filtered paired reads where one read is good but the other trimmmed to nothing (the orphan can still be useful). Note the blank sequence line and blank quality line are essential.

[cjfields]

As I recall, treated as valid zero length sequence. The main usecase was filtered paired reads where one read is good but the other trimmmed to nothing (the orphan can still be useful). Note the blank sequence line and blank quality line are essential.

Okay, that makes sense. This should be easy enough to implement but I need to check whether the current tests expect an exception on these and adjust accordingly.