Skip file existence check if stdin indicator exists

[sterrettJD]

Hi HISAT2 team,

This addresses Issue #451, where piping reads to hisat2 aligner via stdin does not work. Please let me know if you have any questions, or if there are any checks you'd like me to perform regarding this!

Thanks,
John

[mourisl]

Thanks for the reminder. I have examined the code. There are two issues:

1. There is a step in checking the read lengths in the wrapper (https://github.com/DaehwanKimLab/hisat2/blob/master/hisat2#L294-L318), which needs to use about 10,000 reads to generate length distributions and generate the option "--read-lengths" using the hisat2_read_statistics.py script. With the pipe input, the current implementation will cause this step fail so it won't generate the--read-lengths option. I think this is related to how HISAT2 handles some repeat regions based on the code. If we allow piping, then about 10,000 will be missed in the alignment as they are piped in this python script. So I guess your pipeline needs to run the python script explicitly and feeds the result to to the --read-lengths option to have a consistent result when using read file as input.

2. Another minor issue is that the piping should only work for single-end data, so may be the check is still necessary for mate1s and mate2s options.