Length template option 3

[joe-angell]

closer to what i had in mind, the counter output is a bit lame, but could be improved.

[joe-angell]

@dstreett what do you think of this approach?

[joe-angell]

@msettles @dstreett @samhunter This turned into a bigger refactor, removed all the dynamic casts (except from the test code because I'm not too concerned about them, even though most are unnecessary), and simplified the interface as much as I could. Made writer_helper a class that uses a visitor pattern to avoid dynamic_cast.

[msettles]

So quick look (avoiding preparing for next weeks workshop) but the changes seem pretty significant. If I'm reading this correctly it generalizes the concept of "read" more, so if a "Read" has size 1 its single-end if size 2 must be paired end, this actually would allow for more possibilities for types of reads, ala PacBio which can have more than 2 reads (here called a subread) per read, but they mean something a little different. An additional annotation to the read could specify if reads in file are "paired" (ends of a fragment) or "subreads" (circular fragment sequencing).

[joe-angell]

So quick look (avoiding preparing for next weeks workshop) but the changes seem pretty significant. If I'm reading this correctly it generalizes the concept of "read" more, so if a "Read" has size 1 its single-end if size 2 must be paired end, this actually would allow for more possibilities for types of reads, ala PacBio which can have more than 2 reads (here called a subread) per read, but they mean something a little different. An additional annotation to the read could specify if reads in file are "paired" (ends of a fragment) or "subreads" (circular fragment sequencing).

I added a new interface to readbase which lets you iterate over the reads by way of a vector. So if you want to filter all reads longer than 100 or whatever that is easy to do, but we still have Single and Paired end reads, that didn't change. Do you think these changes cause problems for implementing PacBio?

[msettles]

No, instead I think we could naturally extend the idea to something like Pacbio_subreads, so have single_end, paired end and subreads. And applying read type relevant processing per app depending on the type of read.

[joe-angell]

That sounds reasonable, I need to read up on how they work but can help with that if you want.