Features() -- should they accept lines or arrays?

[daler]

Features() -- should they accept lines or arrays?

[brentp]

makes sense. from initial sketching, i think taking an array may work better.
then the split is only done once.

one thing to note is that we should know which the classes/columns after we've done an intersection, so then, it might be easier to do:

bedtool(fn, feature_classes=this.feature_classes + other.feature_classes)

on internal operations, then it only needs to guess for the initial class.
it'll be a bit more complicated than that after accounting for the extra columns for overlap and such, but...

[brentp]

as a motivating example: https://gist.github.com/887065

that has an analysis i'd like to make easier. it's already somewhat facilitated by the new Feature sub-classes, but when i do an intersection and all i really want is the name and the distance, it's more work than it should be, and error prone--i have to guess and check the column offsets.