There is a new package, ivs that implements interval range algebra, that works on generic R types, and should work fine with non-integers.
There would be some awesome advantages to using this over IRanges.
- IRanges has a very custom data structure, and we have to add all of our data to be metadata columns of the IRanges object, and switch between them often.
- ivs instead can store the intervals as part of a tibble or data.frame
- The m/z and frequency values are not naturally integers, so we have to multiply and round out to transform them to be integers for the ranges.
- I think ivs will be able to use the frequency values naturally, without any transformation to integers.
So as long as using ivs isn't any slower than IRanges, it would probably make the code easier to maintain and reason over.
I propose we should at least test this using the noise exclusion procedure. We can evaluate the code differences, and check for any changes in execution speed in a limited part of the codebase.
There is a new package,
ivsthat implements interval range algebra, that works on generic R types, and should work fine with non-integers.There would be some awesome advantages to using this over
IRanges.So as long as using ivs isn't any slower than IRanges, it would probably make the code easier to maintain and reason over.
I propose we should at least test this using the noise exclusion procedure. We can evaluate the code differences, and check for any changes in execution speed in a limited part of the codebase.