ToDo

[nevrome]

• Check if an input date is in the range of the calCurve to reject it with a meaningful error message
• Add an option to allow for calibration of dates even outside of the range of the calibration curve
• Provide an HDR output that is better machine-readable (-q is fine for humans, but --hdrFile should get a more formal structure)
• Maybe: Calculate median ages as a possible output option (started to work on simpler summary output (start-median-end) #11)
• Add median age and simplified ranges to the cli plot axis
• Allow for user-provided calCurves
• Store calcurves in a more clever way (like this?)
• Sum calibration
• Normal distribution vs. Student's t-distribution (both could be provided with a switch)
• Resolution option to allow for faster calibration on a lower output resolution. One setting might be to skip calCurve interpolation completely
• Generally better error handling for the CLI
• Document the library
• Understand the Bchron algorithm and copy it, if it gives a performance advantage
• Maybe the matrix multiplication does not have to be performed explicitly, but can be done "one the fly" when the calCurve matrix is constructed. That's less didactically valuable, but could be faster
• Discuss potential changes for performance gains with more experienced Haskell developers
• A conda release might make the software more attractive for the bioinf crowd
• Maybe further performance improvements are possible with even better array processing libraries like accelerate or repa
• Add unit tests
• Reimplement calcurve interpolation to make it clearer, simpler and faster
• A simple command line bar chart of the calibrated distribution would be a nice gimmick
• Submit to hackage
• Reexport the important functions in a dedicated, user-facing module
• Settle on a type and function interface for the calibration that is clear, easy, short, versatile and extensible

Reminder: If full strictness becomes an issue at some point, then there is an easy way out

[zypeh]

suggestion: you may try using Z-Data for the vector manipulation. It also include parser utility too, just not sure if the parsing phrase is the bottleneck.

[nevrome]

Thanks for the suggestion @zypeh - there are a couple of good vector and matrix libraries I could try, but I fear I don't have the time to experiment extensively. Z-Data looks pretty well maintained, though.