Closed Vindaar closed 3 years ago
Commits https://github.com/Vindaar/TimepixAnalysis/commit/6cae2199494f6e6b6d9b4ed9a38b2c67e0b5ba9a and https://github.com/Vindaar/TimepixAnalysis/commit/1d296a4f684121ef4485ba16abc7a24872dbb7a3 have basically implemented this. Only the 1D datasets like eventNumber, etc. are stored as int64
still. It's a little problematic to choose smaller values for some of the datasets, if we
Calculated properties are all still stored as float64
, but it's questionable whether that should be changed.
Closing as current solution is fine.
Partly trivial, partly quite a bit of work. Currently we store all resulting data in
reconstruction.nim
as 64-bit integers and floats. This wastes a huge amount of space, because for most datasets these ranges are never used.Rewriting itself is trivial, but breaks compatibility with existing files and might introduce more bloat when dealing with the datasets, because we cannot combine actions based on
int
andfloat
type, but need to differentiate more.