Instead of storing the information about all gas gain slices as attributes, which for long runs can accumulate to several hundred attributes (which is extremely slow to overwrite, because deleting attributes in large files with many attributes is ridiculously slow), we should store it as a dateset.
To not end up with many datasets, we can now use compound datatypes in nimhdf5. Just have a dataset with seq[GasGainIntervalData] as a compound dataset, where each row corresponds to one slice. Easy deletion and overwriting!
Vindaar
commented
9 months ago
Instead of storing the information about all gas gain slices as attributes, which for long runs can accumulate to several hundred attributes (which is extremely slow to overwrite, because deleting attributes in large files with many attributes is ridiculously slow), we should store it as a dateset.
To not end up with many datasets, we can now use compound datatypes in
nimhdf5. Just have a dataset withseq[GasGainIntervalData]as a compound dataset, where each row corresponds to one slice. Easy deletion and overwriting!