ageorgou
commented
4 years ago Unfortunately, it seems that we can't specify the compression when we write the file, but rather we need to do it when creating each Dataset.
However, since most of our data is acquisitions, we can change NWBFile::add_time_series_data to use a wrapped Dataset, as in the link above. I think that should take care of the bulk of changes.
Current size differences between generated NWB1 and NWB2 files that we write using pynwb are (at least to a large part) explained by the fact that PySilverLabNWB does not do any compression, while the NWB1 data is compressed with GZIP level 4.
https://pynwb.readthedocs.io/en/latest/tutorials/general/advanced_hdf5_io.html#compression-and-other-i-o-filters explains how pynwb can be used to specify compression of a dataset.