When performing a PSE the variance metrics are stored in separate folders and H5 files.
If we'd like to reproduce the PSE figures we have to read all the values from different files.
We will want to easily load them through the command line (command or library profile).
Also because we are storing data separately, when we are generating the PSE figures, we are actually going on the disk and opening multiple H5 files to get the colours.
We ended up in this situation, because we are treating DataTypeMeasure as any other DT in TVB, and it is produced as a result of an analysis, which comes as a separate step in the workflows.
This storage was a side effect that we like: Discrete PSE gets filled at runtime.
- have a reader method that does exactly this in command line
- write the DT-Measures in DB (and possibly on TimeSeries as metadata)
- find a time when the ops are finished to compute the Measure when the group is over and save it at once in a single H5 (but then we will no longer have the PSE filling at runtime)
- store DT-Measures as metadata on TS and find a place to only gather them in a single H5 at the end (this would mean PSE has a double way of reading data: when ops are completed all or not)