321/process nwb2 #334
321/process nwb2 #334
Conversation
…nit_name The PatchClampSeries units are hardcoded since 6b2a628f (Hardcode units of PatchClampSeries subclasses (#1036), 2019-08-05) [1] in pynwb. Therefore we must accept and translate these units as well. [1]: NeurodataWithoutBorders/pynwb@6b2a628
The getRawDataSourceType function was used to adapt the behaviour depending on the data source. But as we now don't do that anymore, and it currently breaks wiht MIES NWBv2 which don't have experiment_description set, let's just remove it.
Nowadays the version is written as NWB-2.2.0 so we need to expect that prefix as well.
Compare pipeline outputs from nwb2 against pipeline output from nwb1 file where both nwb1 and nwb2 were produced from the same pxp data
sgratiy
force-pushed
the
321/process-nwb2
branch
from
292176b to
115b1fb
Compare
sgratiy
force-pushed
the
321/process-nwb2
branch
from
115b1fb to
14801b4
Compare
|
If you are interested, I have hundreds of MIES generated NWB2 files at /allen/aibs/technology/waynew/ephys/output/Z/ephys/20200207_nwb2/ |
| self._nwb_data = val | ||
|
|
||
| def __init__(self): | ||
| """ This class does work on __init__, so this hack is used to inject |
There was a problem hiding this comment.
What do you mean by "class does not work on init"?
If you provide parameters like this MiesDataSet(nwb_file=NWB_file, ontology=ontology) with nwb file and ontology then it should work.
|
|
||
| def __init__(self): | ||
| """ This class does work on __init__, so this hack is used to inject | ||
| nwb_data & other dependencies. |
There was a problem hiding this comment.
If I understand correctly, you propose in the future to refactor DataSet classes to inject nwb_data set object and other dependencies. This is great for testing, of course, but when I pulled out the ontology from inside the DataSet classes, scientists told me that it was not convenient for them. They did not want to think about creating an ontology object and inject it when creating a DataSet. They want to have easy way of creating DataSet objects. I guess this can be done with some helper functions or alternative constructors.
| obt = ds.extract_sweep_record(sweep) | ||
|
|
||
| misses = [] | ||
| for key in exp.keys(): |
There was a problem hiding this comment.
dictdiffer.diff function works pretty well to compare dicts
|
|
||
| try: | ||
| value = d[key] | ||
| except KeyError: |
There was a problem hiding this comment.
d.get(key) is imo a much more succinct and explicit way to do this- get() returns none by default, or you can specify a default value, e.g. d.get(key, None). It is very slightly slower though, so if this is done a whole lot maybe just leave it.
ipfx/nwb_reader.py
Outdated
| @@ -111,17 +113,17 @@ def get_unit_name(stim_attrs): | |||
| def get_long_unit_name(unit): | |||
| if not unit: | |||
| return "Unknown" | |||
| if unit.startswith('A'): | |||
| elif unit.startswith('A') or unit == 'amps': | |||
There was a problem hiding this comment.
is this as specific of a check we need? startswith seems possibly ambigious...
There was a problem hiding this comment.
I specified unit names explicitly as you suggested. thanks.
Address #321