Edit this page

NWB Workshops and Hackathons

:rewind: Back to the projects list

Currently edited online in https://hackmd.io/z6GJrRSdRySMo1LAppabJw

Validation and Introspection

Session Chair: Yaroslav Halchenko (Dartmouth College) and Thomas Braun (byte physics e.K.)

Needs, requirements and expectations for validation and introspection of NWB files

Participants

Objectives

Different “levels”/types of validation

NWB Schema validation (currently implemented)

In principle, any stored by PyNWB/MatNWB file should pass this stage of validation

User input “best practices”

Validate toward “best practices”. nwb-inspector ATM is the implementation for some of such:

“Ontologies” for the fields/values

Some of those could be addressed by more restricted data types, e.g. ISO datetime duration for “age” or any other “duration”.

Data Validation

Catch obviously (to human curator) eye incorrectly entered/encoded data (e.g. incorrect unit, e.g. V vs mV). ATM not possible to encode range within NWB schema, to e.g. encode the range.

Question of the units: SI with multipliers. Satra mentioned https://people.csail.mit.edu/jaffer/MIXF/MIXF-10 which is character string encoding for numerical values and units, incorporates SI, etc.

Proposed discussion topics

Participants should add topics of possible interest for discussion here

Approach and Plan

Introducing “ontologies” while retaining compatibility

(Ben) Ontologies could not be introduced for a field without breaking compatibility (i.e. forcing more than free form string) by adding a supplementary “Ontologies” table with fields

Current state

Validation

pynwb provides validation of .nwb files to follow specific schema. Could also be invoked from command line via python -m pynwb.validate [FILES] but at best should be invoked per each file separately since the whole process would fail as soon as the current file fails validation.

dandi validate (and also during upload) provide a way to validate files (delegates actual validation to pynwb) from command line:

$> dandi validate v2.0.1/test_*
...
v2.0.1/test_CurrentClampStimulusSeries.nwb: ok
v2.0.1/test_DecompositionSeries.nwb: 1 error(s)
  DynamicTable/colnames (processing/test_mod/LFPSpectralAnalysis/bands.colnames): incorrect type - expected 'text', got 'ascii'
v2.0.1/test_Device.nwb: ok
v2.0.1/test_DynamicTable.nwb: 1 error(s)
  Units/colnames (units.colnames): incorrect type - expected 'text', got 'ascii'
v2.0.1/test_ElectricalSeries.nwb: 1 error(s)
  DynamicTable/colnames (general/extracellular_ephys/electrodes.colnames): incorrect type - expected 'text', got 'ascii'
v2.0.1/test_ElectrodeGroup.nwb: ok
...
Summary: Validation errors in 14 out of 31 files

Introspection

dandi ls provides easy way to see basic information about any .nwb file. For pre-v2 nwb files it would at least provide NWB version information. E.g.:

$> dandi ls v2.0.1/test_Cluster*
PATH                             SIZE    IDENTIFIER            SESSION_DESCRIPTION                                   SESSION_START_TIME   NWB  ND_TYPES
v2.0.1/test_Clustering.nwb       17.8 kB TEST_Clustering       a file to test writing and reading a Clustering       1971-01-01/12:00:00  2.0b Clustering
v2.0.1/test_ClusterWaveforms.nwb 20.0 kB TEST_ClusterWaveforms a file to test writing and reading a ClusterWaveforms 1971-01-01/12:00:00  2.0b ClusterWaveforms, Clustering (2)
Summary:                         37.7 kB                                                                             1971-01-01/12:00:00>
                                                                                                                     1971-01-01/12:00:00<

Progress and Next Steps

Materials

Background and References