Open yarikoptic opened 19 hours ago
Yes, n/a
applies to all columns, and that is how the validator has handled it the whole time. Proposing nan
or another alternative for numeric columns would not change the need for tools to work with n/a
in historical datasets. I'm okay with the suggested reordering, if that clarifies things.
Describe your problem in detail.
ATM specification for Tabular files https://bids-specification.readthedocs.io/en/stable/common-principles.html#tabular-files states
So, in the best reading of it, it mandates use of explicit
n/a
for a missing value in any (not only "String values" column) column. Asn/a
is not a standard placeholder, that unnecessarily complicates loading of such files using anything which expects numeric values for the column (e.g.onset
).Describe what you expected.
I have not investigated this further yet and do not have any specific recommendation ATM (e.g. after looking how pandas would expect to have float.nan to be defined in tsv etc). Just raising a possible discussion point.
At least we might want to reorder sentences to remove possible misassociation with string only columns, i.e. to have it
BIDS specification section
https://bids-specification.readthedocs.io/en/latest/...