Open Bjwebb opened 5 years ago
This is an issue for OCDS, as (1) not all arrays of objects have identifiers and (2) identifiers are not always required in that context (and might be missing anyhow).
The Government Transparency Institute ran into issues when OCDS data was missing identifiers (e.g. a publisher omits a required id
field). Their experience was that Flatten Tool would mint new identifiers, without any feedback to the user, which made analysis more difficult.
I haven't verified that Flatten Tool has this behavior, but if so, there should be feedback to the user.
Thanks for the feedback, @jpmckinney . Between your previous comment and this one, I think this is something that should be given some attention relatively soon. I'll add something to I&R.
When nested identifiers are missing in the source data, then data is pushed onto sheets without enough information about what parent it related to. (See example about sectors within transactions below). Notably it is not possible to unflatten the data back into the original shape.
This is particularly a problem for IATI, because the standard doesn't specify such identifiers for all nested objects (whereas OCDS does).
https://github.com/OpenDataServices/flatten-tool/issues/178 is the same issue, but specifically for multilingual narratives.
In https://github.com/OpenDataServices/flatten-tool/issues/177#issuecomment-474921498 @stevieflow wrote: