These are chat archives for frictionlessdata/chat

14th
Sep 2018
Zane Selvans
@zaneselvans
Sep 14 2018 11:09
The online datapackage viewer/creator here: http://create.frictionlessdata.io/ Seems to be failing to validate a package with a tabular data resource that explicitly lists the empty string ("") as one of the possible missingValues giving the error Invalid type: undefined (expected string) -- in the preview sidebar looking at missingValues it shows up as null Is this the expected behavior?
Zane Selvans
@zaneselvans
Sep 14 2018 13:58
I've finally gotten a data package constructed with a bunch of rich metadata and validated it locally using data validate but when I data push validation fails at datahub.io for some reason. It looks like it's not treating it as a tabular-data-package.
Zane Selvans
@zaneselvans
Sep 14 2018 14:13
Forcefully setting "profile": "data-package" rather than tabular-data-package (even though the latter is what's correctly inferred) allows it to get further, but now it fails on creating views, with a bunch of errors of the variety
ERROR :Failed to cast row: Field "PORTABLE_OPERATION" can't cast value "False" for type "boolean" with format "default"
Zane Selvans
@zaneselvans
Sep 14 2018 15:26
Seems like it's trying to validate/cast an internal post-conversion datapackage using the specified allowable boolean values from the originally uploaded datapackage, rather than converting those to the new normalized schema instead. datahq/datahub-qa#239
But, did finally manage to get something pushed! https://datahub.io/zaneselvans/pudl-msha
My gosh did it swell up though. The original data was about 29MB uncompressed. Compressed it's about 6MB. But the CSV version on DataHub is 56MB. The JSON version is 173MB, and even the zipped version containing both is 21MB.