| Topic | Decision | Date | Related Jira Issues |
|---|
| SAS and R use different date epochs for integer-based date representations creating an interoperability problem. | - Standard: For Dataset-JSON, given that SAS and R use different date epochs to calculate integer dates, Dataset-JSON will represent dates as ISO 8601 datetimes. Additional metadata will be added to inform the conversion software tools to convert the dates to an integer where appropriate.
- Tools: The conversion software tools will manage converting dates to and from the ISO format.
- UG: Using ISO 8601 date formats in considered a JSON best practice.
| |
|
| When running the R conversion tool, if the display format attribute is missing in the Dataset-JSON dataset, the numeric date variable is not recognized as a date but as an integer. This suggests that display formats are not robust enough to identify numeric date variables. | - We need additional metadata to represent target data types.
|
|
|
| Precision mismatches seem to occur when comparing floating point values with many digits after the decimal. | |
|
|
| If ITEMGROUPDATASEQ is not part of the source or target datasets, can the conversion software handle this? | - UG, Tools: Software conversion tools add ITEMGROUPDATASEQ, and it does not have to be in the original or target datasets. The metadata describing the ITEMGROUPDATASEQ variable will always be the same so can be added by the conversion software.
|
|
|
| For XPT files, define.xml or OID metadata was never a requirement. For Dataset-JSON additional metadata requirements have been added and it was assumed that everyone was already aware of how to work with it. | - UG: Similar to Finding 11, we will create a User's Guide to highlight best practices for dealing with the needed metadata.
- Tools: We may also recommend that tools generate default OIDs and other values to accommodate cases where the complete metadata is not yet available (for discussion).
|
|
|
|
|
|
|
