Dear KNIME team,
I love the new Date&Time implementation in KNIME -- especially the distinction in multiple types.
However, I noticed some things that used to work in the legacy nodes but don't in the new ones:
- Date types accepted as x-axis column in Line Chart (JFreeChart)
- calendar week pattern ww.yyyy recognizable in String to Date&Time node
The attached workflow compares these issues. It would be great to close this gap in the future so that one day we can do without the legacy nodes.
There are a couple of node that don’t use the new date&time types yet. We decided to first change the most frequently used nodes and will continue with the others.
As for your second problem, in my opinion the new behaviour is more correct than the old one. If you only have a week and a year it’s not possible to convert this into a date without “inventing” data. Therefore the new nodes refuse the conversion (or don’t accept the pattern).
I guess previously the first day of the week was added? Is this the behaviour you would expect?
Thanks for your comments. Glad to hear that more nodes will be supported with the new date&time types.
Regarding the calendar week conversion: I see your point, but I think “inventing” data occurs at many conversions. For example, if you convert a date to a date&time, the 00:00:00 time is also invented. The same logic can be applied when converting calendar weeks to dates (you are right, I would expect the Monday of that week to be returned – the legacy node behaves this way).
In analytics on industrial datasets, calendar weeks are often used (especially in data coming from a Business Warehouse). It would be a pitty if we need to teach our users to still use the legacy nodes for a long time instead of the in general much better new nodes.
But in other countries the week starts with Sunday… We will have a thought about it, how to best handle such cases.