Closed christophsax closed 3 years ago
For spatial classes, we include language asking for documentation of workflows for multiple spatial classes/data types. Is it overkill to request that that TS packages have this, (e.g., vignettes and examples include conversion from different TS classes?)
My feeling would be that that is not necessary. Conversion between spatial classes and/or types is hardly ever trivial, and an understanding of at least how that can be achieved in a given context is very generally helpful. Time series conversion is as simple as tsbox
, which i think effectively obviates any need to actually document the process. Thoughts @christophsax?
I don't have a clear opinion on this. Requiring documentation on this may be an overkill indeed.
On the other hand, I've seen many users of time series software who struggle with basic "ts"
objects. They read data from a tabular excel, then they access the column vectors of the resulting data frame and convert vectors manually into "ts"
objects. This sometimes still happens after I point them to tsbox.
So an example / documentation of how to get from tabular raw data to a specific time series class of TS package may be certainly useful for starters. Perhaps a dedicated vignette in tsbox could also help here.
The PR is based on an email exchange with @mpadge. Thanks for this! These are very good points on the use of time series.
I am not 100% persuaded why you want to ask for as many time series classes as possible.
Yes, you get it cheaply if you use
tsbox::ts_<some_class>
at the beginning. But I imagine some packages trying to achieve this on their own, and writing their own converters. This takes them down the rabbit hole. I would prefer their function to consistently use one particular time series class that fits their needs, rather than using some shaky converters. Consistency is more important than flexibility.