[R-sig-Geo] spacetime and sf

[Facundo Muñoz]

Thank you Edzer, I understand the trade-offs.

I hadn't thought of fixing the variable name for the time attribute. But
since time can be eventually represented as either dates, or week
numbers, or day numbers, or months, etc. I think it is best to let the
variable adopt whatever name the user feels more natural.

On the other hand, I don't expect my users to perform spatial operations
with their point pattern (other than perhaps filtering observations). So
it might not be that much effort to maintain the appropriate methods for
a specific class inheriting from `sf`.

Best wishes

        ƒacu.-


On 6/4/19 6:02 PM, Edzer Pebesma wrote:
> Dear ƒacu,
>
> spacetime will remain the way it is, just like sp, too much stuff
> depends on how it is now.
>
> For STFDF objects in spacetime, a follow-up package is stars: it
> supports raster and vector data cubes, and can deal with more cube
> dimensions than STFDF ever could; it also gets better dealing with e.g.
> large satellite images.
>
> For the STIDF (irregular: s/t point patterns) there is currently no
> follow-up package that builds on top of sf. As you note, its
> functionality would be pretty much limited to registering which column
> denotes time in an attribute, and add a class label. So far, I haven't
> considered that worth the effort: you may have to add lots of method
> instances to have that information survive all the current methods
> available for sf objects.
>
> Have you thought about giving the time variable a fixed name? Or is
> there a need for maintaining multiple time variables in sf objects?
>
> Best,
>
> On 6/4/19 5:01 PM, Facundo Muñoz wrote:
>> Dear Edzer,
>>
>> How do you conceive the future of the package `spacetime` and more
>> generally, the representation of spatio-temporal data, in light of the
>> transition from `sp` to `sf`?
>>
>> For context, I'm working on a package that handles spatio-temporal point
>> patterns. For my analyses I have relied on `sf` with some additional
>> attribute for the dates. However, for packaging I think it is safer to
>> rely on explicit classes for spatio-temporal data (so I don't have to
>> worry all the time about which is the temporal variable, in which
>> format, etc.). But it seems that only `spacetime` provides such classes
>> generically, which in turn relies on `sp` for the spatial dimension. Is
>> this going to continue to be supported or are there other plans?
>>
>> Thanks in advance
>>
>>         ƒacu.-
>>
>
> _______________________________________________
> R-sig-Geo mailing list
> R-sig-Geo using r-project.org
> https://stat.ethz.ch/mailman/listinfo/r-sig-geo