Inconsistent handling of vectors with unknown values about anytime

[eddelbuettel]

That's an R thing I can do nothing about. A vector is forced / required / defined to be of one type, and (AFAIK) the first value determines the type for the whole vector. anytime() and anydate() then receive that vector (as a whole) and do operations given the representation type of the vector.

I am fairly sure we discussed this here before so you may additional information if you search among the closed issues.

The other issue is that initially we took a numeric (as in your vec2) to be a 'transcribed' character (as here) and reconverted to char. Turns out that is a bad idea because we really need to be like R itself and treat 5 (or 5L) as the offset to the epoch so that anydate(5) gets us Jan 6, 1970. But because honour old behaviour and APIs we added an option oldHeuristic -- but apparently only for anytime:

> anytime(20120213, oldHeuristic=TRUE)      
 [1] "2012-02-13 CST"                                      
 > 

If you want to port that to anydate or shimmy something else in please follow up with a proposal. Otherwise I think we can close this. If you disagree, or have other material, please feel free to reopen. Thanks of trying anytime !

PS This works:

> anydate(anytime(20120213, oldHeuristic=TRUE))   
 [1] "2012-02-13"                                                          
 >  

PPS I was not quite correct on oldHeuristic. It is related to char <-> numeric but the values you have here don't fly, we get bad values. The overall issue remains: you may need to take care of leading NA values.