Better way to handle NaT (not-a-time) #3414
Description
Description of the problem
Both Python numpy and Matlab can have a datetime called NaT, i.e., not-a-time. It's similar to NaN but for time. Not sure if Julia also has something similar.
GMT can't handle NaT. It's OK for command line, but not for external interfaces like PyGMT.
Currently, when a NaT string is passed to GMT, it reports an error and keeps running.
Unable to parse 1 datetime strings (ISO datetime format required)
We should at least change the error to warning.
What's worse is:
Lines 12769 to 12771 in 3003c50
GMT converts the datetime strings as double values relative to the epoch time internally. In the codes above, if
dt[row]="NaT", the function gmt_scanf() returns GMT_IS_NAN and the variable t_vector[row] is untouched. The value of t_vector[row] is 0, which means 1970-01-01T12:00:00. The plot is incorrect if we're plotting a calendar plot around 1970-01-01.
Ping @PaulWessel, @joa-quim and @weiji14.