Re: encode-time vs decode-time

[Stefan Monnier]

>> So maybe we should stop documenting "decoded time" objects as being of
>> the form (SEC MINUTE HOUR DAY MONTH YEAR DOW DST UTCOFF SUBSEC), but
>> instead say that it's a `decoded-time` object on which you can use
>> the `decoded-time-*` accessors (and provide a corresponding pcase pattern
>> to replace the `(,sec ,minute ...) one).
>
> That sounds like a reasonable way to move forward.
>
> I'm not a pcase expert. How would one go about doing the pcase stuff? That
> is, how does one provide a pattern so that code like (pcase time
> ((decoded-time :year y :month m) (some-expr-involving y & m))) can extract
> the year and month components Y and M from the broken-down time TIME that
> was created via (make-decoded-time :year y :month m)? Is there an example of
> doing this sort of thing somewhere?

With the cl-defstruct we use for decoded-time, we can already do:

    (pcase-let (((cl-struct decoded-time (second s) (minute m) hour day) time))
      ...)

so we could define

    (pcase-defmacro decoded-time (&rest fields)
      `(cl-struct decoded-time ,@fields))

OTOH it doesn't work in plain `pcase` currently because of a bug in the
`cl-struct` pcase pattern (it expands to code which does (cl-typep XX
'decoded-time) but the way `decoded-time` is defined makes it unusable
with `cl-typep`).


        Stefan