srfi 18: time->seconds does not round to nearest second, returns exact

[mxork]

The implementation of time->seconds rounds the representation of the time? to the nearest second. The srfi seems pretty clear that this isn't the intended behavior:

(time->seconds time)                                  ;procedure
Converts the time object time into an exact or inexact real number representing the number of seconds elapsed since some implementation dependent reference point.

    (time->seconds (current-time))  ==>  955039784.928075   ;; <------

It also makes the srfi's interface unusable for timeouts <1s, which is goofy.

My change does the obvious thing. It uses an exact (rational) representation for microseconds, because I think that (seconds->time (time->seconds T)) should equal T. That said, using an inexact? to match current-second from (scheme time) seems tasteful as well.

[ashinn]

Thanks! Unless there is precedent for using exact rationals, I think it's better to use inexact here for better compatibility. Chibi can also be compiled without rational support.

[mxork]

Inexact does seem like the right thing, doesn't it. Patched, and this output looks sensible:

(import (scheme small) (srfi 18))

(define now (car (current-time)))
(define rt (seconds->time (time->seconds now)))

(display now)
(newline)
(display rt)

; output:
; #<timeval 140171138348800>
; #<timeval 140171138352096>