Millisecond information for `java.sql.Time` is lost when converting to/from `LocalTime`

[jeff303]

Consider the following REPL session, in which version 0.3.2 of the library is being used (as per the deps.edn guide here).

(import java.time.temporal.ChronoField)
(import java.time.temporal.Temporal)
(def my-local-time (t/local-time 1 12 13 456000000))
(t/sql-time my-local-time)
#inst "1970-01-01T07:12:13.000-00:00"

Notice that the millisecond portion (456) is lost. In principle, though, it is possible to build the java.sql.Time instance to have that information:

(let [millis-of-day (.get my-local-time ChronoField/MILLI_OF_DAY)]
  (java.sql.Time. millis-of-day))
#inst "1970-01-01T01:12:13.456-00:00"

The same problem occurs in reverse as well:

(def my-sql-time 
  (let [millis-of-day (.get my-local-time ChronoField/MILLI_OF_DAY)]
    (java.sql.Time. millis-of-day)))
(t/local-time my-sql-time)
#object[java.time.LocalTime 0xda22aa "19:12:13"]

Again, it's possible to preserve that information (probably more elegantly than this):

(import java.time.LocalTime)
(LocalTime/ofNanoOfDay (* 1000000 (.getTime my-sql-time)))
#object[java.time.LocalTime 0x10d98940 "01:12:13.456"]

[frenchy64]

This happens due to the implementation of java.sql.Time/valueOf:

https://github.com/openjdk/jdk/blob/eeb345a286115213f8a75dfe54cdcc39dfca597a/src/java.sql/share/classes/java/sql/Time.java#L252-L255