On 1/01/17 10:42 PM, Yitzchak Gale wrote:
I have heard that the Smalltalk time library got the semantics of time right long before even the Haskell time library, but I have not seen it myself in detail.
What do you mean by "the" Smalltalk library? The classic Smalltalk-80 library included Date and Time classes for calendar calculations and a Delay class for suspending. The Date and Time classes were not sophisticated. The ANSI Smalltalk standard includes a unified DateAndTime class and a Duration class. When I tried to implement the standard class I found it impossible, because the standard requires both that DateAndTime be in UTC and that arithmetic work for an unspecified range of timepoints, apparently requiring you to predict leap seconds indefinitely far into the future. Despite requiring UTC, it does not let you specify second=60 when creating a timestamp. You can specify the *offset* of local time from UTC, but implementations are allowed to impose a limit or -12 hours to +12 hours, which excludes part of my country. Worse, you specify an *offset*, not a *zone*, yet an implementation is required to intuit, possibly by some form of clairvoyance, the *zone* name relative to a time. Since offsets do not uniquely determine zones, this is of course impossible. Frankly, it's a mess that looks like it was designed by someone who thought they were expert but weren't. Of course actual Smalltalk implementations don't pay all that much attention to the standard. Possibly "the" Smalltalk library is the Chrono package, in which case ignore my rant about the standard.