Standardize support for dates / times / time zones #1191
Conversation
… as standardized in dhall-lang/dhall-lang#1191
|
Here is the matching change to the Haskell implementation: dhall-lang/dhall-haskell#2247 |
The two test files were backwards
… based on feedback from @mmhat and @TristanCacqueray
… based on a discussion with @mmhat
|
Great work! I appreciate the thoroughness of the tests |
… as suggested by @Nadrieril
… as suggested by @Nadrieril
… as suggested by @mmhat Specifically, we don't permit a standalone `Z`, but we permit `Z` as a suffix for a `Time` or `Date`+`Time`
It turns out that 20 digits of precision doesn't necessarily fit into an `Int` :)
|
What good does this without the ability for full timestamps? As the text notes, leap seconds aren’t even supported, so you cannot have a (valid!) timestamp that goes to In addition, just having the parts of a timestamp (dates, times, timezone offsets) doesn’t mean it is clear how to construct a canonical date from it. |
|
Okay, I misread it slightly, it looks like I can have a “normal” timestamp in the surface language, but then it is interpreted as a record that contains the time, date and timezone. Which is a bit … weird, but okay. My biggest problem with this is that time/date logic is not as simple as having a record with three fields, you need a bunch of operations that can convert opaque types into each other (e.g. UTC into UT1 into UNIX timestamps). |
Fixes #80