Ecto / Postgres timestamp before ‘year 0’

I’m currently trying to store dates in Postgresql that are before ISO8601 ‘year 0’ and am running into issues. Postgresql stores dates internally as JD and can correctly insert/remove dates. I’m storing the date generally as NaiveDateTime and it works fine for anything > year 0.

For instance, the following query works fine in SQL.

sql> insert into table1 set t = '1000-01-01BC 00:00:00'; 

However, trying to load this in Ecto doesn’t seem to work as we’re referring to a calendar function that doesn’t have any representation of times before 0.

    (stdlib 3.13) calendar.erl:257: :calendar.last_day_of_the_month/2
    (stdlib 3.13) calendar.erl:133: :calendar.date_to_gregorian_days/3
    (stdlib 3.13) calendar.erl:155: :calendar.datetime_to_gregorian_seconds/1
    (postgrex 0.15.8) lib/postgrex/extensions/timestamp.ex:48: Postgrex.Extensions.Timestamp.encode_elixir/1
    (postgrex 0.15.8) lib/postgrex/type_module.ex:897: Postgrex.DefaultTypes.encode_params/3
    (postgrex 0.15.8) lib/postgrex/query.ex:75: DBConnection.Query.Postgrex.Query.encode/3

-spec datetime_to_gregorian_seconds(DateTime) -> Seconds when
      DateTime :: datetime(),
      Seconds :: non_neg_integer().
datetime_to_gregorian_seconds({Date, Time}) ->
    ?SECONDS_PER_DAY*date_to_gregorian_days(Date) +
        time_to_seconds(Time).

What’s the appropriate way for dealing with this? As far as I know, erlang/elixir doesn’t have any methodology for monkey-patching, the only thing I could think of is to store these dates in an alternative format as a workaround. Any thoughts?

I think thats a bug in Postgrex (I will open an issue and a PR) since t:NaiveDateTime supports negative years I think its reasonable that Postgrex should.

As an alternative I would suggest implementing your own Ecto custom type (if you’re using Ecto) so you can do your own encoding.

EDIT: This issue is already on the Postgrex issue tracker

2 Likes

Awesome. I looked through the code and I patched the library in this branch which could serve as a starting point for a PR. It seems to be working (at least for me).

It appears to me only two files needed modification, but I suppose the patch could be bigger to cleanup all instances of :calendar. And I have no idea what Jose is talking about “internal rata die” representation, so I’m just going to use the branch for now and watch and learn.