Support the `interval` type when appropriate Rust types are developed

[mikedilger]

Is there a reason { name: date, oid: 1082 } is not included? I realize a timestamp is more precise, but (at the very least) existing databases may use date.

I tried to add support for it at https://github.com/mikedilger/rust-postgres/commit/7f4637956f65c9086705c444ba93d30c4f593177 but I'm not sure that's right. When I try using it, it triggers a RefCell already borrowed error (#34 or rust-lang/rust#13246)

[sfackler]

I believe I left it off because it doesn't match up 1:1 with an existing Rust type. Timespec is a significantly higher precision type, and I'm a bit leery of silently lossy conversions in the Rust -> Postgres direction. That being said, mapping it to Timespec isn't the end of the world, though I wouldn't be surprised if date had a different wire format than timestamp, which would require some changes to the ToSql and FromSql implementations.

[sfackler]

Another example is the interval Postgres type. I initially thought that it would be a straightforward 1-1 mappinng with the std::time::Duration Rust type. However, interval stores a number of microseconds and separately a number of days and months. The days and months aren't converted into microseconds until the interval is matched with a time, as the length of a day or month will vary. We can easily implement ToSql for Duration, but any FromSql implementation would be very wrong.

[mikedilger]

I solved my immediate issue by using timestamp in the database.

I agree Timespec is the wrong rust type to map Postgres date to.

Yes, there is no Rust type that maps to Postgres interval yet. Someone was supposed to be working on a Jodatime-like / JSR-310 datetime system. Hopefully that will have a ReadablePeriod type, which maps to Postgres interval.

[cbruun]

Is rust-chrono a good choice?

[sfackler]

I've added support mapping TIMESTAMP WITH TIME ZONE to DateTime<UTC>, TIMESTAMP to NaiveDateTime, DATE to NaiveDate, and TIME to NaiveTime under the chrono feature. There's still no equivalent for INTERVAL as far as I know.

[TheServerAsterisk]

@sfackler would there be interest if I wrote a "simple" mapping for the interval type? More specifically it would have the following interface and other additions if required:

pub fn new(microseconds: i64, days: i32, months: i32) -> Interval;
pub fn microseconds(&self) -> i64;
pub fn days(&self) -> i32;
pub fn months(&self) -> i32;
pub fn to_sql_standard(&self) -> String; 
pub fn to_postgres(&self) -> String; 
pub fn to_postgres_verbose(&self) -> String; 
pub fn to_iso_8601(&self) -> String; 

and will also implement the Clone, Copy, PartialEq, Eq, Debug, Add, and Sub traits.

Any input would be appreciated.

[voronaam]

This affects me as well. My current workaround is

conn.execute("DELETE FROM test WHERE ts < now() - $1::text::interval",
                    &[&format!("{}ms", retention)]);

The ::text part is required for the library to map the type and it is later converted into an interval on the postgres side.

The @TheServerAsterisk suggestion makes sense to me. Though I'd suggest having a new variant which takes in std::time::Duration. But this is not important.

[piperRyan]

I have a dedicated interval type here if anyone is interested.

[jcaesar]

How foolish is it to wish for std::time::Duration to be converted to intervals?

[sfackler]

A Duration could be converted to a Postgres interval, but not vice versa: #60 (comment).

[sanpii]

There's still no equivalent for INTERVAL as far as I know.

chrono::Duration?