RUST-1571 Update ServerId to be int64 #894
Conversation
| @@ -165,7 +171,8 @@ impl Connection { | |||
| pub(crate) fn info(&self) -> ConnectionInfo { | |||
| ConnectionInfo { | |||
| id: self.id, | |||
| server_id: self.server_id, | |||
There was a problem hiding this comment.
I'm of two minds here whether truncation or omission is better when the server_id doesn't fit in an i32. @isabelatkinson do you have an opinion here?
There was a problem hiding this comment.
I think it would be a breaking/unexpected change for users who are currently relying on this field to be Some (even if it's a truncated value) to stop populating it.
There was a problem hiding this comment.
I agree with Isabel. The i64 overflow issue affects a small amount of users and would bother more users if we just removed support for the original server_id without warning.
src/cmap/conn/mod.rs
Outdated
| @@ -46,9 +46,15 @@ pub struct ConnectionInfo { | |||
| pub id: u32, | |||
|
|
|||
| /// A server-generated identifier that uniquely identifies the connection. Available on server | |||
| /// versions 4.2+. This may be used to correlate driver connections with server logs. | |||
| /// versions 4.2+. This may be used to correlate driver connections with server logs. This | |||
| /// could be a truncated value based on input server_id. | |||
There was a problem hiding this comment.
Nit: how about "If the connection ID sent by the server is too large for an i32, this will be a truncated value." (or "will be None." if we go the other way) for wording? I think it's useful for our users when we can be more precise about the behavior.
Co-authored-by: Isabel Atkinson <isabelatkinson@gmail.com>
ServerId is assigned to ConnectionId here: https://github.com/mongodb/mongo-rust-driver/blob/main/src/cmap/establish/handshake/mod.rs#L457. To fix the int32 overflow issue, I added a new ServerId int64 value.