Fix unified_log handling of timestamp formats
#8451
Conversation
The log entries timestamp is a double, we need to ensure that it's cast to an integer when returning it as a result to sqlite and using a timestamp constraint, otherwise sqlite will filter out all the entries that have the decimal part in them.
|
|
||
| r["timestamp"] = BIGINT([[entry date] timeIntervalSince1970]); | ||
| r["timestamp"] = BIGINT( | ||
| static_cast<std::uint32_t>([[entry date] timeIntervalSince1970])); |
There was a problem hiding this comment.
In the spirit of not losing information, do you think it makes sense to also pass a float through? (timestamp_float or something)
There was a problem hiding this comment.
Sure (I would call it timestamp_double to respect the true type).
|
There's an additional problem I did not consider. When writing a constraint like The solution here is to actually pass the timestamp in the constraint + 1, if the constraint is using |
|
This should do, although there are still some interesting questions around the |
unified_log handling of timestamp formats
The log entries timestamp is a double, we need to ensure that it's cast to an integer when returning it as a result to sqlite and using a timestamp constraint, otherwise sqlite will filter out all the entries that have the decimal part in them.