Closed
Description
runLoggerLoggingT flushes at the end of the action. As of Blammo-1.1.2.3, this means that it incurs a 0.1-second thread delay. This means that, for example, if you have an app setup that looks like this:
loadApp f = checkpointCallStack $ do
appSettings@AppSettings {..} <- loadAppSettings
initialFaktorySettings <- Faktory.envSettings
appLogger <- newLogger appLogSettings
appAwsEnv <- runLoggerLoggingT appLogger AWS.envDiscover
appSqlPool <- runLoggerLoggingT appLogger makePostgresPool
_...... This seems like a reasonable thing to write, but it incurs 0.2 seconds of unnecessary pausing. It seems that runLoggerLoggingT carries the unstated assumption that it will only be run once, at the outermost layer of an application, and this does not seem to hold true in the wild. Blammo-1.1.2.3 would be an unacceptable performance regression for us and probably for others, and I think we ought to deprecate it.
cc @pbrisbin
Metadata
Metadata
Assignees
Labels
No labels