New Redis & HTTP handlers

[hartfordfive]

@joeybloggs I've added these two new handlers. Please note this is an initial version and I'd like to improve them both in the near future by adding a feature that allows multiple items to be buffered and sent in a MULTI command to Redis, and allow the option for multiple events to be buffered and sent at once for the HTTP handler.

I haven't added any tests for the HTTP handler as I'm not completely certain how to test for that yet. Would it be reasonable to have spawn a local HTTP server listening for events for this test?

I also plan on adding a file handler, which allows you to log events to a specified file. Let me know if this code is reasonable, and let me know if you have any suggestions for improvements/modifications!

[deankarn]

Thanks @hartfordfive!!

hoping to have a look later tonight.

as for testing the http handler, I think it's totally reasonable to spawn a local http server instance using go's httptest package https://golang.org/pkg/net/http/httptest/

[deankarn]

Wow you did a ton of work! @hartfordfive

there are a few things I would change, but all very minor. like:

• in examples use log.Debug(...) and log.Info(...) instead of manually creating Entry and calling HandleEntry(...)
• (this ones my fault) changed from e.WG.Done() to e.Consumed() hiding the backend sync mechanism and allowing for changes without breaking compatibility.
• perhaps instead of calling fmt.Println(...) on error maybe just call log.Error() that way it will get logged somewhere, like another registered logger ( console logger perhaps )
• bufferSize is a uint and so will always be >= 0 so can just directly assign to new struct.

    if bufferSize >= 1 {
        h.buffer = bufferSize
    }

Question

Because handlers work off of a channel, many workers could be created to handle logs; do you think an option should be added to specify a number of workers?

I will help further as the week progresses

[hartfordfive]

@joeybloggs I'll make all those small suggested modifications. As for the number of workers, I totally agree and I'll add that also. One other potential modification would be to add the "runtime. GOMAXPROCS()" function to your log.go file to allow the use for one or more CPU cores? This would definitely help with message throughput when it comes to the HTTP or Redis handler.

[deankarn]

Thanks @hartfordfive

As for runtime.GOMAXPROCS() definitly shouldn't play around with that setting in a library. As of Go 1.5 it is by default set to the number CPU Cores in the system.

Morover this is a process wide setting and would affect the entire running application.

[hartfordfive]

@joeybloggs Good point on that. As for the tests, I'll have those finished hopefully within the next few days.

[deankarn]

No rush and thanks again

[hartfordfive]

@joeybloggs Regarding the Redis handler, that test suite requires a local instance considering there is no equivalent testing package as with httptest. Otherwise, I can't see another concrete method of testing it. What's your opinion on that?

[deankarn]

I use Semaphore CI for my testing which has a redis database running on the default port, as long as the tests point to 127.0.0.1 and default redis port it should work

[deankarn]

Hey @hartfordfive , sorry I've been a bit busy getting the library to v2.0 but just thought I'd let you know and the http handler made it into the release!

I will be looking into the redis handler soon enough, I'm just trying to find a way to do the following:

1) Reuse the redis connection (tcp) and reconnect when disconnected, for efficiency. 2) The ability to use TLS redis connection which I don't see an easy way with the redis driver used.