Use Redis as storage

[erfantarighi]

I added a RedisStorage class which is faster and more efficient, I strongly recommend using RedisStorage as the default storage. because in high traffic using the Django default database can harm or make the backend slow (MySQL has a connection limit)

[jkarneges]

Conceptually looks good.

Comments:

• It looks like event_counter is double-incremented when appending an event.
• Ideally, appending should be transactional. Might need some watch statements. Or could do it as a lua script.
• Consider pipelining reads, otherwise there would be a round trip per event.
• I wonder about configuring using a URL and initializing with Redis.from_url(), instead of configuring using a dict (or maybe allow both).
• I guess there's no need for an explicit trim method since redis will auto expire via TTLs?

[erfantarighi]

Thanks for the comments @jkarneges

• Yes, it's my bad, I will fix it
• There are no transactional actions in Redis, Redis is a single-thread software which means you won't face concurrency issues
• I used pipelines on the append function
• Yes, we can use the from_url function but it's not that safe for input, developers may be confused about setting the URLs especially if they have some characters such as # in their passwords, in this case, they should encode (Using URLEncode) their URLs and then set them.
• Yes, I removed it because I used Redis TTLs

[jkarneges]

Well, there are transactions of sorts using multi/exec. I forgot that the python lib does this by default when pipeline mode is used.

I'll merge this but I strongly suggest pipelining the reads too. Otherwise, a single call to get_events() may do many round trips (e.g. fetching 100 items is 100 round trips).

[williamrexhardin]

This change is particularly interesting. Will there be a release available soon with these changes? 5.1.1?

[jkarneges]

Released 5.2.0 with this change.