<antirez>

It’s not yet stable but it’s soon to become, and comes with a long list of things that will make Redis more useful for we users: finally Redis 4.0 Release Candidate 1 is here, and is bold enough to call itself 4.0 instead of 3.4. For me semantic versioning is not a thing, what I like instead is try to communicate, using version numbers and jumps, what’s up with the new version, and in this specific case 4.0 means “this is the shit”.

It’s just that Redis 4.0 has a lot of things that Redis should have had since ages, in a different world where one developer can, like Ken The Warrior, duplicate itself in ten copies and start to code. But it does not matter how hard I try to learn about new vim shortcuts, still the duplicate-me thing is not in my chords.

Redis is often used for caching, in a setup where a fixed maximum memory to use is specified. When new data arrives, we need to make space by removing old data. The efficiency of Redis as a cache is related to how good decisions it makes about what data to evict: deleting data that is going to be needed soon is a poor strategy, while deleting data that is unlikely to be requested again is a good one.

In other terms every cache has an hits/misses ratio, which is, in qualitative terms, just the percentage of read queries that the cache is able to serve. Accesses to the keys of a cache are not distributed evenly among the data set in most workloads. Often a small percentage of keys get a very large percentage of all the accesses. Moreover the access pattern often changes over time, which means that as time passes certain keys that were very requested may no longer be accessed often, and conversely, keys that once were not popular may turn into the most accessed keys.

WARNING: Long pretty useless blog post. TLDR is that I wrote, just for fun, a text editor in less than 1000 lines of code that does not depend on ncurses and has support for syntax highlight and search feature. The code is here: http://github.com/antirez/kilo.

Screencast here: https://asciinema.org/a/90r2i9bq8po03nazhqtsifksb

For the sentimentalists, keep reading…

A couple weeks ago there was this news about the Nano editor no longer being part of the GNU project. My first reaction was, wow people still really care about an old editor which is a clone of an editor originally part of a terminal based EMAIL CLIENT. Let’s say this again, “email client”. The notion of email client itself is gone at this point, everything changed. And yet I read, on Hacker News, a number of people writing how they were often saved by the availability of nano on random systems, doing system administrator tasks, for example. Nano is also how my son wrote his first program in C. It’s an acceptable experience that does not require past experience editing files.

I’m spending days trying to get a couple of APIs right. New APIs about modules, and a new Redis data type.
I really mean it when I say *days*, just for the API. Writing drafts, starting the implementation shaping data structures and calls, and then restarting from scratch to iterate again in a better way, to improve the design and the user facing part.

Why I do that, delaying features for weeks? Is it really so important?
Programmers are engineers, maybe they should just adapt to whatever API is better to export for the system exporting it.

It was a matter of time but it eventually happened. In the Redis 1.0 release notes, 7 years ago, I mentioned that one of the interesting features for the future was “loadable modules”. I was really interested in such a feature back then, but over the years I became more and more skeptic about the idea of adding loadable modules in Redis. And probably for good reasons.

Modules can be the most interesting feature of a system and the most problematic one at the same
time: API incompatibilities between versions, low quality modules crashing the system, a lack

I’m aboard of a flight bringing me to San Francisco. Eventually I purchased the slowest internet connection of my life (well at least for a good reason), but for several hours I was without internet, as usually when I fly.

I don’t mind staying disconnected for some time usually. It’s a good time to focus, write some code, or a blog post like this one. However when I’m disconnected, what makes the most difference is not Facebook or Twitter or Github, but the lack of text messages.

At this point text messages are a fundamental thing in my life. They are also probably the main source of distraction. I use messages to talk with my family, even just to communicate between different floors. I use messages with friends to organize dinners and vacations. I even use messages with the plumber or the doctor.

It took more than expected, but finally we have it, Redis 3.2.0 stable is out with changes that may be useful to a big number of Redis users. At this point I covered the changes multiple time, but the big ones are:

* The GEO API. Index whatever you want by latitude and longitude, and query by radius, with the same speed and easy of use of the other Redis data structures. Here you can find the API documentation: http://redis.io/commands/#geo. Thank you to Matt Stancliff for the initial implementation, that was reworked but is still at the core of the GEO API, and to the developers of ARDB for providing the geo indexing code that Matt used.

Today Redis is 7 years old, so to commemorate the event a bit I passed the latest couple of days doing a fun coding marathon to implement a new crazy command called BITFIELD.

The essence of this command is not new, it was proposed in the past by me and others, but never in a serious way, the idea always looked a bit strange. We already have bit operations in Redis: certain users love it, it’s a good way to represent a lot of data in a compact way. However so far we handle each bit separately, setting, testing, getting bits, counting all the bits that are set in a range, and so forth.

Yesterday night I was re-reading Redlock analysis Martin Kleppmann wrote (http://martin.kleppmann.com/2016/02/08/how-to-do-distributed-locking.html). At some point Martin wonders if there is some good way to generate monotonically increasing IDs with Redis.

This apparently simple problem can be more complex than it looks at a first glance, considering that it must ensure that, in all the conditions, there is a safety property which is always guaranteed: the ID generated is always greater than all the past IDs generated, and the same ID cannot be generated multiple times. This must hold during network partitions and other failures. The system may just become unavailable if there are less than the majority of nodes that can be reached, but never provide the wrong answer (note: as we'll see this algorithm has another liveness issue that happens during high load of requests).

Martin Kleppmann, a distributed systems researcher, yesterday published an analysis of Redlock (http://redis.io/topics/distlock), that you can find here: http://martin.kleppmann.com/2016/02/08/how-to-do-distributed-locking.html

Redlock is a client side distributed locking algorithm I designed to be used with Redis, but the algorithm orchestrates, client side, a set of nodes that implement a data store with certain capabilities, in order to create a multi-master fault tolerant, and hopefully safe, distributed lock with auto release capabilities.