Redis

Redis is a data structure server. It can be used to speed up operations that might be too simple for a heavy handed traditional database. It supports strings, lists, ordered and unordered sets, and hashes. This is meant to provide an overview of the data structures and commands available to Redis. The full list of commands is available in the documentation.

There's also a guide on MDN with suggestions on how to use Redis.

# Strings

Strings can store values of type byte string, integer (signed platform size), or floating-point (64-bit doubles). Strings support simple key-value store operations such as GET, SET, and DEL. The DEL command returns the number of items that were deleted.

Values can be incremented/decremented by 1 with INCR/DECR or by an arbitrary amount with INCRBY/DECRBY, or INCRYBYFLOAT for floats. If these are used on a key that doesn't exist or is an empty string, it's assumed to have a value of zero.

Strings also support string operations such as APPEND for concatenation, substring reading and writing with GETRANGE and SETRANGE.

Strings can also be manipulated as bitstring values with GETBIT and SETBIT, the population count can be retrieved with BITCOUNT, and bitwise operations can be performed with BITOP.

# Lists

Linked-lists support left/right push and pop with LPUSH/RPUSH and LPOP/RPOP. Specific elements can be retrieved with LINDEX, where -1 is the last index, and a range with LRANGE. Lists can be trimmed so that they only include the items in the range provided to LTRIM.

There are also potentially-blocking list manipulation commands that are often used for the development of messaging and task queues.

The operations BLPOP and BRPOP which pop the left/right-most item from the first non-empty list provided, blocking until one of the source lists is empty, or until the provided timeout in seconds 1.

The operation RPOPLPUSH pops the right-most item from the source and pushes it onto the left end of the destination, also returning the item to the user. Similarly, BRPOPLPUSH does the same but blocks until the source is not empty, or until the provided timeout in seconds.

# Sets

Sets can be added to and removed from with SADD and SREM. The SPOP command removes and returns a random item from the set. The SMOVE command can be used to move items between sets. Membership tests are possible with SISMEMBER. The number of items in a set (its cardinality) can be retrieved with SCARD.

All of the members from the set can be fetched with SMEMBERS. An arbitrary number of distinct random items can be retrieved with SRANDMEMBER, though if the arbitrary number is negative the items returned may not necessarily be distinct.

Set operations such as intersection, union, and difference are possible with SINTER, SUNION, and SDIFF. There are also STORE-suffixed variants to these set operations that store their results in a specified key instead of returning them.

# Ordered Sets

Ordered sets hold unique keys called members and associated values called scores that are limited to floating-point numbers (doubles) and are used to store the items in ascending order.

Members can be added and removed with ZADD and ZREM. The set's cardinality can be retrieved with ZCARD. More specifically, ZCOUNT can be used to count the number of items whose scores fall within the provided range.

The ZINCRBY command can be used to increment the score of a member by a specific amount. The ZSCORE command can be used to get an item's score. The ZRANGEBYSCORE command can be used to get the items whose scores fall within the provided range. It's also possible to remove items from the set based on whether or not they fall within a provided range of scores or ranks with ZREMRANGEBYSCORE and ZREMRANGEBYRANK.

The ZRANK command can be used to get the position of an item in the ascending order of the set. The ZRANGE command can be used to retrieve the items whose ranks fall within the provided range, with or without scores.

Given that the ordering of ordered sets is ascending, many commands have analogues that operate on the set as if it were in descending order, such as ZREVRANK and ZREVRANGE.

There are set operations ZINTERSTORE and ZUNIONSTORE whose semantics differ from regular set operations due to the added complexity of score-handling. These commands can take as option an operation to perform to aggregate the scores of items with the same keys, the default of which is to add them but can also be to take the min or max. Another option that can be passed are weights that are used to multiply the scores of the corresponding set before passing them to the aggregation operation. If these ordered-set operations are used on a regular set, the scores of the items in that set are interpreted as being 1.

# Hashes

Hashes can be thought of as a separate namespace within Redis. They support setting, getting, and deleting values with HSET, HGET, and HDEL. There are HMSET and HMGET variants that allow for multiple fields to be retrieved or key-value pairs to be set. The HLEN command yields the number of key-value pairs in the hash.

The HGETALL command can get all of the key-value pairs, though it's possible to get only one or the other with HKEYS and HVALUES. The existence of a key in a hash can be checked with HEXISTS.

HINCRBY/HINCRBYFLOAT can be used to increment a value by a specific amount.

# Publish-Subscribe

Redis supports publish-subscribe operations. There are straightforward commands SUBSCRIBE, UNSUBSCRIBE, and PUBLISH. There are also (un)subscribe commands that apply to channels that match a given pattern: PSUBSCRIBE and PUNSUBSCRIBE.

If a client gets disconnected and a message is sent before it can reconnect, it will never see the message. For this reason, it may be desirable to use a message broker such as RabbitMQ. The client-output-buffer-limit option is used to configure the maximum client buffer size.

# Sorting

Sorting can be performed on lists, sets, and ordered sets---even according to data in other structures.

conn.rpush('sort-input', 23, 15, 110, 7)

conn.hset('d-7', 'field', 5)
conn.hset('d-15', 'field', 1)
conn.hset('d-23', 'field', 9)
conn.hset('d-110', 'field', 3)

conn.sort('sort-input', by='d-*->field')
# ['15', '110', '7', '23']

conn.sort('sort-input', by='d-*->field', get='d-*->field')
# ['1', '3', '5', '9']

# Transactions

Simple transactions are possible in Redis by using the MULTI command to designate that the following commands are to be part of a transaction, followed by the commands themselves and EXEC, which prompts Redis to execute all of the commands sequentially. Transactions also help to reduce round trips between Redis and the client.

# Key Expiration

Keys can be deleted automatically by Redis after a certain time or at a specific time by using key expiration. The EXPIRE command can be used to make a key expire in a given number of seconds, while EXPIREAT allows a specific time to be specified. The TTL command gets the amount of time remaining before expiration. The PERSIST command removes the expiration from a key.

There are P-prefixed variants such as PTTL, PEXPIRE, and PEXPIREAT which operate at the millisecond level.


  1. This of course reminds me of the select system call used for synchronous I/O multiplexing. 

April 19, 2014