Warning

This branch is not longer active. Both the ZDIFFSTORE command and the timeseries data-structure have been implemented using lua scripts. If timeseries is what you are looking for, check the timeseries data structure for univariate timeseries and the column timeseries for numeric multivariate timeseries suitable for modelling financial timeseries. Stdnet uses redis >= 2.6 from the standard (antirez) distribution.

Stdnet Redis branch (Discontinued)

During the development of stdnet we came across several design decisions, some more critical than others, and most of them were resolved coding the client only. However, more exotic features, such as the time series structure, required hacking on the server side.

For this reason we have the stdnet-redis branch.

Sorted sets

For some reasons redis does not have a zdiffstore command. For 99% of applications this is not an issue, but when using stdnet with models which have an implicit sorting it becomes one.

ZDIFFSTORE

Computes the difference between the first sorted set and all the successive sorted sets given by the specified keys, and stores the result in destination. It is mandatory to provide the number of input keys (numkeys) before passing the input keys and the other (optional) arguments:

ZDIFFSTORE destination numkeys key [key ...] [withscore]

If the optional argument withscore is True (default is False), elements are removed from the first sorted sets only if the score is matched.

Time-series

Time-series is an important data-structure not yet supported in redis, it is represented by a unique sorted associative container, that is to say it associates ordered unique times to values.

Values, which can be anything you like, are ordered with respect to times (you could use unix timestamp, but any double value would do), and can be accessed by times or rank (the order of times in the time-series). Times are unique so that there will be only one value associated with a given time.

Internally, a time-series is implemented using the same skiplist implementation as ordered sets. Values are added to a skip list which maintain sorting with respect to times.

Performance:

O(log(N)) on INSERT REMOVE and RETRIEVAL via times

There are currently seven commands which cover the basic operations of a timeseries. I tried to keep them to a minimum but there is scope to add more.

TSLEN

Size of timeseries at key, that is the number of rows in the timeseries:

tslen key

TSADD

Add items to timeseries at key:

tsadd key time1 value1 time2 value2 ...

If value at time is already available, the value will be updated.

TSEXISTS

Check if time is in timeseries at key:

tsexists key time

TSRANK

Returns the rank (index position) of time in timeseries at key:

tsrank key time

TSGET

Get value at time in timeseries at key:

tsget key time

TSRANGE

Range by rank:

tsrange key start end <flag>

Where start and end are integers following the same Redis conventions as zrange, <flag> is an optional string which can take two values: withtimes or novalues:

tsrange key start end           -> return values
tsrange key start end withtimes -> return (time,value)s
tsrange key start end novalues  -> return times

TSRANGEBYTIME

Range by times:

tsrangebytime key time_start time_end <flag>

Where time_start and time_end are double (timestaps) and <flag> is the same as in TSRANGE.

TSCOUNT

Count element in range by time:

tscount key time_start,time_end

This command is similar to ZCOUNT for sorted sets.

Source code changes

• Added 2 files in src: t_ts.h and t_ts.c.
• Modified redis.c to add extra commands to the command table and added the t_ts.h include.
• Modified Makefile so that t_ts.c is compiled.
• Modified object.c in decrRefCount and added t_ts.h include.
• Modified db.c in typeCommand and added t_ts.h include.
• Modified rdb.c in rdbSaveObject and rdbLoadObject and added t_ts.h include.

t_zset.c

• Modified so that t_ts.c can use its internals.
• Modified zunionInterGenericCommand function to accommodate the ZDIFFSTORE command.

Tests

Added 1 file in tests/unit/type: ts.tcl.

To run the timeseries tests:

make test TAGS="ts"