redis-tagging

Fast and simple tagging of (sorted) items

Redis-Tagging

Fast and simple tagging of (sorted) items.

  • Maintains the order of tagged items with the help of Redis Sorted Sets.
  • Unions and intersections on tags while also maintaining the order.
  • Fast and efficient paging over thousands of results with support of limit, offset.
  • Namespaces to keep multiple "buckets" of tags on the same server.
  • Counters for each tag in a namespace.
  • REST interface via REST Tagging
  • Test coverage

Tagging and efficient querying of items with unions and intersections is no fun with traditional databases.

Imagine a SQL database with concerts that need to be output ordered by date. Each item is tagged with tags like chicago, rock, stadium, open-air. Now let's try to get the following items:

  • 10 concerts (orderd by date) in chicago (limit=10, tags=["chicago"]) and the total amount of concerts in chicago.
  • The next 10 concerts, skipping the first 10, (limit=10, tags=["chicago"], offset=10) and the total amount.
  • 40 concerts in detroit, chicago or cleveland (limit=40, tags=["detroit", "chicago", "cleveland"], type="union") and the total amount.
  • 50 concerts that are rock and in a stadium (limit=50, tags=["rock", "stadium"]) and the total amount.
  • The top 20 tags used and the amount of items tagged with each.

Those queries together with the maintenance of tables and indexes can be a pain with SQL. Enter Redis and its fast in-memory set opererations.

Here is how Redis-Tagging will make the tagging of items in external databases fast and easy:

  • When storing an item in your database you still store the tags but in a normal string field for reference and easy output. You will no longer use this field in a WHERE-statement. No additional tables for tags and tag associations are needed.
  • You post the id, a score (for sorting) and the list of tags to Redis-Tagging whenever you add, update or delete an item. The score could be a date timestamp or any other number you use for sorting.
  • Redis-Tagging will output all results (e.g. all items with tags chicago and rock) as a list of IDs ordered correctly by the score you supplied.
  • You use this list of IDs to get the actual items from your database.

So with little changes you will end up with a lot less code, tables and need to maintain a complex structure just to support fast tagging.

If you want to use the REST interface to access Redis Tagging from a non NodeJS application please have a look at: REST Tagging

npm install redis-tagging

Parameters for RedisTagging:

  • host (String): optional (Default: "127.0.0.1") The Redis server
  • port (Number): optional (Default: 6379) The Redis port
  • options, optional Default: {}. Additional options. See: https://github.com/mranney/node_redis#rediscreateclientport-host-options
  • nsprefix (String): optional (Default: "rt") The namespace prefix used for all keys created by Redis Tagging
  • client: optional An external RedisClient object which will be used for the connection.
var RedisTagging = require("redis-tagging");
var rt = new RedisTagging({host: "127.0.0.1", port: 6379, nsprefix: "rt"} );

Important: Redis-Tagging works with items from your database (whatever you might use). Its purpose is to make tag based lookups fast and easy.
A typical item in your database should include an id (the primary key) and a list of tags for this items. You could store this as a JSON string (e.g. ["car", "bmw", "suv", "x5"].
You'll want to try to keep your db in sync with the item ids stored in Redis-Tagging.

Go through the following examples to see what Redis-Tagging can do for you:

This will create an item with the id itm123.
Note: There is no partial update of tags for an item. You always write the full list of tags.

rt.set(
    {
        bucket: "concerts",
        id: "itm123",
        tags: ["new york", "stadium", "rock", "open-air"],
        score: 1356341337
    },
    function (errresp) {
        if (resp === true) {
            // item was saved 
        }
    }
);

Returns all tags for an item id.

Note: This method is usually not needed if you store the tags for each item in your database.

rt.get(
    {
        bucket: "concerts",
        id: "itm123"
    },
    function (errresp) {
        // resp countains an array of all tags 
        // For the above set example resp will contain: 
        // ["new york", "stadium", "rock", "open-air"] 
    }
);

Note: This is the same as using set with an empty array of tags.

rt.remove(
    {
        bucket: "concerts",
        id: "itm123"
    },
    function (errresp) {
        if (resp === true) {
            // item was removed 
        }
    }
);
rt.allids(
    {
        bucket: "concerts"
    }
    ,
    function (errresp) {
        // resp countains an array of all ids 
    }
);

The main method. Return the IDs for one or more tags. When more than one tag is supplied the query can be an intersection (default) or a union. type=inter (default) only those IDs will be returned where all tags match. type=union all IDs where any tag matches will be returned.

Parameters object:

  • bucket (String)
  • tags (Array) One or more tags
  • limit (Number) optional Default=100 (0 will return 0 items but will return the total_items!)
  • offset (Number) optional Default=0
  • withscores (Number) optional Default=0 Set this to 1 to output the scores
  • order (String) optional Default ="desc"
  • type (String) optional "inter", "union" Default: "inter"
rt.tags(
    {
        bucket: "concerts",
        tags: ["berlin", "rock"],
        limit: 2,
        offset: 4
    },
    function (errresp) {
        // resp contains: 
        //  {"total_items":108, 
        //  "items":["8167","25652"], 
        //  "limit":2, 
        //  "offset":4} 
    }
);

The returned data is item no. 5 and 6. The first 4 got skipped (offset=4). You can now do a

SELECT * FROM Concerts WHERE ID IN (8167,25652) ORDER BY Timestamp DESC

Return the top n tags of a bucket.

rt.toptags(
{
        bucket: "concerts",
        amount: 3
    },
    function (errresp) {
        // resp contains: 
        //  { 
        //      "total_items": 18374, 
        //      "items":[ 
        //          {"tag":"rock", "count":1720}, 
        //          {"tag":"pop", "count":1585}, 
        //          {"tag":"New York", "count":720} 
        //      ] 
        //  } 
    }
);

List all buckets with at least one item stored in Redis.

Important: This method uses the Redis keys command. Use with care.

rt.buckets(
    function (errresp) {
        // resp contains an array with all buckets 
    }
);

Removes a single bucket and all items

rt.removebucket(
    {
        bucket: "concerts"
    },
    function (errresp) {
        if (resp === true) {
            // bucket was removed 
        }
    }
);
  • Make sure your DB has the following fields for the items you want to tag (names don't need to match exactly):
    • id: A primary key to quickly find your item.
    • score: Any number you use to sort your data. This is usually a date. If you saved a date in date-format you need to convert it to a numeric timestamp.
    • tags: A list of tags for this item. It is up to you how you store this. Usually a normal string field is sufficient.
  • Do a set for each item to populate the Redis-Tagging data.
  • When you insert / update / delete items in your DB make sure you also tell Redis-Tagging about it.
  • Now use the methods described above to make intersections and get the IDs back.
  • Use the IDs to get the actual records from your DB and display them as usual.
  • Enjoy.

See https://github.com/smrchy/redis-tagging/blob/master/CHANGELOG.md

NameDescription
node-cacheSimple and fast Node.js internal caching. Node internal in memory cache like memcached.
rsmqA lightweight message queue for Node.js that requires no dedicated queue server. Just a Redis server.
redis-sessionsAn advanced session store for Node.js and Redis
rsmq-workerHelper to implement a worker based on RSMQ (Redis Simple Message Queue).
redis-notificationsA Redis based notification engine. It implements the rsmq-worker to safely create notifications and recurring reports.
task-queue-workerA powerful tool for background processing of tasks that are run by making standard http requests.
obj-schemaSimple module to validate an object by a predefined schema
connect-redis-sessionsA connect or express middleware to use redis sessions that lets you handle multiple sessions per user_id.
systemhealthNode module to run simple custom checks for your machine or it's connections. It will use redis-heartbeat to send the current state to Redis.
soyerSoyer is small lib for serverside use of Google Closure Templates with node.js.
grunt-soy-compileCompile Goggle Closure Templates (SOY) templates including the handling of XLIFF language files.
backlunrA solution to bring Backbone Collections together with the browser fulltext search engine Lunr.js

Please see the LICENSE.md file.