alexdown / rethinkdb-view

train of thoughts about how to implement the SQL VIEW/MATERIALIZED VIEW concept in rethinkDB

VIEW/MATERIALIZED VIEW in RethinkDB

This is just a train of thoughts about how to reproduce the functionality of a SQL VIEW/MATERIALIZED VIEW in rethinkDB.

Setup

You need the following software:

• RethinkDB (download it from http://rethinkdb.com). Install and start it (command is global, but remember to use a test data directory, something like rethinkdb -d /data/rethinkdb_data/). You can then connect to the web console at http://localhost:8080/#. You can query the db from the Data Explorer tab.
• Node.js, and the rethinkdb package (install via npm) to run test.js (the example)

The data model is:

• One db test1 with one table sales. You can add elements in it in the form:

    {
        "desc" : "fourth sale in october",
        "date" : ISODate("2014-10-14T23:00:00.000Z")
    }

  by running

    // execute only once
    r.dbCreate("test1")
    r.db("test1").tableCreate("sales")
    
    // execute many times to populate the table with entries with different dates
    r.db('test1').table("sales").insert({desc: "second sales in november", date: r.time(2014, 11, 6, '+08:00')})

Presentation

The "presentation" (such a pretentious word) goes like that:

1. You can easily create the aggregation by querying in ReQL: r.db('test1').table("sales").group([r.row("date").year(), r.row("date").month()]).count()

2. However, there is no equivalent to the mongo $out, so your only option is query via a client and then re-insert the data in a new table

3. And, as in mongo, feature requests for views and materialized views have been created but not yet prioritized.

4. The option of monitoring for changes on a table is available also in rethinkDB. It is kind of more "official" than the mongoDB equivalent "hack" of tailing the oplog. It is called change feeds

5. You can create a cursor that is returning an object every time a specific table is updated. In the example below (also in test.js, I create a changefeed on the sales table:

   connect().then(function(conn){
   	r.table('sales').changes().run(conn)
   	.then(function(cursor){
   		// do something with the cursor
   	});
   });

6. What you can do with the cursor is iterate on it using forEach.

   cursor.each(function(err, item){

   the item object is as follows:

   {
   	old_val: {...},
   	new_val: {...}	
   }
   

7. Outside of the db also here, like in mongoDB. However, a BIG BIG plus is that each change "notification" have access to the old value of each object in addition to the new one (in mongodb, there was no way to retrieve the old value of the row from the oplog, as it contains only the changes to apply).

Bedtime reading, if you like the topic:

• http://rethinkdb.com/docs/rethinkdb-vs-mongodb/
• http://rethinkdb.com/docs/comparison-tables/

SQL

As a reference, the SQL equivalent of the data model above is (PostgreSQL):

• sales table:

  CREATE TABLE sales ( date date, "desc" character varying)
  

• sales table entries:

  INSERT INTO sales VALUES ('2014-09-03', 'another sale')
  

• monthlySalesReport query/view (to create the view, prefix the query with CREATE VIEW AS):

  SELECT 
    concat(date_part('year', date), '-', date_part('month', date)) as id,
    count(*)
  FROM
    sales
  GROUP BY id