sails-cassandra

Apache Cassanda database adapter for Sails/Waterline. It works with 2.x and 3.x versions of Cassandra.

Implements:

• Semantic
  • .create()
  • .createEach()
  • .find()
  • .count()
  • .update()
  • .destroy()
• Queryable
• Migratable
  • .define()
  • .describe()
  • .drop()
• Iterable
  • .stream()
• SQL
  • .query()

1. Installation

Install from NPM:

$ npm install sails-cassandra --save

2. Configuring Sails

Add the cassandra configuration to the config/connections.js file. The basic options are as follows:

module.exports.connections = {

  my-cassandra-connection: {

    module        : 'sails-cassandra',

    // typical sails/waterline options (see comment below)
    user          : 'username',
    password      : 'password',

    // cassandra driver options
    contactPoints : [ '127.0.0.1' ],
    keyspace      : 'keyspace name',
    ...
  }
};

And then change default model configuration in the config/models.js:

module.exports.models = {
  connection: 'my-cassandra-connection'
};

Adapter configuration may contain any of Cassandra client options. However, you probably will only need contactPoints and keyspace to get started and the adapter will provide reasonable defaults for the rest.

Authentication information for cassandra-driver is typically supplied in authProvider option, however sails-cassandra adapter will also recognize user and password options and convert them into authProvider overriding an existing value. This also means that if you wish to use your own authProvider you will need to remove user and password from the configuration.

3. Running Tests

You can set environment variables to override the default database configuration for running tests as follows:

$ WATERLINE_ADAPTER_TESTS_PASSWORD=yourpass npm test

Default settings are:

{
  contactPoints: [ process.env.WATERLINE_ADAPTER_TESTS_HOST || '127.0.0.1' ],
  user: process.env.WATERLINE_ADAPTER_TESTS_USER || 'root',
  password: process.env.WATERLINE_ADAPTER_TESTS_PASSWORD || '',
  keyspace: process.env.WATERLINE_ADAPTER_TESTS_DATABASE || 'test'
}

Note: Default name of the keyspace for running tests is test. Make sure you created it in your database before executing npm test.

4. Implementation Notes

This section describes behaviour of Apache Cassandra adapter that is distinct from other database types.

4.1. Naming of tables and columns

Column and table names in Cassandra are case insensitive and this ambiguity makes it difficult to map between attribute names that are case sensitive and column names that are not. There are two possible workarounds for this:

1. There is a mechanism in Apache Cassandra to make table/column names case sensitive by including them in double quotes. This may seem as a good idea on the surface but it does not cover a use case when tables are not created by sails/waterline but by an external process.

2. Converting table/column names to lower case is an another approach. This way we always convert table/column names to lower case before mapping them to/from attribute names. This is not very elegant but it works and this is the current preferred approach.

4.2. Autoincrement

The autoincrement feature was plaguing ORM frameworks right from their inseption as it requires 1-2 extra queries in order to retrieve new record identifier from underlying database into the framework. It also does not work very well with sharding and replication.

Cassandra database does not support autoincrement, however it achieves the same functionality in a much more efficient way by using time based UUIDs (a.k.a. Type 1 UUID) for primary keys.

Sails/Waterline supports autoincrement and its implementation is heavily influenced by MySQL database. The sails-cassandra adapter makes an attempt to achieve the same functionality using the following rules:

1. Model attribute that represents primary key may have autoIncrement property set to true.

2. This automatically forces attribute type to string and supersedes any other declarations. The adapter will give a warning message is there is a discrepancy.

3. The value of the primary key cannot be overridden by create() or update() calls once autoIncrement property is enabled. You will see a (non-lethal) warning message if such attempt is made.

Note: This logic is inconsistent with the current Sails/Waterline specifications as it requires autoIncrement field to be of type integer. Please use discretion. Also, see this issue.

4.3. Type conversion between Cassandra and Sails/Waterline

The following table represents mappings between Sails/Waterline model data types and Apache Cassandra data types:

| Sails/Waterline Type | JS Type | Cassandra Type | |:---------------------|:---------|:----------------------------------| | string | String | text (UTF-8 text) | | text | String | text (UTF-8 text) | | integer | Number | bigint (64-bit signed integer) | | float | Number | double (64-bit float) | | date | Date | timestamp | | datetime | Date | timestamp | | boolean | Boolean | boolean | | binary | Buffer | blob | | array | Array | list | | json | ??? | text (UTF-8 text) | | email | String | ascii (US-ASCII character string) | | autoIncrement=true | String | timeuuid |

The following table may be used as a guideline when creating Sails/Waterline models for existing tables:

| Cassandra Type | Type Id | Driver JS type | Waterline JS Type | Waterline Type | |:---------------|:-------:|:---------------|:------------------|:---------------------| | ascii | 1 | String | String | string | | bigint | 2 | Long | Number or NaN | integer | | blob | 3 | Buffer | Buffer | binary | | boolean | 4 | Boolean | Boolean | boolean | | counter | 5 | Long | Number or NaN | integer | | date | 17 | LocalDate | Date | date | | decimal | 6 | BigDecimal | Number or NaN | float | | double | 7 | Number | Number | float | | float | 8 | Number | Number | float | | inet | 16 | InetAddress | String | string | | int | 9 | Number | Number | integer | | list | 32 | Array | Array | array | | map | 33 | Object/ES6 Map | Null | not supported (null) | | set | 34 | Array/ES6 Set | Null | not supported (null) | | smallint | 19 | Number | Number | integer | | text | 10 | String | String | text | | time | 18 | LocalTime | Number | integer | | timestamp | 11 | Date | Date | datetime or date | | timeuuid | 15 | TimeUuid | String | string | | tynyint | 20 | Number | Number | integer | | tuple | 49 | Tuple | Array | array | | uuid | 12 | Uuid | String | string | | varchar | 13 | String | String | text | | varint | 14 | Integer | Number or NaN | integer |

4.4. Use of indexes

Apache Cassandra require index on a column that is used in where clause of select statement and unlike other database it will produce and exception if the index is missing.

Sails/Waterline allows to set index or unique properties on model attributes. The sails-cassanda adapter will respect these attributes and it will create indexes for attributes with index or unique attributes set to true.

Note: that Apache Cassandra have no notion of unique constraint and the uniqueness has to be enforced either by Sails/Waterline core or in your own code. The unique attribute property is considered an alias for index and both are treated in the exactly same way.

4.5. Search criteria

Apache Cassandra only supports subset of operation in selection criteria in comparison to relational databases and this section describes what is currently supported.

4.5.1. Key Pairs

This is an exact match criteria and it is declared as follows:

Model.find({firstName: 'Joe', lastName: 'Doe'});

It is supported and it will be executed as follows:

SELECT id, first_name, last_name
  FROM users
  WHERE first_name = 'Joe' AND last_name = 'Doe'
  ALLOW FILTERING;

Please also refer to Use of Indexes above.

4.5.2. Modified Pair

This criteria:

Model.find({age: {'>': 18, 'lessThanOrEqual': 65});

will be converted to CQL query that may look like this:

SELECT id,first_name,last_name
  FROM users
  WHERE age > 18 AND age <= 65
  ALLOW FILTERING;

and supported operations are as follows:

| Operation | Shorthand | Supported | |:-----------------------|:---------:|:---------:| | 'lessThan' | '<' | Yes | | 'lessThanOrEqual' | '<=' | Yes | | 'greaterThan' | '>' | Yes | | 'greaterThanOrEqual' | '>=' | Yes | | 'not' | '!' | No | | 'like' | none | No | | 'contains' | none | No | | 'startsWith' | none | No | | 'endsWith' | none | No |

4.5.3. In Pairs

This criteria:

Model.find({title: ['Mr', 'Mrs']});

will be rendered into the following CQL statement:

SELECT id, first_name, last_name
  FROM users
  WHERE title IN ( 'Mr', 'Mrs' )
  ALLOW FILTERING;

Note: that IN criterion works differently in Apache Cassandra. It is subject of certain limitations and is considered a pattern to be avoided.

4.5.4. Not-In Pair

Not supported since Apache Cassandra does not support NOT IN criterion, so this construct:

Model.find({name: {'!': ['Walter', 'Skyler']}});

will cause adapter to throw an exception.

4.5.5. Or Pairs

Not supported since Apache Cassandra has no OR criterion, so this construct:

Model.find({
  or : [
    {name: 'walter'},
    {occupation: 'teacher'}
  ]
});

will cause the adapter to throw an exception.

4.5.6. Limit, Sort, Skip

Only limit is curently implemented and works as expected. sort and skip are not supported and silently ignored if provided.

5. Version History

• 0.12.x - Supports Apache Cassandra v2.x and v3.x, uses cassandra-driver v3.x, requires Node.js v4+
• 0.10.x - Supports Apache Cassandra v2.x, uses cassandra-driver v2.x, requires Node.js v0.10+

6. License

See LICENSE.md file for details.