lb4-connector-dynamodb

Link

Refer to original https://github.com/tmpaul/jugglingdb-dynamodb

Forked from ggcasuso's loopback-connector-dynamodb

Overview

DynamoDB Connector for loopback (compatible with datasource-juggler) forked from this repository

Using the adapter with DynamoDB Local

• During the testing/development phase of your application's lifecycle, it is a good idea to use DynamoDB local. DynamoDB local is a java archive file that runs on your machine, and it does a very good job at mocking the original database. Download the file here.
• To run DynamoDB Local, use this command from the terminal in the directory where you extracted the tar file: java -Djava.library.path=./DynamoDBLocal_lib -jar DynamoDBLocal.jar [options]

Adapter Options

• -port: port_number (8000 by default)
• --inMemory: Run in memory
• --shared: Shared DB

Schema/Model file (DynamoDB Local)

1    var dynSettings = {
2      host: "localhost",
3      port:"8000",
4      accessKeyId: "mykey",
5      secretAccessKey:"secret"
6    };

Schema/Model Options

• host: Address of the dynamodb server. Defaults to "localhost".
• port: Port number of the dynamodb server. Defaults to "8000".
• region: DynamoDB server region. Defaults to "ap-southeast-1".
• accessKeyId: Your access key id. Defaults to "fake".
• secretAccessKey: Your secret access key. Defaults to "fake".
• maxRetries: Number of connection retries. Defaults to 0.
• logLevel: Log level. Defaults to "debug".

Model Definition

1    var Schema = require('jugglingdb').Schema;
2    var schemaDynamo = new Schema('dynamodb', dynSettings);
3
4    var User = schemaDynamo.define('User', {
5      id: { type: String, keyType: "hash", uuid: true},
6      name: { type: String },
7      age: { type: Number},
8      isDead: { type: Boolean},
9      DOB: { type: Date, keyType: "range"},
10      tasks: { type: String, sharding: true, splitter: "60kb"}
11    }, {
12      table: "User"
13    });

Using the adapter with DynamoDB remote

• Put your AWS access key and secret access key IDs along with the region in credentials.json in the root folder of your app. For example,

1    {
2      "accessKeyId": "xxxxxxxxxxxxxx",
3      "secretAccessKey": "xxxxxxxxxxxxxxxxxx",
4      "region": "us-east-1"
5    }

If this file is missing, the adapter will try to read host, port , IDs and key from the values you pass in the schema/model file. See below for an example.

Schema/Model file (DynamoDB Remote)

• The adapter first looks for environment variables AWS_ACCESS_KEY_ID & AWS_SECRET_ACCESS_KEY. If these values are found, extra settings like region, maxRetries are loaded from dynSettings. Example:

1    var dynSettings = { region: "us-east-1", logLevel: "info" }

• If environment variables cannot be located, settings are loaded by adapter from credentials.json. Any extra settings like log level can still be passed with dynSettings.

Model Definition

1    var dynSettings = { logLevel: 'info' }
2    var Schema = require('jugglingdb').Schema;
3    var schemaDynamo = new Schema('dynamodb', dynSettings);
4
5    var User = schemaDynamo.define('User', {
6      id: { type: String, keyType: "hash", uuid: true},
7      name: { type: String },
8      age: { type: Number},
9      isDead: { type: Boolean},
10      DOB: { type: Date, keyType: "range"},
11      tasks: { type: String, sharding: true, splitter: "60kb"}
12    }, {
13      table: "User",
14      ReadCapacityUnits: 15,
15      WriteCapacityUnits: 20,
16    });

Table Creation

• Table creation is done at the time of model definition. If the table already exists in the database, no action will be taken, otherwise create table command is issued. When a create table command is issued to DynamoDB, the table status will be CREATING. The table will be ready for read or writes only when the table status changes to ACTIVE. To make sure that tables are ready for your read/write operations, the adapter continuously checks for the table status. This behavior can be adjusted as follows:

1    {
2      table : "User",
3      ReadCapacityUnits : 15,
4      WriteCapacityUnits: 20,
5      tableStatus : { waitTillActive: true, timeInterval : 2000 }
6    });

• According to the above example, the table status is checked every 2000 ms.

• If tableStatus property is not specified, the adapter automatically checks table status every 5000 ms. To turn this off, specify waitTillActive as false.

Created Event

• Unlike other adapters, dynamodb adapter creates table when schema.define is called. Since the schema.define function does not accept a callback, listen to the created event to check for table creation. The adapter also fires off a secondary created-modelName event right after created where modelName is the model's name in LOWERCASE. The secondary event can be used to resolve ambiguities as to which model was created.

• Additionally, if checking for table status is enabled, the created event is emitted after the table status is ACTIVE.

One model defined in a file

1    schemaDynamo.adapter.emitter.on("created", function() {
2        // Do stuff with user...
3    });

Multiple model definitions at different places/functions in a file

• If you have multiple model definitions at different places or even different functions in the same file, consider listening to the created-modelName event. Example:

1    schemaDynamo.adapter.emitter.on("created-user", function() {
2        // Do stuff with user...
3    });

Multiple model definitions at same place/function in a file

• If there are multiple models in your schema defined together in the same function or the same place, use the following code instead:

1    // Assume there are 4 models. Check test/relations.test.js for an example.
2    var modelCount = 0;
3    schemaDynamo.adapter.emitter.on("created", function() {
4        modelCount++;
5        if (modelCount == 4) {
6            // ... Do stuff ...
7        }
8    });

• Important Note

1    During model definition, all errors are thrown because `schema.define` does not accept a callback.

Explanation

• DynamoDB stores information in the form of tables. Each table has a collection of items. Think of items as a row in a table. Each item in turn is a collection of attributes. Each attribute is stored in the database as a key:value pair.

Key Schema

• DynamoDB primary key is a composite key consisting of a hash key and a range key. For a given item in the table, the key schema can specify one hash key or one hash key and a range key.

• To specify a given attribute as hash key, use keyType: "hash" along with its definition as shown in the above example.

• DynamoDB expects a hash key for every item that is created. If this value is not available at the time of creation, there is an option to use a UUID generator to generate the hash key. To use this option, specify uuid: true in the model definition.

• Similarly, attribute becomes a range key if keyType is set to "range".

USAGE

KEYS

HASH KEY ONLY

• If a model only has a hash key, any attribute can be specified as the hash key. However, if uuid is set to true, then the attribute name must be id. This restriction comes from JugglingDB.

• If you forget to include a hash key for the model, automatically an attribute called id is generated with uuid set to true.

• If no unique ID generation is present, the value of the hash key must be provided at the time of creating an item in the table. In this case, the attribute name can be anything; not just id.

1    var User = schemaDynamo.define('User', {
2      someId: {
3        type: String,
4        keyType: "hash",
5        uuid: true
6      }, ...
7      //Not allowed
8
9    var User = schemaDynamo.define('User', {
10      someId: {
11        type: String,
12        keyType: "hash"
13      }, ...
14      // Allowed

HASH & RANGE KEYS

• If a model has both hash & range keys, a primary key attribute called id must be present in the table. The attribute name cannot be anything else other than id, or the adapter will throw an error. Unlike the above given case, id attribute does not get created for the object.

• The primary key must be defined as follows:

1    var User = schemaDynamo.define('User', {
2        id: {
3          type: String,
4          keyType: "pk",
5          separator: "--oo--"
6        },
7        companyId: {
8          type: Number,
9          keyType: "hash"
10        },
11        name: {
12          type: String
13        },
14        age: {
15          type: Number,
16          keyType: "range"
17        },
18        ...
19    });

• The separator is used to define the primary key based on the hash and range keys. If the hash key is 1 and the range key is xyz, then according to the above example, the primary key id will be 1--oo--xyz. Any random separator can be used to store the primary key, but make sure that you do not include separators like ### or ?. These separators might cause problems in the view pages of the model, wherein they will be interpreted as part of the url. The default separator is --x--.

• The important thing to note is that the primary key is purely a virtual attribute to identify a particular item. It does not get persisted in the database.

DATATYPES

• DynamoDB supports only String, Binary, and Number datatypes along with their corresponding Sets.
• The adapter currently supports String, Number, Date and Boolean datatypes.
• Null, undefined and empty strings are handled by the adapter. However, invalid date or missing numbers might cause DynamoDB to throw an error.
• Date is stored internally as a number, and boolean is stored as a string => true or false.

READ/ WRITE CAPACITY UNITS

• Provisioned Throughput for each table can be specified by using the set property function of Jugglingdb. In the above example, the read and write capacity units are set for User.
• The defaults for read and write capacity units are 5 and 10 respectively.

DATABASE LIMITATIONS

• DynamoDB has an item size limit of 64 kb. Typically data is stored in the form of big strings in NoSQL tables (e.g objects with complex data structures). Eventually these strings might exceed 64 kb in size. To overcome this limitation, the adapter uses a sharding technique. Sharding is done if sharding is set to true in the model property.

1     tasks: {
2        type: String,
3        sharding: true,
4        splitter: "63kb"
5    }

• When an attribute is being sharded, the attribute is separated from the parent table, and is stored in a different (child) table. According to the example above, tasks attribute will be stored in a new table called User_tasks. The primary key of User table id, is stored in the new table as user#id, and a new range key is assigned for every chunk that exceeds a given size. So if splitter is set to 60 kb, and the tasks string exceeds 60 kb, the structure of child table will be as shown below:

1    {
2      user#id: { "S" : "xxxxyyyyy" },
3      tasks#ID: { "N" : "1" },
4      tasks: { "S" : "This attribute has been broken" }
5    }
6
7    {
8      user#id: { "S" : "xxxxyyyyy" },
9      tasks#ID: { "N" : "2" },
10      tasks: { "S" : "down into two different pieces" }
11    }

• The value of splitter can be anywhere from 1 to 63 kb. This is to make sure that there is enough room to store the primary key of parent table and the range key in the child table.

• The attribute value (String) is broken down based on the size specified by the splitter attribute. For example, in the above given model definition, the tasks string is broken down into 60 kb chunks. Each chunk is stored as a new item with a range key and the primary key of parent item.

• Due to the large size of items being retrieved/written to the child tables, these tables require more read and write capacity compared to the original table. Read and write capacities for the child table can be specified as follows:

1    tasks: {
2      type: String,
3      sharding: true,
4      splitter: "63kb",
5      read: 15,
6      write: 20
7    }

• If read and write are not specified, the read / write capacities of the parent table is used.

• When the main item is being retrieved from the database, the adapter queries each child table, and builds back the string. As a result, the data structure is still intact after retrieval.

CRUD OPERATIONS

Create

1    // Only hash key
2
3    User = db.define('User', {
4      email: { type: String, keyType: "hash"},
5      name: { type: String }
6    });
7
8    // Both hash and range keys
9
10    Book = db.define('Book', {
11      id : { type: String, keyType: "pk"},
12      title : { type: String, keyType: "hash"},
13      subject : { type: String, keyType: "range"},
14    });
15
16    var user = new User();
17    var book = new Book();
18
19    user.email = "john@doe.com";
20    user.name = "John Doe";
21
22    /* Note that book's `id` is not specified. `id` being
23    a primary key is automatically created from the hash and range keys:
24    title and subject
25    */
26
27    book.title = "A Lost Cause";
28    book.subject = "Fiction";
29
30    User.create(user, function(err, _user) {
31    ...
32    console.log(_user);
33    /*
34    {
35      id : "john@doe.com",
36      // Note that id is set to the same value as hashKey. This ensures that views don't break
37      email: "john@doe.com",
38      name: "John Doe",
39    }
40    */
41    });
42
43    Book.create(book, function(err, _book) {
44    ...
45    console.log(_book);
46    /*
47    {
48      id : "A Lost Cause--x--Fiction", // Value is simply returned by create. It does not exist in database.
49      title: "A Lost Cause",
50      subject: "Fiction"
51    }
52    */
53    });

Find

1    var id = "0e203f96-8edc-437a-b1f0-625a584a49bd";
2    User.find(id, function (err, user) {
3      ...
4    });

All

• All method internally uses query operation if hash , hash/range keys are provided. Otherwise it will use scan operation. Both operations are limited to a result set size of 1 mb by DynamoDB.
• To overcome this limitation , the adapter uses an async doWhilst loop to continuously fetch data. Therefore, be careful while using all, especially if you have large datasets.
• An upcoming version will include an iterator which will allow you to fetch data in batches.

1    User.all({
2      where : {
3        age : { gt : 20 }
4        /*
5        or age : { between : [10,20] } - Between
6        or age : { eq : 20 }    - Equal to
7        or age : [10,20]     - In range 10 to 20
8        or age : { le : 35}  - Less than or equal to
9        */
10      }
11    }, function(err, users){
12    ...
13    });

UpdateAttributes

1
2    // Model instance.updateAttributes(...)
3
4    user.updateAttributes({name : "No Name"}, function(err, updatedUser){
5    ...
6    });

Save

• If the object does not exist in the database, invoking save will call create method first.

1    user.save(function(err, savedUser){
2    ...
3    });

Destroy

1    user.destroy(function(err){
2    ...
3    });

Running Tests

• It is strongly recommended to run tests using DynamoDB local. Run tests by issuing npm test command.
• mocha and should are required for the tests to run.
• Default test logLevel is debug. You can suppress output of the test by changing the logLevel in test/init.js to error.

Upcoming features

• Support for String, Number sets.

Bugs, Features, Enhancements etc.

• The adapter is still in its development stage, and as a result the functionality is still lacking in some areas & few bugs are expected. Please create issues in the github repository to address these. Also please try to include a test case or error log if you are reporting a bug. Good luck!