loopback-connector-azure-search

LoopBack connector for Microsoft Azure Search.

How to install?

Install the connector package to your project:

npm install --save loopback-connector-azure-search

Set-up a new data source that uses the connector. In datasources.json, put the following:

{
  "myDataSourceName": {
    "name": "myDataSourceName",
    "connector": "loopback-connector-azure-search",
    "host": "YOUR AZURE SEARCH HOST, TYPICALLY https://myAccount.search.windows.net:443/",
    "adminOrQueryKey": "YOUR ADMIN/QUERY KEY",

    "allowNonUrlSafeIDs": true false,
    "retry": {
        "max": 3,
        "waitInMilliseconds": 50
    }
  }
}

You can name your data source as you like, here we used myDataSourceName as an example. You can find values for YOUR AZURE SEARCH HOST and YOUR ADMIN/QUERY KEY in Azure portal. Please consult Azure documentation if needed. Rest of the properties are optional and they can be left undefined. However, we recommend setting a retry policy as shown above which automatically retries failed requests. Currently, only a constant wait retries are supported.

When configuring model properties, you may want to set specific settings for Azure Search. For example:

{
  "azureSearch": {
    "indexName": "my-index",
    "analyzers": [
        "See: https://docs.microsoft.com/en-us/rest/api/searchservice/custom-analyzers-in-azure-search"
    ]
  },

  "myViewProperty": {
    "azureSearch": {
      "propertyName": "myDatabaseProperty",
      "searchable": true false,
      "filterable": true false,
      "sortable": true false,
      "facetable": true false,
      "retrievable": true false,
      "analyzer": "See: https://docs.microsoft.com/rest/api/searchservice/language-support"
    }
  }
}

Special Azure Search settings are set via azureSearch configuration object, but this is not required -- connector will use its default values if nothing else is specified in your model configuration. Also, if you do not specify values for searchable, filterable, sortable, facetable, retrievable and analyzer properties, connector will use Azure Search default values which may change in future. Please consult LoopBack documentation for more details on how to configure models and Azure documentation on indexing options.

There are special Azure Search properties that can be mapped via propertyName configuration property. These special fields are just mappings so they are not created to index and they can only be accessed in special contexts. The available mapping properties are described in a table below.

Special propertyName value	Mapped value
@odata.count	When using limit and offset filters, the query is recognized as a pagination query. You can receive the total number of matching items in an index with this property.
@search.score	Azure Search gives a score for each result. You can access the score value via this property.

When you have specified your models, you must create Azure Search indices from models. You can do this by calling autoupdate() connector method. When you change model properties, you can update indices by calling autoupdate() again. However, Azure Search has some limitations on what kind of updates to indices can be applied. In those cases, you will receive an error and are forced to use automigrate() that removes old indices and then creates the new ones. But keep in mind that this will remove all indices and all the data that they may contain! Please consult LoopBack documentation for more details.

You are now ready to start using Azure Search in your LoopBack application!

Supported features

The following LoopBack's base connector methods are implemented:

• connect()

• disconnect()

• autoupdate()

• automigrate()

• create()

• updateOrCreate()

• replaceOrCreate()

• all()

• save()

• count()

Connector debug string is loopback:connector:azure-search that can be enabled through DEBUG environment variable.

Type mappings

The type mappings of LoopBack and Azure Search indices are described in a table below.

LoopBack type	Azure Search type	Notes
string	Edm.String
array	Collection(Edm.String)
boolean	Edm.Boolean
date	Edm.DateTimeOffset
GeoPoint	Edm.GeographyPoint
number	Edm.Int64, Edm.Int32 or Edm.Double	By default, Edm.Double is used. You can change this by setting azureSearch.integerProperty = 32 or azureSearch.integerProperty = 64 to map to Emd.Int32 or Emd.Int64 respectively. Alternatively, you can set azureSearch.integerProperty = true which defaults to Edm.Int64.

Attempt to use an unsupported type will result an error.

Special notes about filtering

All LoopBack-style filtering features are implemented (see https://loopback.io/doc/en/lb3/Querying-data.html). In addition, the following extra filtering operators are implemented:

• any operator can be used to filter array properties. For example: { where: { property: { 'any x': { x: 'variable' } } } }, where x is a variable name you get to choose that is bound to a local scope where you can apply other filtering rules. This example ensures that property array contains at least one item with a value 'variable'.
• all operator can be used to filter array properties. Usage: { where: { property: { 'all x': { x: 'variable' } } } }, where x is a variable name you get to choose that is bound to a local scope where you can apply other filtering rules. This example ensures that property array contains only items with a value 'variable'.
• If you need to perform a full-text search from all fields, you can use a special * wildcard field to indicate this: { where: { '*': 'search keyword' } }.
• You can select searchable fields with searchFields filter. It works as fields filter, but instead filtering results, it only affects properties where search is made.

Change log

• 0.0.1 -- Initial release.
• 0.1.0 -- Add support for retry policies. Lots of improvements and bug fixes.
• 0.2.0 -- Add support for unsafe keys via allowNonUrlSafeIDs setting property.
• 0.3.0 -- Add support for fields and order filters.
• 0.3.1 -- Support analyzer field property.
• 0.4.0 -- Support analyzers, searchAnalyzer and indexAnalyzer index properties. Add searchFields filter. Fix fields filter. Inject special Azure Search properties to view properties.
• 0.4.1 -- Fix incorrectly handled star property. Fix how full-text search are escaped.
• 0.5.0 -- Add support for document delete. Use full words in full-text search. Fix minor bugs that violated the specification.
• 0.5.1 -- Add option defaultIndexAction that sets the default indexing action if it is not provided with data.
• 0.5.2 -- Fix null value where filtering. Fix logical query where filters.
• 0.5.3 -- Retry all other but 400 (Bad Request) errors.
• 0.6.0 -- Fix like operators. Change how like operators are parsed: now they internally use search.ismatch inside a filter clause unless search property is a special wildcard property. In order to use full-text search from other properties, filterable must be enabled for a property.