infer_storage_schema plugin

  • Article

This plugin infers schema of external data, and returns it as CSL schema string. The string can be used when creating external tables. The plugin is invoked with the evaluate operator.

Authentication and authorization

In the properties of the request, you specify storage connection strings to access. Each storage connection string specifies the authorization method to use for access to the storage. Depending on the authorization method, the principal may need to be granted permissions on the external storage to perform the schema inference.

The following table lists the supported authentication methods and any required permissions by storage type.

Authentication method Azure Blob Storage / Data Lake Storage Gen2 Data Lake Storage Gen1
Impersonation Storage Blob Data Reader Reader
Shared Access (SAS) token List + Read This authentication method isn't supported in Gen1.
Microsoft Entra access token
Storage account access key This authentication method isn't supported in Gen1.

Syntax

evaluate infer_storage_schema( Options )

Learn more about syntax conventions.

Parameters

Name Type Required Description
Options dynamic ✔️ A property bag specifying the properties of the request.

Supported properties of the request

Name Type Required Description
StorageContainers dynamic ✔️ An array of storage connection strings that represent prefix URI for stored data artifacts.
DataFormat string ✔️ One of the supported data formats.
FileExtension string If specified, the function will only scan files ending with this file extension. Specifying the extension may speed up the process or eliminate data reading issues.
FileNamePrefix string If specified, the function will only scan files starting with this prefix. Specifying the prefix may speed up the process.
Mode string The schema inference strategy. A value of: any, last, all. The function infers the data schema from the first found file, from the last written file, or from all files respectively. The default value is last.

Returns

The infer_storage_schema plugin returns a single result table containing a single row/column holding CSL schema string.

Note

  • Storage container URI secret keys must have the permissions for List in addition to Read.
  • Schema inference strategy 'all' is a very "expensive" operation, as it implies reading from all artifacts found and merging their schema.
  • Some returned types may not be the actual ones as a result of wrong type guess (or, as a result of schema merge process). This is why you should review the result carefully before creating an external table.

Example

let options = dynamic({
  'StorageContainers': [
    h@'https://storageaccount.blob.core.chinacloudapi.cn/MobileEvents;secretKey'
  ],
  'FileExtension': '.parquet',
  'FileNamePrefix': 'part-',
  'DataFormat': 'parquet'
});
evaluate infer_storage_schema(options)

Output

CslSchema
app_id:string, user_id:long, event_time:datetime, country:string, city:string, device_type:string, device_vendor:string, ad_network:string, campaign:string, site_id:string, event_type:string, event_name:string, organic:string, days_from_install:int, revenue:real

Use the returned schema in external table definition:

.create external table MobileEvents(
    app_id:string, user_id:long, event_time:datetime, country:string, city:string, device_type:string, device_vendor:string, ad_network:string, campaign:string, site_id:string, event_type:string, event_name:string, organic:string, days_from_install:int, revenue:real
)
kind=blob
partition by (dt:datetime = bin(event_time, 1d), app:string = app_id)
pathformat = ('app=' app '/dt=' datetime_pattern('yyyyMMdd', dt))
dataformat = parquet
(
    h@'https://storageaccount.blob.core.chinacloudapi.cn/MovileEvents;secretKey'
)