你当前正在访问 Microsoft Azure Global Edition 技术文档网站。如果需要访问由世纪互联运营的 Microsoft Azure 中国技术文档网站，请访问 https://docs.azure.cn。

materialize()

项目
08/12/2024

Applies to: ✅ Microsoft Fabric ✅ Azure Data Explorer ✅ Azure Monitor ✅ Microsoft Sentinel

Captures the value of a tabular expression for the duration of the query execution so that it can be referenced multiple times by the query without recalculation.

Syntax

materialize(expression)

Learn more about syntax conventions.

Parameters

Name	Type	Required	Description
expression	`string`	✔️	The tabular expression to be evaluated and cached during query execution.

Remarks

The materialize() function is useful in the following scenarios:

To speed up queries that perform heavy calculations whose results are used multiple times in the query.
To evaluate a tabular expression only once and use it many times in a query. This is commonly required if the tabular expression is non-deterministic. For example, if the expression uses the rand() or the dcount() functions.

Note

Materialize has a cache size limit of 5 GB. This limit is per cluster node and is mutual for all queries running concurrently. If a query uses materialize() and the cache can't hold any more data, the query will abort with an error.

Tip

Another way to perform materialization of tabular expression is by using the hint.materialized flag of the as operator and partition operator. They all share a single materialization cache.

Tip

Push all possible operators that reduce the materialized dataset and keep the semantics of the query. For example, use common filters on top of the same materialized expression.
Use materialize with join or union when their operands have mutual subqueries that can be executed once. For example, join/union fork legs. See example of using join operator.
Materialize can only be used in let statements if you give the cached result a name. See example of using let statements).

Examples of query performance improvement

The following example shows how materialize() can be used to improve performance of the query. The expression _detailed_data is defined using materialize() function and therefore is calculated only once.

Run the query

let _detailed_data = materialize(StormEvents | summarize Events=count() by State, EventType);
_detailed_data
| summarize TotalStateEvents=sum(Events) by State
| join (_detailed_data) on State
| extend EventPercentage = Events*100.0 / TotalStateEvents
| project State, EventType, EventPercentage, Events
| top 10 by EventPercentage

Output

State	EventType	EventPercentage	Events
HAWAII WATERS	Waterspout	100	2
LAKE ONTARIO	Marine Thunderstorm Wind	100	8
GULF OF ALASKA	Waterspout	100	4
ATLANTIC NORTH	Marine Thunderstorm Wind	95.2127659574468	179
LAKE ERIE	Marine Thunderstorm Wind	92.5925925925926	25
E PACIFIC	Waterspout	90	9
LAKE MICHIGAN	Marine Thunderstorm Wind	85.1648351648352	155
LAKE HURON	Marine Thunderstorm Wind	79.3650793650794	50
GULF OF MEXICO	Marine Thunderstorm Wind	71.7504332755633	414
HAWAII	High Surf	70.0218818380744	320

The following example generates a set of random numbers and calculates:

How many distinct values in the set (Dcount)
The top three values in the set
The sum of all these values in the set

This operation can be done using batches and materialize:

Run the query

let randomSet = 
    materialize(
        range x from 1 to 3000000 step 1
        | project value = rand(10000000));
randomSet | summarize Dcount=dcount(value);
randomSet | top 3 by value;
randomSet | summarize Sum=sum(value)

Result set 1:

Dcount
2578351

Result set 2:

value
9999998
9999998
9999997

Result set 3:

Sum
15002960543563

Examples of using materialize()

Tip

Materialize your column at ingestion time if most of your queries extract fields from dynamic objects across millions of rows.

To use the let statement with a value that you use more than once, use the materialize() function. Try to push all possible operators that will reduce the materialized dataset and still keep the semantics of the query. For example, use filters, or project only required columns.

    let materializedData = materialize(Table
    | where Timestamp > ago(1d));
    union (materializedData
    | where Text !has "somestring"
    | summarize dcount(Resource1)), (materializedData
    | where Text !has "somestring"
    | summarize dcount(Resource2))

The filter on Text is mutual and can be pushed to the materialize expression. The query only needs columns Timestamp, Text, Resource1, and Resource2. Project these columns inside the materialized expression.

    let materializedData = materialize(Table
    | where Timestamp > ago(1d)
    | where Text !has "somestring"
    | project Timestamp, Resource1, Resource2, Text);
    union (materializedData
    | summarize dcount(Resource1)), (materializedData
    | summarize dcount(Resource2))

If the filters aren't identical, as in the following query:

    let materializedData = materialize(Table
    | where Timestamp > ago(1d));
    union (materializedData
    | where Text has "String1"
    | summarize dcount(Resource1)), (materializedData
    | where Text has "String2"
    | summarize dcount(Resource2))

When the combined filter reduces the materialized result drastically, combine both filters on the materialized result by a logical or expression as in the following query. However, keep the filters in each union leg to preserve the semantics of the query.

    let materializedData = materialize(Table
    | where Timestamp > ago(1d)
    | where Text has "String1" or Text has "String2"
    | project Timestamp, Resource1, Resource2, Text);
    union (materializedData
    | where Text has "String1"
    | summarize dcount(Resource1)), (materializedData
    | where Text has "String2"
    | summarize dcount(Resource2))

通过