tom-operations

作者: microsoft

使用 TOM(表格物件模型)包裝器進行工作的指南。當需要以程式方式修改語意模型時使用此功能。

npx skills add https://github.com/microsoft/semantic-link-labs --skill tom-operations

TOM (Tabular Object Model) Operations

This skill covers working with the TOM wrapper in Semantic Link Labs for programmatic semantic model management.

When to Use This Skill

Use this skill when you need to:

  • Read or modify semantic model metadata
  • Add/remove tables, columns, measures, relationships
  • Work with calculation groups and items
  • Manage partitions and data sources
  • Implement model management features

Overview

The TOM wrapper provides a Pythonic interface to the Tabular Object Model (TOM), which is the object model for Analysis Services semantic models.

Key Files

FilePurpose
src/sempy_labs/tom/__init__.pyModule exports
src/sempy_labs/tom/_model.pyTOMWrapper class implementation

Basic Usage

Connecting to a Semantic Model

from sempy_labs.tom import connect_semantic_model

# Read-only connection
with connect_semantic_model(
    dataset="My Semantic Model",
    readonly=True,
    workspace="My Workspace"
) as tom:
    # Read model metadata
    for measure in tom.all_measures():
        print(f"Measure: {measure.Name}")

# Read-write connection (requires XMLA read/write enabled)
with connect_semantic_model(
    dataset="My Semantic Model",
    readonly=False,
    workspace="My Workspace"
) as tom:
    # Modify model
    measure = tom.model.Tables["Sales"].Measures["Revenue"]
    measure.Description = "Total revenue in USD"
    # Changes are saved when context exits

Important Notes

  1. XMLA Endpoint: Read-write operations require XMLA read/write enabled on the capacity
  2. Context Manager: Always use with statement - changes are saved on exit
  3. Service Principal: Supported via service_principal_authentication context

TOMWrapper Properties and Methods

Core Properties

with connect_semantic_model(dataset, workspace=workspace) as tom:
    # Access the model
    model = tom.model

    # Get dataset info
    dataset_id = tom._dataset_id
    dataset_name = tom._dataset_name
    workspace_id = tom._workspace_id
    workspace_name = tom._workspace_name

    # Check compatibility level
    compat_level = tom._compat_level

Iterator Methods

MethodReturns
all_columns()All columns in all tables
all_calculated_columns()All calculated columns
all_calculated_tables()All calculated tables
all_calculation_groups()All calculation groups
all_measures()All measures
all_partitions()All partitions
all_hierarchies()All hierarchies
all_levels()All hierarchy levels
all_calculation_items()All calculation items
all_functions()All user-defined functions

Example: Iterating Objects

with connect_semantic_model(dataset, workspace=workspace) as tom:
    # List all measures
    for measure in tom.all_measures():
        print(f"Table: {measure.Parent.Name}, Measure: {measure.Name}")

    # List all columns
    for column in tom.all_columns():
        print(f"Table: {column.Table.Name}, Column: {column.Name}, Type: {column.DataType}")

    # List all partitions
    for partition in tom.all_partitions():
        print(f"Table: {partition.Table.Name}, Partition: {partition.Name}")

Reading Model Metadata

Tables

with connect_semantic_model(dataset, workspace=workspace) as tom:
    for table in tom.model.Tables:
        print(f"Table: {table.Name}")
        print(f"  Description: {table.Description}")
        print(f"  Is Hidden: {table.IsHidden}")

Columns

with connect_semantic_model(dataset, workspace=workspace) as tom:
    for column in tom.all_columns():
        print(f"Column: {column.Name}")
        print(f"  Table: {column.Table.Name}")
        print(f"  Data Type: {column.DataType}")
        print(f"  Is Hidden: {column.IsHidden}")

Measures

with connect_semantic_model(dataset, workspace=workspace) as tom:
    for measure in tom.all_measures():
        print(f"Measure: {measure.Name}")
        print(f"  Table: {measure.Parent.Name}")
        print(f"  Expression: {measure.Expression}")
        print(f"  Format String: {measure.FormatString}")

Relationships

with connect_semantic_model(dataset, workspace=workspace) as tom:
    for rel in tom.model.Relationships:
        print(f"From: {rel.FromTable.Name}[{rel.FromColumn.Name}]")
        print(f"To: {rel.ToTable.Name}[{rel.ToColumn.Name}]")
        print(f"Cross Filter: {rel.CrossFilteringBehavior}")

Modifying Model Metadata

Update Measure Properties

with connect_semantic_model(dataset, readonly=False, workspace=workspace) as tom:
    measure = tom.model.Tables["Sales"].Measures["Revenue"]
    measure.Description = "Total revenue"
    measure.FormatString = "$#,##0.00"
    measure.DisplayFolder = "Financial Metrics"

Update Column Properties

with connect_semantic_model(dataset, readonly=False, workspace=workspace) as tom:
    column = tom.model.Tables["Date"].Columns["Month"]
    column.IsHidden = False
    column.Description = "Calendar month"
    column.SortByColumn = tom.model.Tables["Date"].Columns["MonthNumber"]

Update Table Properties

with connect_semantic_model(dataset, readonly=False, workspace=workspace) as tom:
    table = tom.model.Tables["Sales"]
    table.Description = "Sales transactions"
    table.IsHidden = False

Adding Model Objects

Add a Measure

import Microsoft.AnalysisServices.Tabular as TOM

with connect_semantic_model(dataset, readonly=False, workspace=workspace) as tom:
    table = tom.model.Tables["Sales"]

    new_measure = TOM.Measure()
    new_measure.Name = "Total Sales"
    new_measure.Expression = "SUM(Sales[Amount])"
    new_measure.FormatString = "$#,##0.00"
    new_measure.Description = "Sum of all sales amounts"

    table.Measures.Add(new_measure)

Add a Calculated Column

import Microsoft.AnalysisServices.Tabular as TOM

with connect_semantic_model(dataset, readonly=False, workspace=workspace) as tom:
    table = tom.model.Tables["Products"]

    calc_column = TOM.Column()
    calc_column.Name = "Profit Margin"
    calc_column.Type = TOM.ColumnType.Calculated
    calc_column.Expression = "[Revenue] - [Cost]"
    calc_column.DataType = TOM.DataType.Decimal

    table.Columns.Add(calc_column)

Add a Relationship

import Microsoft.AnalysisServices.Tabular as TOM

with connect_semantic_model(dataset, readonly=False, workspace=workspace) as tom:
    from_column = tom.model.Tables["Sales"].Columns["ProductKey"]
    to_column = tom.model.Tables["Products"].Columns["ProductKey"]

    relationship = TOM.SingleColumnRelationship()
    relationship.Name = f"{from_column.Table.Name}_{to_column.Table.Name}"
    relationship.FromColumn = from_column
    relationship.ToColumn = to_column
    relationship.CrossFilteringBehavior = TOM.CrossFilteringBehavior.OneDirection

    tom.model.Relationships.Add(relationship)

TOMWrapper Helper Methods

The TOMWrapper class includes many helper methods for common operations:

Display Methods (return DataFrames)

with connect_semantic_model(dataset, workspace=workspace) as tom:
    # List all measures
    df = tom.list_measures()

    # List all columns
    df = tom.list_columns()

    # List all relationships
    df = tom.list_relationships()

    # List all partitions
    df = tom.list_partitions()

Add Methods

with connect_semantic_model(dataset, readonly=False, workspace=workspace) as tom:
    # Add measure
    tom.add_measure(
        table_name="Sales",
        measure_name="Total Sales",
        expression="SUM(Sales[Amount])",
        format_string="$#,##0.00"
    )

    # Add calculated column
    tom.add_calculated_column(
        table_name="Products",
        column_name="Margin",
        expression="[Revenue] - [Cost]"
    )

Update Methods

with connect_semantic_model(dataset, readonly=False, workspace=workspace) as tom:
    # Update measure expression
    tom.update_measure(
        table_name="Sales",
        measure_name="Total Sales",
        expression="SUMX(Sales, Sales[Quantity] * Sales[Price])"
    )

Service Principal Authentication

from sempy_labs import service_principal_authentication
from sempy_labs.tom import connect_semantic_model

# Using Service Principal
with service_principal_authentication(
    tenant_id="...",
    client_id="...",
    client_secret="..."
):
    with connect_semantic_model(
        dataset="My Model",
        readonly=False,
        workspace="My Workspace"
    ) as tom:
        # Operations use Service Principal credentials
        for measure in tom.all_measures():
            print(measure.Name)

Azure Analysis Services Support

The TOMWrapper also supports Azure Analysis Services:

with connect_semantic_model(
    dataset="MyDatabase",
    workspace="asazure://westus2.asazure.windows.net/myserver",
    readonly=True
) as tom:
    for table in tom.model.Tables:
        print(table.Name)

Best Practices

Do's

  • ✅ Always use context manager (with statement)
  • ✅ Use readonly=True when only reading metadata
  • ✅ Use iterator methods (all_measures(), etc.) for efficient traversal
  • ✅ Save changes by exiting the context normally

Don'ts

  • ❌ Don't open read-write connections when only reading
  • ❌ Don't forget to handle exceptions (changes won't save on exception)
  • ❌ Don't hold connections open longer than necessary

Error Handling

from sempy.fabric.exceptions import FabricHTTPException

try:
    with connect_semantic_model(
        dataset="My Model",
        readonly=False,
        workspace="My Workspace"
    ) as tom:
        # Modifications...
        pass
except FabricHTTPException as e:
    print(f"API error: {e}")
except Exception as e:
    print(f"Error: {e}")
    # Changes NOT saved due to exception

Additional Resources

ResourceURL
TOM ReferenceMicrosoft Docs
Sample NotebookTabular Object Model.ipynb
API DocumentationReadTheDocs

來自 microsoft 的更多技能

oss-growth
microsoft
開源增長駭客角色
agent-framework-azure-ai-py
microsoft
使用Microsoft Agent Framework Python SDK(agent-framework-azure-ai)构建Azure AI Foundry代理。适用于使用AzureAIAgentsProvider创建持久化代理、使用托管工具(代码解释器、文件搜索、网络搜索)、集成MCP服务器、管理对话线程或实现流式响应。涵盖函数工具、结构化输出和多工具代理。
development
airunway-aks-setup
microsoft
在AKS上設定AI Runway——從裸叢集到執行模型。涵蓋叢集驗證、控制器安裝、GPU評估、供應商設定及首次部署。時機:「設定AI Runway」、「上線AKS叢集」、「安裝AI Runway」、「airunway設定」、「部署模型至AKS」、「在AKS上進行GPU推論」、「在AKS上設定KAITO」、「在AKS上執行LLM」、「在AKS上使用vLLM」、「在AKS上設定模型服務」、「AI Runway控制器」。
devops
appinsights-instrumentation
microsoft
使用Azure Application Insights檢測Web應用程式的指南。提供遙測模式、SDK設定與組態參考。適用時機:如何檢測應用程式、App Insights SDK、遙測模式、什麼是App Insights、Application Insights指南、檢測範例、APM最佳實踐。
devops
applicationinsights-web-ts
microsoft
使用Application Insights JavaScript SDK(@microsoft/applicationinsights-web)為瀏覽器/Web應用程式進行檢測。適用於真實使用者監控(RUM)——頁面檢視、點擊、AJAX/fetch依賴、例外、自訂事件,以及與後端OpenTelemetry追蹤關聯的瀏覽器端GenAI代理追蹤。涵蓋SDK載入器指令碼與npm設定、框架擴充(React、React Native、Angular)、點擊分析、遙測初始化器,以及從瀏覽器發出的代理/工具/模型span的OTel GenAI語意慣例。
devops
azure-ai-anomalydetector-java
microsoft
使用適用於 Java 的 Azure AI 異常偵測器 SDK 建置異常偵測應用程式。在實作單變量/多變量異常偵測、時間序列分析或 AI 驅動監控時使用。
development
azure-ai-language-conversations-py
microsoft
使用 azure-ai-language-conversations Python SDK 實作對話語言理解(CLU)。當使用 ConversationAnalysisClient 分析對話意圖與實體、建置 NLP 功能,或將語言理解整合至應用程式時使用。
development
azure-ai-ml-py
microsoft
Azure Machine Learning SDK v2 for Python。用於機器學習工作區、作業、模型、資料集、計算資源與管線。 觸發詞:「azure-ai-ml」、「MLClient」、「workspace」、「model registry」、「training jobs」、「datasets」。
development