驗證 Expectation Suite

Great Expectations 資料品質入門

Davina Moossazadeh

Data Scientist

Validation Definition

Validation Definition - 將 Expectation Suite 與其描述的資料連結的參照

示意圖：Validation Definition 以 Data 與 Expectation Suite 為輸入，輸出為 Validation Results。

¹ https://docs.greatexpectations.io/docs/core/run_validations/create_a_validation_definition/

建立 Validation Definition

使用 ValidationDefinition 類別建立 Validation Definition：

validation_definition = gx.ValidationDefinition(

    name="my_validation_definition",

    data=batch_definition,

    suite=suite, 
)

print(validation_definition)

檢視 Validation Definition

name='my_validation_definition' 
data=BatchDefinition(
    id='1fcb36d6-fac6-4b9a-8ba6-a659978fd59e', 
    name='my_batch_definition', 
    partitioner=None
) 
suite={
  "name": "my_suite",
  "id": "0a123b9c-e370-4b18-b703-785dde88732d",
  "expectations": [],
  "meta": {"great_expectations_version": "1.2.4"},
  "notes": null
} 
id=None

檢視 Validation Definition

print(validation_definition.name)

'my_validation_definition'

print(validation_definition.data)

id='1fcb36d6-fac6-4b9a-8ba6-a659978fd59e' 
name='my_batch_definition' 
partitioner=None

print(validation_definition.suite)

{
  "name": "my_suite",
  "id": "0a123b9c-e370-4b18-b703-785dde88732d",
  "expectations": [],
  "meta": {"great_expectations_version": "1.2.4"},
  "notes": null
}

print(validation_definition.id)

None

檢視 Validation Definition

print(validation_definition.data_source)

assets:
  - batch_definitions:
      - name: my_batch_definition
        partitioner: null
    batch_metadata: {}
    id: 83682084-3bc4-4898-a807-fadc0f911415
    name: 'my_dataframe_asset'
    type: dataframe
id: f71d275e-a5b2-402e-a53c-8dad6975cce5
name: 'my_pandas_data_source'
type: pandas

執行 Validation Definition

使用 Validation Definition 的 .run() 方法執行 Validation，透過 batch_parameters 傳入 DataFrame：

validation_results = validation_definition.run(

    batch_parameters={"dataframe": dataframe}
)

Validation Definition 錯誤

請注意錯誤訊息：

ValidationDefinitionRelatedResourcesFreshnessError:
ExpectationSuite 'my_suite' must be added to the DataContext before it can be 
updated. Please call `context.suites.add(<SUITE_OBJECT>)`, then try your action 
again.

在將 Expectation Suite 加入 Data Context 之前就執行 Validation Definition，會產生錯誤

加入 Expectation Suite

使用 .suites.add() 將 Expectation Suite 加入 Data Context：

suite = context.suites.add(

    suite=suite
)

評估 Validation Definition

validation_results = validation_definition.run(
    batch_parameters={"dataframe": dataframe}
)

Calculating Metrics: 0/0 [00:00<?, ?it/s]

print(validation_results.success)

False

print(validation_results.describe())

評估 Validation Definition

{ "success": false,
  "statistics": {
    "evaluated_expectations": 1, "successful_expectations": 0,
    "unsuccessful_expectations": 1, "success_percent": 0.0
  },
  "expectations": [{
    "expectation_type": "expect_table_row_count_to_equal",
    "success": false,
    "kwargs": {"batch_id": ""my_datasource-my_dataframe_asset", "value": 118000}, 
    "result": {"observed_value": 11866}}
  ],
  "result_url": "https://app.greatexpectations.io/organizations/my_org/data-assets/*/validations/expectation-suites/0a123b9c-e370-4b18-b703-785dde88732d/results/cb093105-6ede-47d4-a141-dee10c632e18"
}

關於 GX 的補充說明

Great Expectations 提供多種流程來完成相似工作
- 例如：Batch Definition vs. Validation Definition
本課程提供整體觀念；其他 GX 實作可能採用不同做法

小抄

將 Expectation Suite 加入 Data Context：

context.suites.add(suite)

建立 Validation Definition：

validation_definition = \
gx.ValidationDefinition(
  name: str, 
  data=batch_definition, 
  suite=suite
)

執行 Validation：

validation_results = \
validation_definition.run(
  batch_parameters={"dataframe": dataframe}
)

檢查 Validation 結果：

validation_results.success
validation_results.describe()

一起來練習吧！

Great Expectations 資料品質入門