Amazon Redshift 將不再支援從修補程式 198 開始建立新的 Python UDFs。現有 Python UDF 將繼續正常運作至 2026 年 6 月 30 日。如需詳細資訊，請參閱[部落格文章](https://aws.amazon.com/blogs/big-data/amazon-redshift-python-user-defined-functions-will-reach-end-of-support-after-june-30-2026/)。
本文為英文版的機器翻譯版本，如內容有任何歧義或不一致之處，概以英文版為準。
# 在 Amazon Redshift 中使用半結構化資料的範例
下列範例示範如何在 Amazon Redshift 中使用 PartiQL 語法處理半結構化資料。您將建立範例資料表來載入一組半結構化資料，然後在各種不同的使用案例中查詢半結構化資料物件。
**注意**
建議您在處理 SUPER 資料類型之前，先設定 `enable_case_sensitive_identifier` 和 `enable_case_sensitive_super_attribute` 組態選項。如需詳細資訊，請參閱[enable\$1case\$1sensitive\$1identifier](r_enable_case_sensitive_identifier.md)及[enable\$1case\$1sensitive\$1super\$1attribute](r_enable_case_sensitive_super_attribute.md)。
## 載入半結構化資料
下列陳述式會建立範例資料表，並將範例 JSON 物件載入 `all_data` SUPER 欄。
```
DROP TABLE IF EXISTS test_json;
SET enable_case_sensitive_super_attribute TO true;
SET enable_case_sensitive_identifier TO true;
CREATE TABLE test_json (all_data SUPER);
INSERT INTO test_json VALUES (JSON_PARSE('
{
"data":{
"pnr":{
"type":"pnr",
"pnrid":"123PQRS-2024-09-20",
"bookingIdentifier":"123PQRS",
"version":"5",
"triggerType":"",
"events":[
{
"eventType":"UPDATED",
"type":"PART",
"id":"123PQRS-2024-09-20-HO-1"
},
{
"eventType":"CREATED",
"type":"ABC",
"id":"123PQRS-2024-09-20-OT-38"
}
],
"create":{
"pnrCreateDate":"2024-09-20T16:56:00Z",
"officeID":"OFFCID1234",
"officeIDCategory":"Email"
},
"lastModification":{
"dateTime":"2024-09-20T17:09:00Z"
},
"PARTDetails":[
{
"path":"1",
"TrainPARTs":[
{
"PARTID":"123PQRS-2024-09-20-HO-1",
"departure":{
"departureStation":"XYZ",
"departureTimeLocal":"2024-10-03T06:30:00",
"departureTimeGMT":"2024-10-03T10:30:00Z"
},
"arrival":{
"arrivalStation":"ABC",
"arrivalTimeLocal":"2024-10-03T08:20:00",
"arrivalTimeGMT":"2024-10-03T15:20:00Z"
},
"marketing":{
"carrierCode":"XX",
"TrainNumber":"100"
},
"operating":{
"carrierCode":"YY",
"TrainNumber":"100-A"
},
"status":"ON",
"aircraft":{
"code":"222"
},
"class":"WC",
"first":"Y",
"seating":[
]
}
]
}
],
"commuterInformation":[
{
"commuterID":"2",
"commuterPNR":"123PQRS-2024-09-20-RO-2",
"commuterTypeCode":"DOM",
"firstName":"JOHN",
"lastName":"MILLER"
}
],
"contactDetail":[
{
"emailContacts":[
{
"id":"123PQRS-2024-09-20-OT-4",
"contact":"JOHNMILLER@EXAMPLE.COM",
"purpose":[
"BUSINESS"
],
"commuter":[
"123PQRS-2024-09-20-RO-2"
],
"language":"EN"
},
{
"id":"123PQRS-2024-09-20-OT-5",
"contact":"HARVEYCORMIER@EXAMPLE.COM",
"purpose":[
"NOTIFICATION"
],
"commuter":[
"123PQRS-2024-09-20-RO-2"
],
"language":"EN"
}
]
},
{
"phoneContacts":[
{
"id":"123PQRS-2024-09-20-OT-3",
"contact":"1234567890",
"purpose":[
"NOTIFICATION"
],
"commuter":[
"123PQRS-2024-09-20-RO-2"
],
"language":""
}
]
},
{
"addressInfo":[
{
"id":"123PQRS-2024-09-20-OT-6",
"addressline":[
"112 PORT STREET"
],
"provinceState":"CA",
"cityName":"SAN JOSE",
"postalCode":"12345",
"countryCode":"USA",
"purpose":[
"MAILING"
],
"commuter":[
"123PQRS-2024-09-20-RO-2"
]
}
]
}
],
"PendingService":[
{
"id":"123PQRS-2024-09-20-OT-26",
"code":"MONO",
"status":"",
"text":"Broken Seat at Coach-No XYZ123 Seat-No 567",
"trainCode":"WC-1",
"TrainsArray":[
"123PQRS-2024-09-20-HO-1"
],
"commuter":[
"123PQRS-2024-09-20-RO-2"
]
},
{
"id":"123PQRS-2024-09-20-OT-27",
"code":"OTHS",
"status":"",
"text":"Broken Seat at Coach-No XYZ567 Seat-No 111",
"trainCode":"WC-1",
"TrainsArray":[
"123PQRS-2024-09-20-HO-1"
],
"commuter":[
"123PQRS-2024-09-20-RO-2"
]
},
{
"id":"123PQRS-2024-09-20-OT-28",
"code":"OTHS",
"status":"",
"text":"Broken Seat at Coach-No XYZ890 Seat-No 123",
"trainCode":"WC-1",
"TrainsArray":[
"123PQRS-2024-09-20-HO-1"
],
"commuter":[
"123PQRS-2024-09-20-RO-2"
]
},
{
"id":"123PQRS-2024-09-20-OT-29",
"code":"OTHS",
"status":"",
"text":"Broken Seat at Coach-No XYZ111 Seat-No 333",
"trainCode":"WC-1",
"TrainsArray":[
"123PQRS-2024-09-20-HO-1"
],
"commuter":[
"123PQRS-2024-09-20-RO-2"
]
}
],
"parts": [
{
"partname": "prop",
"manufacturer": "local parts co",
"quality": 2,
"price": 10.00
},
{
"partname": "prop",
"manufacturer": "big parts co",
"quality": null,
"price": 9.00
},
{
"partname": "prop",
"manufacturer": "small parts co",
"quality": 1,
"price": 12.00
},
{
"partname": "rudder",
"manufacturer": "local parts co",
"quality": 1,
"price": 2.50
},
{
"partname": "rudder",
"manufacturer": "big parts co",
"quality": 2,
"price": 3.75
},
{
"partname": "rudder",
"manufacturer": "small parts co",
"quality": null,
"price": 1.90
},
{
"partname": "wing",
"manufacturer": "local parts co",
"quality": null,
"price": 7.50
},
{
"partname": "wing",
"manufacturer": "big parts co",
"quality": 1,
"price": 15.20
},
{
"partname": "wing",
"manufacturer": "small parts co",
"quality": null,
"price": 11.80
}
],
"count_by_color": [
{
"quality": "high",
"red": 15,
"green": 20,
"blue": 7
},
{
"quality": "normal",
"red": 35,
"green": null,
"blue": 40
},
{
"quality": "low",
"red": 10,
"green": 23,
"blue": null
}
]
}
},
"id":"abcdefgh-ijklmnop-qrstuvwxyz123",
"mainIds":[
{
"ID":"pqrstuvwxyz-aabbcc123",
"Source":"NYC"
}
]
}
'));
```
## 查詢巢狀半結構化資料
下列陳述式使用 PartiQL 的點標記法來擷取 `pnrid` 欄位，該欄位在是最上層 `all_data` 物件內三層深的巢狀欄位。
```
select all_data.data.pnr.pnrid::varchar from test_json;
pnrid
--------------------
123PQRS-2024-09-20
```
下列陳述式使用 PartiQL 的括號標記法，只指定並擷取最上層物件內巢狀 `events` 陣列的第一個元素。
```
SELECT
all_data.data.pnr.events[0]
FROM test_json;
events
---------------------------------
{
"eventType":"UPDATED",
"type":"PART",
"id":"123PQRS-2024-09-20-HO-1"
}
```
下列陳述式只會從 `events` 陣列中擷取指定元素的 `eventType` 屬性。
```
SELECT
all_data.data.pnr.events[0].eventType
FROM test_json;
eventtype
-----------
"UPDATED"
```
下列陳述式
## 使用 `enable_case_sensitive_identifier` 和 `enable_case_sensitive_super_attribute` 搭配半結構化資料
下列範例顯示組態選項 [enable\$1case\$1sensitive\$1identifier](r_enable_case_sensitive_identifier.md) 和 [enable\$1case\$1sensitive\$1super\$1attribute](r_enable_case_sensitive_super_attribute.md) 用於查詢半結構化資料時有何不同。如需這些組態選項的詳細資訊，請參閱 [存取具有大寫和混合大小寫欄位名稱或屬性的 JSON 欄位](super-configurations.md#upper-mixed-case)。
在下列陳述式中，將兩個組態選項重設為其預設值 false，會使查詢傳回 NULL。
```
RESET enable_case_sensitive_identifier;
RESET enable_case_sensitive_super_attribute;
SELECT
all_data.data.pnr.events[0].eventType
FROM test_json;
eventtype
-----------
NULL
```
在下列範例中，您在區分大小寫的屬性前後加上雙引號並將 `enable_case_sensitive_identifier` 設為 true 之後，範例查詢就會傳回所需的結果。
```
RESET enable_case_sensitive_identifier;
RESET enable_case_sensitive_super_attribute;
SELECT
all_data.data.pnr.events[0]."eventType"
FROM test_json;
eventtype
-----------
NULL
SET enable_case_sensitive_identifier TO true;
SELECT
all_data.data.pnr.events[0]."eventType"
FROM test_json;
eventtype
-----------
"UPDATED"
```
在下列範例中，您將 `enable_case_sensitive_super_attribute` 設為 true 但未在區分大小寫的屬性前後加上雙引號之後，範例查詢就會傳回所需的結果。
```
RESET enable_case_sensitive_identifier;
RESET enable_case_sensitive_super_attribute;
SELECT
all_data.data.pnr.events[0].eventType
FROM test_json;
eventtype
-----------
NULL
SET enable_case_sensitive_super_attribute TO true;
SELECT
all_data.data.pnr.events[0].eventType
FROM test_json;
eventtype
-----------
"UPDATED"
```
## 篩選半結構化資料
下列陳述式在計算 `UPDATED` 類型事件數之陳述式的 WHERE 子句中使用 PartiQL 語法，以從陣列內擷取特定屬性的資料。您可以在通常參考欄的查詢中的任何部分使用此語法。
```
SELECT COUNT(*)
FROM test_json
WHERE all_data.data.pnr.events[0].eventType = 'UPDATED';
count
------
1
```
下列範例會在 GROUP BY 和 ORDER BY 子句中使用 PartiQL 的括號和點語法。
```
SELECT all_data.data.pnr.events[0].eventType::varchar,
COUNT(*)
FROM test_json
WHERE all_data.data.pnr.events[0].eventType IS NOT NULL
GROUP BY all_data.data.pnr.events[0].eventType
ORDER BY all_data.data.pnr.events[0].eventType;
eventtype | count
-----------+-------
"UPDATED" | 1
```
## 解除巢狀化半結構化資料
下列陳述式使用 PartiQL 聯結來解除巢狀化 `events` 陣列。請注意，即使陣列的索引數目並非靜態，此聯結仍有作用。
如需在 FROM 子句中使用 UNNEST 來解除巢狀化半結構化資料的範例，請參閱 [UNNEST 範例](r_FROM_clause-unnest-examples.md)。
```
SELECT
a.all_data.data.pnr.type::varchar type_info,
a.all_data.data.pnr.pnrid::varchar pnr_id ,
a.all_data.data.pnr.bookingIdentifier::varchar booking_id,
a.all_data.data.pnr.version::varchar version_info,
b.eventType::varchar event_type,
b.id::varchar event_id
FROM test_json a,
a.all_data.data.pnr.events b;
type_info | pnr_id | booking_id | version_info | event_type | event_id
-----------+---------------------+------------+--------------+------------+-------------------------
pnr | 123PQRS-2024-09-20 | 123PQRS | 5 | UPDATED | 123PQRS-2024-09-20-HO-1
pnr | 123PQRS-2024-09-20 | 123PQRS | 5 | CREATED | 123PQRS-2024-09-20-OT-38
```
## 解除巢狀化巢狀陣列
下列陳述式使用 PartiQL 聯結來解除巢狀化另一個陣列內的巢狀陣列。
如需在 FROM 子句中使用 UNNEST 來解除巢狀化半結構化資料的範例，請參閱 [UNNEST 範例](r_FROM_clause-unnest-examples.md)。
```
SELECT
a.all_data.data.pnr.type::varchar type_info,
a.all_data.data.pnr.pnrid::varchar pnr_id ,
a.all_data.data.pnr.bookingIdentifier::varchar booking_id,
a.all_data.data.pnr.version::varchar version_info,
d.id::varchar email_record_id,
d.contact::varchar email_contact,
e::varchar email_purpose,
f::varchar email_commuter
FROM test_json a,
a.all_data.data.pnr.contactDetail c,
c."emailContacts" d,
d.purpose e,
d.commuter f;
type_info | pnr_id | booking_id | version_info | email_record_id | email_contact | email_purpose | email_commuter
-----------+---------------------+------------+--------------+-------------------------+---------------------------+---------------+-------------------------
pnr | 123PQRS-2024-09-20 | 123PQRS | 5 | 123PQRS-2024-09-20-OT-4 | JOHNMILLER@EXAMPLE.COM | BUSINESS | 123PQRS-2024-09-20-RO-2
pnr | 123PQRS-2024-09-20 | 123PQRS | 5 | 123PQRS-2024-09-20-OT-5 | HARVEYCORMIER@EXAMPLE.COM | NOTIFICATION | 123PQRS-2024-09-20-RO-2
```
## 在子查詢中使用半結構化資料
下列陳述式在 WHERE 子句中使用子查詢，以僅傳回先前範例中結果的子區段。
```
SELECT
a.all_data.data.pnr.type::varchar type_info,
a.all_data.data.pnr.pnrid::varchar pnr_id ,
a.all_data.data.pnr.bookingIdentifier::varchar booking_id,
a.all_data.data.pnr.version::varchar version_info,
d.id::varchar email_record_id,
d.contact::varchar email_contact
FROM test_json a,
a.all_data.data.pnr.contactDetail c,
c."emailContacts" d
WHERE (SELECT COUNT(*) FROM d.purpose e WHERE e = 'BUSINESS') > 0;
type_info | pnr_id | booking_id | version_info | email_record_id | email_contact | email_purpose | email_commuter
-----------+---------------------+------------+--------------+-------------------------+---------------------------+---------------+-------------------------
pnr | 123PQRS-2024-09-20 | 123PQRS | 5 | 123PQRS-2024-09-20-OT-4 | JOHNMILLER@EXAMPLE.COM | BUSINESS | 123PQRS-2024-09-20-RO-2
```
## 使用半結構化資料彙總查詢
下列陳述式使用 COUNT 函式來彙總 `PendingService` 陣列中元素的數目。
```
SELECT
a.all_data.data.pnr.type::varchar type_info,
a.all_data.data.pnr.pnrid::varchar pnr_id ,
a.all_data.data.pnr.bookingIdentifier::varchar booking_id,
a.all_data.data.pnr.version::varchar version_info,
COUNT(*) AS total_pending_service
FROM test_json a,
a.all_data.data.pnr.PendingService c
GROUP BY 1,2,3,4;
type_info | pnr_id | booking_id | version_info | total_pending_service
-----------+--------------------+------------+--------------+-----------------------
pnr | 123PQRS-2024-09-20 | 123PQRS | 5 | 4
```
## 在具體化視觀表中使用半結構化資料
下列陳述式使用先前範例中的陳述式來建立具體化視觀表。當基礎資料表取得新資料時，具體化視觀表會自動重新整理待處理的服務數量。
```
CREATE MATERIALIZED VIEW mv_total_pending_service
AUTO REFRESH YES
AS
SELECT
a.all_data.data.pnr.type::varchar type_info,
a.all_data.data.pnr.pnrid::varchar pnr_id ,
a.all_data.data.pnr.bookingIdentifier::varchar booking_id,
a.all_data.data.pnr.version::varchar version_info,
COUNT(*) AS total_pending_service
FROM test_json a,
a.all_data.data.pnr.PendingService c
GROUP BY 1,2,3,4;
```
## 搭配半結構化資料使用 PIVOT 和 UNPIVOT
下列陳述式在 `partname` 欄上使用 PIVOT 來傳回每個部分的平均價格。
```
SELECT *
FROM
(
SELECT
c.partname::varchar, c.price
FROM test_json a,
a.all_data.data.pnr.parts c)
PIVOT (AVG(price) for partname IN ('prop', 'rudder', 'wing'));
prop | rudder | wing
------------+--------------------+--------
10.33 | 2.71 | 11.50
```
在上一個範例中，結果會轉換為資料欄。下列範例顯示以列 (不是欄) 傳回平均價格的 GROUP BY 查詢。
```
SELECT partname, avg(price)
FROM (
SELECT
c.partname::varchar, c.price
FROM test_json a,
a.all_data.data.pnr.parts c)
WHERE partname IN ('prop', 'rudder', 'wing')
GROUP BY partname;
partname | avg
----------+-------
prop | 10.33
rudder | 2.71
wing | 11.50
```
以下是以 `manufacturer` 作為隱含欄的 PIVOT 範例。
```
SELECT *
FROM (
SELECT
c.quality, c.manufacturer::varchar
FROM test_json a,
a.all_data.data.pnr.parts c) PIVOT (
count(*) FOR quality IN (1, 2, NULL)
);
manufacturer | 1 | 2 | null
-------------------+----+----+------
local parts co | 1 | 1 | 1
big parts co | 1 | 1 | 1
small parts co | 1 | 0 | 2
```
以下是 `quality` 欄上的 UNPIVOT 範例。
```
SELECT *
FROM
(
SELECT
c.quality as quality
FROM test_json a,
a.all_data.data.pnr.parts c)
UNPIVOT (cnt FOR column_header IN (quality));
column_header | cnt
-----------------+----
quality | 2
quality | 1
quality | 1
quality | 2
quality | 1
```