概要

Databricks にて Azure SQL Dataabse のデータ型と対応する Spark Dataframe のデータ型の調査方法を共有します。Azure SQL Database にて一時テーブルを作成し、そのテーブルから Spark Dataframe を作成する方法を実施することで容易に確認できます。下記が Spark Dataframe のデータ型です。float(24) と float(25) で Spark のデータ型が変わるなどの仕様があり、すべてのデータ型の調査しているわけではないです。今後の調査を継続的に実施するために実施手順を共有します。

root
 |-- bit_col: boolean (nullable = true)
 |-- tinyint_col: integer (nullable = true)
 |-- smallint_col: short (nullable = true)
 |-- int_col: integer (nullable = true)
 |-- bigint_col: long (nullable = true)
 |-- numeric_col: decimal(18,0) (nullable = true)
 |-- smallmoney_col: decimal(10,4) (nullable = true)
 |-- money_col: decimal(19,4) (nullable = true)
 |-- decimal_col: decimal(18,6) (nullable = true)
 |-- float_24_col: float (nullable = true)
 |-- float_25_col: double (nullable = true)
 |-- float_53_col: double (nullable = true)
 |-- real_col: float (nullable = true)
 |-- char_col: string (nullable = true)
 |-- varchar_col: string (nullable = true)
 |-- text_col: string (nullable = true)
 |-- nchar_col: string (nullable = true)
 |-- nvarchar_col: string (nullable = true)
 |-- ntext_col: string (nullable = true)
 |-- date_col: date (nullable = true)
 |-- datetimeoffset_col: string (nullable = true)
 |-- datetime_col: timestamp (nullable = true)
 |-- datetime2_col: timestamp (nullable = true)
 |-- time_col: timestamp (nullable = true)
 |-- binary_col: binary (nullable = true)
 |-- varbinary_col: binary (nullable = true)
 |-- image_col: binary (nullable = true)
 |-- varchar_max_col: string (nullable = true)
 |-- nvarchar_max_col: string (nullable = true)
 |-- varbinary_max_col: binary (nullable = true)
 |-- hierarchyid_col: binary (nullable = true)
 |-- xml_col: string (nullable = true)
 |-- uniqueidentifier_col: string (nullable = true)

検証方法

Azure SQL Database の接続情報を定義

server = "ads-01"  ## サーバー名
database = "ads-01"  ## データベース名
username = dbutils.secrets.get(scope="qiita", key="sql_authentificate_user")  ## Azure SQL DB のユーザー名
password = dbutils.secrets.get(scope="qiita", key="sql_authentificate_password")  ## Azure SQL DB のパスワード

Azure SQL Database にて一時テーブルを作成後に Spark Dataframe を作成

# CREATE TABLE の実行
sql = """
IF EXISTS (SELECT * FROM sys.objects WHERE object_id = OBJECT_ID(N'#create_tble_01') AND type in (N'U'))
	DROP TABLE #create_tble_01
;
CREATE TABLE #create_tble_01
(    
    -- 正確な数値型 (Exact numerics)
    bit_col BIT,
    tinyint_col TINYINT,
    smallint_col SMALLINT,
    int_col INT,
    bigint_col BIGINT,
    numeric_col NUMERIC(18, 0),
    smallmoney_col SMALLMONEY,
    money_col MONEY,
    decimal_col DECIMAL(18, 6),

    -- 近似数値型 (Approximate numerics):
    float_24_col FLOAT(24),
    float_25_col FLOAT(25),
    float_53_col FLOAT(53),
    real_col real,

    -- 文字列型 (Character strings):
    char_col CHAR(10),
    varchar_col VARCHAR(100),
    text_col text,

    -- Unicode 文字列型 (Unicode character strings)
    nchar_col NCHAR(10),
    nvarchar_col NVARCHAR(100),
    ntext_col ntext,

    -- 日付と時刻型 (Date and time):
    date_col DATE,
    datetimeoffset_col DATETIMEOFFSET,
    datetime_col DATETIME,
    datetime2_col DATETIME2,
    time_col TIME,

    -- バイナリ文字列型 (Binary strings):
    binary_col BINARY(50),
    varbinary_col VARBINARY(50),
    image_col IMAGE,

    -- 大きな値のデータ型 (Large value data types)
    varchar_max_col VARCHAR(MAX),
    nvarchar_max_col NVARCHAR(MAX),

    -- 大きなオブジェクトのデータ型 (Large object data types)
    varbinary_max_col VARBINARY(MAX),

    -- その他のデータ型 (Other data types):
    hierarchyid_col HIERARCHYID,
    xml_col XML,
    uniqueidentifier_col UNIQUEIDENTIFIER NOT NULL PRIMARY KEY DEFAULT NEWID()
);
"""

df = (
    spark.read.format("sqlserver")
    .option("host", f"{server}.database.windows.net")
    .option("user", username)
    .option("password", password)
    .option("database", database)
    .option("prepareQuery", sql)
    .option("dbTable", "#create_tble_01")
    .load()
)

df.printSchema()

root
 |-- bit_col: boolean (nullable = true)
 |-- tinyint_col: integer (nullable = true)
 |-- smallint_col: short (nullable = true)
 |-- int_col: integer (nullable = true)
 |-- bigint_col: long (nullable = true)
 |-- numeric_col: decimal(18,0) (nullable = true)
 |-- smallmoney_col: decimal(10,4) (nullable = true)
 |-- money_col: decimal(19,4) (nullable = true)
 |-- decimal_col: decimal(18,6) (nullable = true)
 |-- float_24_col: float (nullable = true)
 |-- float_25_col: double (nullable = true)
 |-- float_53_col: double (nullable = true)
 |-- real_col: float (nullable = true)
 |-- char_col: string (nullable = true)
 |-- varchar_col: string (nullable = true)
 |-- text_col: string (nullable = true)
 |-- nchar_col: string (nullable = true)
 |-- nvarchar_col: string (nullable = true)
 |-- ntext_col: string (nullable = true)
 |-- date_col: date (nullable = true)
 |-- datetimeoffset_col: string (nullable = true)
 |-- datetime_col: timestamp (nullable = true)
 |-- datetime2_col: timestamp (nullable = true)
 |-- time_col: timestamp (nullable = true)
 |-- binary_col: binary (nullable = true)
 |-- varbinary_col: binary (nullable = true)
 |-- image_col: binary (nullable = true)
 |-- varchar_max_col: string (nullable = true)
 |-- nvarchar_max_col: string (nullable = true)
 |-- varbinary_max_col: binary (nullable = true)
 |-- hierarchyid_col: binary (nullable = true)
 |-- xml_col: string (nullable = true)
 |-- uniqueidentifier_col: string (nullable = true)