Intro to schema feature

Why We Need Schema

Some NoSQL databases or message queue are not strongly limited schema, so the schema cannot be obtained through the api. At this time, a schema needs to be defined to convert to TableSchema and obtain data.

SchemaOptions

We can use SchemaOptions to define schema, the SchemaOptions contains some configs to define the schema. e.g. columns, primaryKey, constraintKeys.

schema = {
    table = "database.schema.table"
    schema_first = false
    comment = "comment"
    columns = [
    ...
    ]
    primaryKey {
    ...
    }
    
    constraintKeys {
    ...
    }
}

table

The table full name of the table identifier which the schema belongs to, it contains database, schema, table name. e.g. database.schema.table, database.table, table.

schema_first

Default is false.

If the schema_first is true, the schema will be used first, this means if we set table = "a.b", a will be parsed as schema rather than database, then we can support write table = "schema.table".

comment

The comment of the CatalogTable which the schema belongs to.

Columns

Columns is a list of configs used to define the column in schema, each column can contains name, type, nullable, defaultValue, comment field.

columns = [
       {
          name = id
          type = bigint
          nullable = false
          columnLength = 20
          defaultValue = 0
          comment = "primary key id"
       }
]

Field	Required	Default Value	Description
name	Yes	-	The name of the column
type	Yes	-	The data type of the column
nullable	No	true	If the column can be nullable
columnLength	No	0	The length of the column which will be useful when you need to define the length
columnScale	No	-	The scale of the column which will be useful when you need to define the scale
defaultValue	No	null	The default value of the column
comment	No	null	The comment of the column

What type supported at now

Data type	Value type in Java	Description
string	java.lang.String	string
boolean	java.lang.Boolean	boolean
tinyint	java.lang.Byte	-128 to 127 regular. 0 to 255 unsigned*. Specify the maximum number of digits in parentheses.
smallint	java.lang.Short	-32768 to 32767 General. 0 to 65535 unsigned*. Specify the maximum number of digits in parentheses.
int	java.lang.Integer	All numbers from -2,147,483,648 to 2,147,483,647 are allowed.
bigint	java.lang.Long	All numbers between -9,223,372,036,854,775,808 and 9,223,372,036,854,775,807 are allowed.
float	java.lang.Float	Float-precision numeric data from -1.79E+308 to 1.79E+308.
double	java.lang.Double	Double precision floating point. Handle most decimals.
decimal	java.math.BigDecimal	Double type stored as a string, allowing a fixed decimal point.
null	java.lang.Void	null
bytes	byte[]	bytes
date	java.time.LocalDate	Only the date is stored. From January 1, 0001 to December 31, 9999.
time	java.time.LocalTime	Only store time. Accuracy is 100 nanoseconds.
timestamp	java.time.LocalDateTime	Stores a unique number that is updated whenever a row is created or modified. timestamp is based on the internal clock and does not correspond to real time. There can only be one timestamp variable per table.
row	org.apache.seatunnel.api.table.type.SeaTunnelRow	Row type, can be nested.
map	java.util.Map	A Map is an object that maps keys to values. The key type includes int string boolean tinyint smallint bigint float double decimal date time timestamp null , and the value type includes int string boolean tinyint smallint bigint float double decimal date time timestamp null array map row.
array	ValueType[]	A array is a data type that represents a collection of elements. The element type includes int string boolean tinyint smallint bigint float double.

How to declare type supported

SeaTunnel provides a simple and direct way to declare basic types. Basic type keywords include string, boolean, tinyint, smallint, int, bigint, float, double, date, time, timestamp, and null. The keyword names for basic types can be used directly as type declarations, and SeaTunnel is case-insensitive to type keywords. For example, if you need to declare a field with integer type, you can simply define the field as int or "int".

The null type declaration must be enclosed in double quotes, like "null". This approach helps avoid confusion with HOCON's null type which represents undefined object.

When declaring complex types (such as decimal, array, map, and row), pay attention to specific considerations.

• When declaring a decimal type, precision and scale settings are required, and the type definition follows the format decimal(precision, scale). It's essential to emphasize that the declaration of the decimal type must be enclosed in "; you cannot use the type name directly, as with basic types. For example, when declaring a decimal field with precision 10 and scale 2, you specify the field type as "decimal(10,2)".
• When declaring an array type, you need to specify the element type, and the type definition follows the format array<T>, where T represents the element type. The element type includes int,string,boolean,tinyint,smallint,bigint,float and double. Similar to the decimal type declaration, it also be enclosed in ". For example, when declaring a field with an array of integers, you specify the field type as "array<int>".
• When declaring a map type, you need to specify the key and value types. The map type definition follows the format map<K,V>, where K represents the key type and V represents the value type. K can be any basic type and decimal type, and V can be any type supported by SeaTunnel. Similar to previous type declarations, the map type declaration must be enclosed in double quotes. For example, when declaring a field with map type, where the key type is string and the value type is integer, you can declare the field as "map<string, int>".
• When declaring a row type, you need to define a HOCON object to describe the fields and their types. The field types can be any type supported by SeaTunnel. For example, when declaring a row type containing an integer field a and a string field b, you can declare it as {a = int, b = string}. Enclosing the definition in " as a string is also acceptable, so "{a = int, b = string}" is equivalent to {a = int, c = string}. Since HOCON is compatible with JSON, "{\"a\":\"int\", \"b\":\"string\"}" is equivalent to "{a = int, b = string}".

Here is an example of complex type declarations:

schema {
  fields {
    c_decimal = "decimal(10, 2)"
    c_array = "array<int>"
    c_row = {
        c_int = int
        c_string = string
        c_row = {
            c_int = int
        }
    }
    # Hocon style declare row type in generic type
    map0 = "map<string, {c_int = int, c_string = string, c_row = {c_int = int}}>"
    # Json style declare row type in generic type
    map1 = "map<string, {\"c_int\":\"int\", \"c_string\":\"string\", \"c_row\":{\"c_int\":\"int\"}}>"
  }
}

PrimaryKey

Primary key is a config used to define the primary key in schema, it contains name, columns field.

primaryKey {
    name = id
    columns = [id]
}

Field	Required	Default Value	Description
name	Yes	-	The name of the primaryKey
columns	Yes	-	The column list in the primaryKey

ConstraintKeys

Constraint keys is a list of config used to define the constraint keys in schema, it contains constraintName, constraintType, constraintColumns field.

constraintKeys = [
      {
         constraintName = "id_index"
         constraintType = KEY
         constraintColumns = [
            {
                columnName = "id"
                sortType = ASC
            }
         ]
      },
   ]

Field	Required	Default Value	Description
constraintName	Yes	-	The name of the constraintKey
constraintType	No	KEY	The type of the constraintKey
constraintColumns	Yes	-	The column list in the primaryKey, each column should contains constraintType and sortType, sortType support ASC and DESC, default is ASC

What constraintType supported at now

ConstraintType	Description
INDEX_KEY	key
UNIQUE_KEY	unique key

How to use schema

Recommended

source {
  FakeSource {
    parallelism = 2
    result_table_name = "fake"
    row.num = 16
    schema {
        table = "FakeDatabase.FakeTable"
        columns = [
           {
              name = id
              type = bigint
              nullable = false
              defaultValue = 0
              comment = "primary key id"
           },
           {
              name = name
              type = "string"
              nullable = true
              comment = "name"
           },
           {
              name = age
              type = int
              nullable = true
              comment = "age"
           }
       ]
       primaryKey {
          name = "id"
          columnNames = [id]
       }
       constraintKeys = [
          {
             constraintName = "unique_name"
             constraintType = UNIQUE_KEY
             constraintColumns = [
                {
                    columnName = "name"
                    sortType = ASC
                }
             ]
          },
       ]
      }
    }
}

Deprecated

If you only need to define the column, you can use fields to define the column, this is a simple way but will be remove in the future.

source {
  FakeSource {
    parallelism = 2
    result_table_name = "fake"
    row.num = 16
    schema = {
      fields {
        id = bigint
        c_map = "map<string, smallint>"
        c_array = "array<tinyint>"
        c_string = string
        c_boolean = boolean
        c_tinyint = tinyint
        c_smallint = smallint
        c_int = int
        c_bigint = bigint
        c_float = float
        c_double = double
        c_decimal = "decimal(2, 1)"
        c_bytes = bytes
        c_date = date
        c_timestamp = timestamp
      }
    }
  }
}

When we should use it or not

If there is a schema configuration project in Options,the connector can then customize the schema. Like Fake Pulsar Http source connector etc.