sqlalchemy.sql.ddl.CreateColumn

Represent a _schema.Column as rendered in a CREATE TABLE statement, via the .CreateTable construct.

This is provided to support custom column DDL within the generation of CREATE TABLE statements, by using the compiler extension documented in :ref:`sqlalchemy.ext.compiler_toplevel` to extend .CreateColumn.

Typical integration is to examine the incoming _schema.Column object, and to redirect compilation if a particular flag or condition is found:

from sqlalchemy import schema
from sqlalchemy.ext.compiler import compiles

@compiles(schema.CreateColumn)
def compile(element, compiler, **kw):
    column = element.element

    if "special" not in column.info:
        return compiler.visit_create_column(element, **kw)

    text = "%s SPECIAL DIRECTIVE %s" % (
            column.name,
            compiler.type_compiler.process(column.type)
        )
    default = compiler.get_column_default_string(column)
    if default is not None:
        text += " DEFAULT " + default

    if not column.nullable:
        text += " NOT NULL"

    if column.constraints:
        text += " ".join(
                    compiler.process(const)
                    for const in column.constraints)
    return text

The above construct can be applied to a _schema.Table as follows:

from sqlalchemy import Table, Metadata, Column, Integer, String
from sqlalchemy import schema

metadata = MetaData()

table = Table('mytable', MetaData(),
        Column('x', Integer, info={"special":True}, primary_key=True),
        Column('y', String(50)),
        Column('z', String(20), info={"special":True})
    )

metadata.create_all(conn)

Above, the directives we've added to the _schema.Column.info collection will be detected by our custom compilation scheme:

CREATE TABLE mytable (
        x SPECIAL DIRECTIVE INTEGER NOT NULL,
        y VARCHAR(50),
        z SPECIAL DIRECTIVE VARCHAR(20),
    PRIMARY KEY (x)
)

The .CreateColumn construct can also be used to skip certain columns when producing a CREATE TABLE. This is accomplished by creating a compilation rule that conditionally returns None. This is essentially how to produce the same effect as using the system=True argument on _schema.Column, which marks a column as an implicitly-present "system" column.

For example, suppose we wish to produce a _schema.Table which skips rendering of the PostgreSQL xmin column against the PostgreSQL backend, but on other backends does render it, in anticipation of a triggered rule. A conditional compilation rule could skip this name only on PostgreSQL:

from sqlalchemy.schema import CreateColumn

@compiles(CreateColumn, "postgresql")
def skip_xmin(element, compiler, **kw):
    if element.element.name == 'xmin':
        return None
    else:
        return compiler.visit_create_column(element, **kw)


my_table = Table('mytable', metadata,
            Column('id', Integer, primary_key=True),
            Column('xmin', Integer)
        )

Above, a .CreateTable construct will generate a CREATE TABLE which only includes the id column in the string; the xmin column will be omitted, but only against the PostgreSQL backend.