The current logic in Felis's sql module assumes that autoincrement
will be None if it is not set explicitly so for now match this logic
in the datamodel.

Add a module with a new class that encapsulates the creation of
sqlalchemy MetaData from a Pydantic Schema object, along with a test
that dumps the corresponding SQL DDL statements to a file. The logic
for translating into SQA was primarily derived from the existing
visitor class in sql.py, which this is intended to replace eventually.

Add a "create" command that uses the metadata module instead of the
sql module.

Add flags for specifying an alternate schema name and creating the
schema in the database if it does exist already. For MySQL, a new
database will be created. In PostgreSQL, a new schema will be created
within the current database.

Add several new flags to the CLI for dropping a schema if it exists and
creating one if it does not exist.

Add methods implementing this functionality to the metadata module.

This is a generic schema representing sales data with customers and
orders for testing purposes.

Add a test case which loads a simple YAML test file into a schema,
creates SQLAlchemy Metadata from the object, creates the database
objects within an in-memory SQLite database, and then checks that the
reflected MetaData is the same as that created from the schema.

This value is not handled correctly as it ends up being quoted. It will
be fixed on another branch.

This is used by the metadata module to handle a SQA engine or mock
connection. SQL is executed in a with statement for an engine but not
for a mock connection, as it does not support this.

The sql module is being deleted but the variant handling utilities
need to be kept.

Remove the sql module which is replaced by the metadata module.

Remove the "create-all" command from the CLI which is replaced by
"create".

Remove the InsertDump class from the cli module, as it was moved into
the metadata module.

Instead of having a class with internal functions for building a
MetaData object, use a builder pattern for a cleaner interface.

Instead of not setting the schema name, use "main" instead which works
for sqlite. In case a host name is not set on the database engine URL,
force a dry run using the supplied variant.

Test that the metadata from the builder and that reflected from the
database have equivalent index and constraint objects. Add some
additional constraints to the YAML test file.

In some cases, it is not desirable to apply the schema name to either
the metadata or the tables. Add two flags to the builder for
controlling this behavior. The flags are true by default and can be
set to false if the user does not want the schema to be applied.

Change the regular expression to match on numbers rather than any
character between the parentheses.

Fix mypy issue where TextIOBase does not have a visible name attribute.

Provide a utility method for getting the SQLAlchemy datatype from a
column object in the Pydantic model, including the variant type
handling.

This test reads in a test schema, creates a SQLAlchemy MetaData object
from it, and then compares the schema to the metadata to check that
they have the equivalent, expected information.

Remove exception checking for each constraint type as it is
unnecessary. In the unlikely case that a constraint was not built
properly, throw an exception.

This field is required so that a PostgreSQL datatype can be optionally
provided in the column information.

This setting confuses mypy and the syntax seems clearer when using the
enum object instead.

All of the other fields use snake case and aliases for camelcase
variables, so change primary key to match.