Schema migrations and reflect_table

[lukasz-madon]

There are a few examples of migrations in the documentation and tutorials. There seems to be two ways of doing migrations

1. Explicit SQL
2. ORM via table()

   users = table('users',
        column('first_name', String),
        column('last_name', String),
        column('full_name', String)
    )
    conn = op.get_bind()
    conn.execute(
        users.update().values(
            full_name=sa.func.concat(users.c.first_name, ' ', users.c.last_name)
        )
    )

There 3rd option would be to use reflect_table

users = op.reflect_table('users')
users.update().values(

Is reflection OK to use in this case? What are the pros and cons? Can reflect_table bit me with without explicit typing?

[CaselIT]

Hi,

Well the main drawback of reflect table is that you need to have a db available, meaning you can't use alembic in offline mode, ie to generate sql files to apply to the db by some other mean.

Other than that I would think it should be safe if you use only alembic to modify the schema of the db

[lukasz-madon]

I guess you also lose typechecking - reflection can make conversion implicit.

[CaselIT]

In a way yes, but at the moment table() or Table don't offer much type checking, at least in python using mypy and similar tools.
If you meant more the sql type specified into table etc, then yes when reflecting a table you get the type in the db