add VARCHAR, NVARCHAR placeholder types to SQL Server dialect, support first class compilation

[Asher Leon Manning]

At line 2523 in sqlalchemy\sql\compiler ,

text += "(%d)" % type_.length

causes a TypeError, as the '%d' expects a number, not 'max'

[Mike Bayer]

please specify what database backend you are targeting.

reference: https://gerrit.sqlalchemy.org/#/c/19/

needs tests if someone has resources to contribute

[Asher Leon Manning]

I encountered the bug with SQL Server 2008, but 'max' is allowed in all versions of SQL Server since 2005.

[Mike Bayer]

If this is SQL Server, please use sqlalchemy.dialects.mssql.VARCHAR() and don't specify a length.

[Asher Leon Manning]

It should not be necessary to make modifications to a database to use SQLAlchemy on it.

The error was encountered in the following code

    connectstring =  ('mssql+pyodbc:///?odbc_connect=DRIVER.....')
    title = 'table_title'
    en = sqlalchemy.create_engine(connectstring)
    md= sqlalchemy.MetaData()
    table = sqlalchemy.Table('%s' %title, md, autoload=True, autoload_with=en)
    for c in table.columns:
        print c.type

If the database contains a column of type Integer(max) or Varchar(max), etc, it will throw this error.

[Mike Bayer]

ah! a code sample illustrating a bug ! well that is different! moment

[Mike Bayer]

the SQL Server dialect does not contain VARCHAR or NVARCHAR at all, as "max" is not part of its implementation. so this disables the ability to select for the right compiler.

[Mike Bayer]

OK nevermind, this was handled differently as part of #3504. that is, we don't leak the "max" detail into the type.

[Mike Bayer]

Duplicate of #3504.

[Mike Bayer]

please use:

print repr(c.type)

or:

from sqlalchemy.dialects import mssql
print c.type.compile(dialect=mssql.dialect())

until you can upgrade to 1.1.