[Sqlalchemy-tickets] Issue #4163: cx_oracle requires method to force use of NCHAR and setinputsizes for strings (zzzeek/sqlalchemy)

[Michael B. <iss...@bi...>]

New issue 4163: cx_oracle requires method to force use of NCHAR and setinputsizes for strings
https://bitbucket.org/zzzeek/sqlalchemy/issues/4163/cx_oracle-requires-method-to-force-use-of

Michael Bayer:

thread at:

https://groups.google.com/d/msg/sqlalchemy/vsJBLakBP4Y/PVz5zSKCAAAJ

Anthony's info for a similar issue at:

https://github.com/oracle/python-cx_Oracle/issues/119

potential APIs:

    e = create_engine("oracle://...", use_nchar_for_unicode=True)

or:


```
#!python

from sqlalchemy.dialects.oracle import NVARCHAR2

...   Column('data', NVARCHAR2(50, use_nchar=True))
```

A more open-ended approach, since we occasionally have issues with setinputsizes being needed or not:


```
#!python

e = create_engine("oracle://...")
e.dialect.setinputsizes(Unicode, cx_Oracle.NCHAR)

```

the third approach could also accommodate for user-defined types.   it would need some changes in DefaultDialect.set_input_sizes as well.

it's not that easy to let users just override get_dbapi_type() because that is only called upon the "impl" type that is mostly dialect-private.





the former is way easier to use but I don't know if people are going to want this applied to all datatypes across the board.