Re: RFC: Extend psycopg2.connect to accept all valid parameters?

On Thu, Nov 17, 2011 at 11:41 AM, Federico Di Gregorio <fog@dndg.it> wrote:
> On 17/11/11 12:39, Daniele Varrazzo wrote:
>> On Thu, Nov 17, 2011 at 11:18 AM, Federico Di Gregorio <fog@dndg.it> wrote:
>>
>>> > Never said doing it in Python is wrong. In fact anything that isn't
>>> > time-critical (type conversions, etc.) at this point is OK in Python.
>> I was also thinking that having the pair connect()/_connect() is
>> perfect for regression testing: _connect() can be replaced with a stub
>> to test the arguments conversion without really connecting.
>
> Wunderful. But please don't rename the C function. Just "import as", to
> avoid breaking API (not that I ever encountered Python code using
> _psycopg.so directly but one never knows...)

I wanted to rename it because I've dropped its support to the keyword
arguments: the interface (which is not an API: if sb is using
_psycopg.so he is doing at his own risk) is broken anyway. And because
we are not going to use the C keyword codepath, I want to drop it
altogether: it is not going to be maintained anymore.

-- Daniele