On Jan 28, 2014, at 10:57 PM, Erich Blume <[email protected]> wrote:

> Section 2.1 ( http://www.sqlite.org/datatype3.html ) has a deterministic 
> algorithm for transforming an unknown column type in to it's inverse-affinity 
> real type. It seems to me that we could modify the code in base.py around 820 
> (I'm at home and working from memory) to implement that algorithm if the 
> chosen schema is sqlite and if the column type is not found in ischema_names. 
> I'd be happy to cook up a patch for that if it's not too objectionable to 
> anyone.

if that algo is global to all types then yes, I’ll accept a pull request for 
that.  The algo should come up with one of the five affinities as that name 
INTEGER/TEXT/NONE/REAL/NUMERIC and then pull those right out of ischema_names.

> 
> I'll try and see if I can find out how the author created this database with 
> the incorrect column type names. If s/he used an official binary to create it 
> then I feel it's probably important for SQLAlchemy to support it.

it was likely just some random code somewhere adapted from a different 
database.  nevertheless the “DOUBLE” type is very common in SQL so it should be 
supported.

Attachment: signature.asc
Description: Message signed with OpenPGP using GPGMail

Reply via email to