>On Wed, 2006-04-19 at 08:24, M.-A. Lemburg wrote: >> I've found the qmark style to be sufficient for everything
This is more about convenience (same as Python positional arguments vs. keyword arguments). Celko's book on SQL tree structures is full of examples using :names, with the same name appearing sevveral times in the same query. But I guess it's not that difficult to write an SQL tokenizer that can translate named to qmark and build a map from name to indices. > >I agree. Considering that many modules already support qmark, and those >that currently use 'format'/'pyformat' would have to move to something >else if these are dropped, I'd like to suggest that supporting 'qmark' >style should become a required feature. Modules would be free to also >implement 'named' and/or 'numeric', but supporting 'qmark' should be >mandatory. But how should a module support multiple notations? - by tokenizing any SQL statement to check which paramstyle was used - by having an option on the connection or the cursor - by having an additional method which requires qmark placeholders (this one would be backwards compatible) And while we're discussing annoying differences between drivers: aren't there some drivers that define def execute (stmt, *args) instead of def execute (stmt, args = []) Daniel -- Daniel Dittmar SAP Labs Berlin [EMAIL PROTECTED] _______________________________________________ DB-SIG maillist - [email protected] http://mail.python.org/mailman/listinfo/db-sig
