Mike Looijmans
nlv11281 at natlab.research.philips.com
Thu Oct 26 08:38:54 EDT 2006
Being the one who moved all that code around, I'm probably the one to explain why. In the original version, there was a 'bad' way of detecting whether objects were files or not. If one used the callback mechanism, and provided a file-like object, this object would not be recognised as a file by that code, and that either resulted in the whole file being read into memory as a string or in some obscure exception being thrown. Mike Looijmans Philips Natlab / Topic Automation Graham Dumpleton wrote: > I'd probably hold off trying to work out any patches yourself. The more > I look at FieldStorage the less I understand why some of the changes > were made. I'm not talking about how a dictionary based index was > added, but more how some of the code was shifted around when it > probably didn't need to, with the result being that the code then became > incompatible with versions of Trac which were out at the time. It would > not have been that hard to maintain compatibility with Trac and still get > the desired gains they were after. I can see myself therefore doing a bit > or reorganisation of the code to fix the identified issue, but also > reinstate > Trac compatibility.
|