Fredrik Lundh wrote: > Delaney, Timothy (Tim) wrote: > >> The most important problem here is that you've lost the stack trace from >> the original exception. > > which, in many cases, is a good thing, especially if you replace "int" > with a call to some low-level support library. if I call API "foo" to > open a data file of some kind, I'm not necessarily interested in an > "unsubscriptable error" exception on line 129 in barpath.py, nor is it > helping me figure out what's wrong in my program.
I always prefer to have the stack trace, since it *is* useful to me to figure out what is wrong with my program (or possibly, the library I'm using). > or do you expose internal implementation details to your end users > too, to make sure they don't lose any information? No - but I do log any errors, with stack traces, and as much useful data as possible. That way it's more likely that I can get a useful bug report (they send the log file). Replacing an exception *may* not get in the way if you're wrapping a single statement in the exception. I can't think of another case where it would not get in the way. Chained exceptions change this somewhat, because you don't lose the original information. Tim Delaney -- http://mail.python.org/mailman/listinfo/python-list
