Magnus Lycka wrote: > This won't help if we have problems keeping the whole > structure / call graph in memory at one time.
Ah, I had assumed that the main problem was just that the resulting DOM was too big. Still, I don't think there would be a problem if you just constructed a very small tree and then wrote out XML based upon that, instead of using an external library for this. For example, if your only elements are node and leaf, you can store that distinction in 1 bit, as opposed to a library's naive implementation using the strings "node" and "leaf", probably 40 bits in each case. Similarly you can calculate the aggregate attributes on the fly when it comes to outputting the tree instead of storing them. -- Ben Sizer -- http://mail.python.org/mailman/listinfo/python-list
