On 22 July 2013 14:15, Dan Douglas <[email protected]> wrote: > On Monday, July 22, 2013 01:28:21 PM Cedric Blancher wrote: >> Forwarding the proposal. Wendy's idea to use varname.__ as name to >> reference a compound/type variable's parent is IMO *great*. > > In the general case of compounds it's fine. In the case of objects created > from user-defined types, it's far better and customary to have a reference to > the object, and then refer to sub-variables relative to that (from the outside > in, not from the inside working out.).
I think it still has merit for both cases. But IMO it should help with the major issue that building large trees inside a compound variable where a single compound variable at the base contains a compound variable array, which itself harbours many more nesting levels of compound variables. The current performance of such a construct SUCKS compared to java or python (or compared to a real Unix filesystem), regardless of the cache David has implemented, because constructing and using such a tree always requires the absolute variable name. Using the Wendy/Cedric proposal of nameref parent=varname.__ to reference the parent compound or type variable is a very good way to (finally) introduce relative variable names top build and access variable trees. Lionel _______________________________________________ ast-developers mailing list [email protected] http://lists.research.att.com/mailman/listinfo/ast-developers
