That isn't what I meant. If there was a a point (and I'm not really sure that I'm even trying to make one), the point was that Google makes heavy use of reduce-like functionality, essentially implementing a distributed reduce across a cluster. From what I hear, they use a lot of Python and hired van Rossum for a reason. It just seems odd (don't read anything into this than mere cocked eyebrows) that the language designer that they hired and obviously have a lot of trust in would decide that reduce was essentially pointless. Google's distributed reduce seems to point in opposite way.
However, if reduce could be rolled into the list comprehension syntax, that would be even better. Or take it that extra step and roll a grouping functionality in there too, then you would have map, reduce, group, and filter all in one construct. You could call it select (joins are merely indexes into other structures). -j Steve Holden wrote: > Jason Nordwick wrote: >> I use reduce to also do indexing, hashing with upsert semantics of lists of >> key-value pairs, transitioning through a state table, etc... >> >> Somebody else pointed out to me how odd it is of Python to be ditching >> reduce when Guido van Rossum was hired by Google, and Google is literally >> built on map and reduce (http://en.wikipedia.org/wiki/Mapreduce). >> > > That seems a bit literal. Just because they use a tool called MapReduce > that doesn't imply that thay chose to implement it with the Python map() > and reduce() functions. It's a distributed application, in case you > hadn't noticed ... > > regards > Steve -- http://mail.python.org/mailman/listinfo/python-list
