Braxton McKee
Braxton McKee
Right now, we have an incomplete implementation that doesn't actually let you do much with const dicts. We should implement the full set of semantics, including adding, converting from typed...
Our methodology for serializing core python objects works for lots of things, but not certain kinds of types and methods. For instance ``` import numpy import typed_python.Codebase as Codebase sc...
Otherwise, we end up with a mutable inside of a value, which makes no sense, prevents hashing, etc etc.
right now we resize the hashtable in the interpreter, which is much slower. This is just a performance optimization we have not done. We should provide compiled forms of both...
Not surprisingly, our str.split implementation is much slower than python's internals for lots of small strings. A big part of it is the malloc implementation (tcmalloc speeds it up by...
Right now, we attempt to compile everything. If we see something we don't understand in the compiler, we attempt to convert everything to 'object' and call the interpreter. This is...
We have `hash` implemented at the compiler level for integers, strings, bytes, floats, and TupleOf. But we need it for everything else, along with comprehensive tests for correspondence between the...
As part of #110, we're allowing the creation of a common base Type. When a class has a member that's just Type, we should be able to generate code that...
right now if you try to send a very large string (say, 100gb), the deserializer will allocate space for it. We should allow clients to impose strict limits to ensure...