David A. Ham
I sort of think it's a shame to lose concise infix operators in favour of only using verbose function interfaces. If the users think that the flattened version is easier...
Obviously any PR will need tests.
superceded by #195
The idea here is that `remove_complex_node` is only called when compiling for real mode, so encountering a complex value is an error.
I think they're very niche and could be done as an external operator if someone really needed them.
Presumably all of these are syntactic sugar for combinations of indexing on the one hand, and `as_vector` and `as_tensor` on the other. That being the case, this seems fairly unobjectionable.
How do we know the LRU cache is big enough?
Isn't that a memory leak?
I think there is a tough lifetimes issue going on here. UFL needs to hold onto references for the duration of a tree traversal, lest the DAG explode. On the...
That's not recursion safe. Imagine someone puts some UFL inside an ExternalOperator, and then takes a derivative. The derivative application method of the external operator might well itself call the...