Nested Bytecode Blocks

[DavePearce]

An important issue is how to handle nested byte code blocks. After some thought, I've concluded this is primarily a UI issue. We want an API which doesn't have LoopEnd labels, and allows easy access to the contents of a nested block. There are two obvious approaches:

• Every bytecode still has a single index within the outermost block. This means you can iterate all byte codes at the top level. This is rather confusing, however.
• Every bytecode has a n-dimensional index reflecting its nesting level. This is perhaps more intuitive, but makes it hard to describe branches out of a block into a parent block without using labels.

In thinking about this, I'm also wondering about labels for branches. Use of labels gives rise to the relabelling problem. That is, importing one block into another (e.g. inline pre-/post-conditions) means we have to relabel the incoming blocks to avoid name clashes. Perhaps this should be considered a minor issue?

[DavePearce]

We can also simplify the issue of handling source annotations by using attribute maps?

[DavePearce]

Basically, done.