[Mlir-commits] [mlir] [mlir][vector] Update the internal representation of in_bounds (PR #100336)

[Benjamin Maxwell]

MacDue wrote:

> I have some concerns about keeping a dictionary of attribute syntax while making it appear as something different: that seems like counter-intuitive to the users who think that they see an `attr-dict` but the in-memory representation does not match what we see.

I would be surprised if people think that much when they see something like `foo = [1, 2, 3, 4]` in the `attr-dict`. I'd expect most people would think "array of integers". They would not specifically think of `ArrayAttr` of `IntegerAttr` without knowing a lot about how MLIR's parsing works (especially because the same syntax may not mean ArrayAttr` of `IntegerAttr` outside the `attr-dict`).

Also, as this is an inherent attribute over in C++ you can see the type of the attribute is a dense array. 

https://github.com/llvm/llvm-project/pull/100336