Seems reasonable to me, but I thought much about isPartOfExpression. @weswigham do you have opinions about PropertyAssignment being true for isPartOfExpression?

After skimming the code and uses of isPartOfExpression it still seems like a good idea.

isPartOfExpression feels like a best-effort cheapish thing. If you 100% must know if a node is within an expression context, you can't actually always trust it, I think. Its more like isPossiblyPartOfExpression. Its implementation just feels... Underspecified a little?

Having said that, it's mostly used by the language service and test harness, so changing it is unlikely to have too many significant ramifications.

It looks like that won't work because it tries to make us baseline the type of a PropertyAccess. isPartOfExpression is misnamed -- it's not returning whether something is part of an expression, it's only true if the node itself is an expression. It should be called isExpression, but the function currently named isExpression is specific to transforms and can't look at the parent.