johndtill / neb Goto Github PK

The current implementation of the parse tree assumes that function names are string literals (or identifiers). For example, f(x) resolves to call a function f, which could be stored as a string. However, there are situations where the function key is not a string literal, e.g. the test phrase ⁜f⏴∂ϕ(r,t)⏵⏴∂t⏵=⁜∑⏴i=1⏵⏴3⏵⁜∑⏴j=1⏵⏴3⏵⁜f⏴∂⏵⏴∂⁜_⏴x⏵⏴i⏵⏵[⁜_⏴D⏵⏴ij⏵(ϕ, r)⁜f⏴∂ϕ(r,t)⏵⏴∂⁜_⏴x⏵⏴j⏵⏵] uses ⁜_⏴D⏵⏴ij⏵ as a function key. The current parser treats the arguments as implicit multiplication by a range instead of a function call.

The project needs to decide what fields a Node should have. The current design using a QString is easiest, but copies a lot of data, and might not have enough information to pinpoint error locations while processing the parse tree.

The error model currently uses exceptions, which is easy, but using exceptions for control flow is likely not ideal. Need to think through the design.