I think the real issue here is not whether there is a separate type for tables and arrays, but whether the arrays are homogenous (all elements must have the same type). In most dynamic languages, the arrays are heterogeneous. For example, Python has a separate array type, but if you want homogenous arrays you have to reach for something like numpy.

The main catch is that if the optimization guesses wrong and a different type is inserted into the table afterwards, then it would incurr an O(n) operation to transfer all the data to a deoptimized table.

Another caveat is that Lua can have more than one internal representation for the same type, and those have different type tag variants. For instance: strings can be represented internally as either short or long strings; Functions can be Lua closures, C closures, or perhaps even an object with a __call metamethod; Objects can be either tables or userdata.

This seems likely to create some inexplicable performance elbows where you have 1000 strings, but there's one code path that replaces one with a number, and now the whole array needs to be copied. Tracking that down won't be fun.

It makes a lot of sense, and but then you have two code paths for tables.

The Lua folks want a simple codebase, so they (knowingly) leave a lot of performance on the table in favor of simplicity.

Here's the rest of that paragraph for you:

"However, this attribute is a gcc extension not present in ISO C. Moreover, even in gcc it is not guaranteed to work [3]. As portability is a hallmark of Lua, this almost magical solution is a no-go."

`__attribute__((packed))` wouldn't help here since the issue is about Lua's array/hash hybrid table design and memory allocation strategy, not C struct padding.