[BeetleB]

So you would be equally critical of overloading [] for maps?

Sorry, hard for me to relate, as I've overloaded [] (in, say, Python) to make life easy on everyone. People loved it.

I hope you're aware that there is a long standing debate on whether overloading operators is good/bad, and it comes down to personal preference?

[fooker]

> So you would be equally critical of overloading [] for maps?

No, I'm not sure how you got that impression. Overloading is great.

It's also confusing when it does something completely different from what you intuitively expect.

[BeetleB]

If you're an old timer, you expect [] to index into an array, and you definitely do not expect it to do a lookup in a map/dict. Older languages just didn't include a dict-like structure in the language (technically, even C++ didn't).

Do you not see that for them, overloading [] for dictionary lookups is "something completely different from what you intuitively expect"?

[tialaramex]

More strikingly, C++ doesn't distinguish what Rust would call IndexMut and just Index, the use of the [idx] operator in a context where we'll mutate things and one where we don't want that.

Rust's HashMap implements Index because answer = map[name] is a perfectly reasonable thing to do, and if there is no key matching name then we panic, makes sense but it does not provide IndexMut so that you could write map[name] = answer because the edge cases are non-obvious so better to make you write what you meant.

C++ hash tables implement operator[] but the result is mutable, in order that map[name] = 123.0 can work which in Rust would be IndexMut and isn't provided. Because this is true, the index operation always succeeds, if you ask for map["not-present"] it creates the hash table entry for "not-present" and tries to store a default value ready to update it if you later assign to the reference.