Order of precedence

[chqrlie]

Changing the order of precedence of the C operators is problematic:

• it causes more confusion for little benefit.
• it is inconsistent with the C syntax (breaks the philosophy Should be easy to learn for C programmers)
• it is inconsistent with other languages that inherit the C operators: c++, java, javascript, csharp, kotlin... that have huge momentum, making C2 quite unfriendly to learn and port code to.
• it is unnecessary to fix the problem.

I propose you keep the same order of precedence, but require explicit parentheses for subexpressions when mixed with other operators whose precedence is deemed counter-intuitive or at least confusing. This is what clang and gcc do with strict warning levels and does not break compatibility: code that compiles in C2 would parse to the same AST in the other languages.

a + b >> c + d would require parentheses to lift the confusion between a + (b >> c) + d and (a + b) >> (c + d).

Same for a & b == c which should always be written (a & b) == c

a || b && c would become a || (b && c) or (a || b) && c depending on the programmer's intent, which is non obvious.

But a & b & c, a && b && c... would be accepted as unambiguous.

I think the is a better approach for the philosophy Should help to avoid common mistakes.

[bvdberg]

Again, this was discussed a Lot. I think the C system is a big cause for bugs. Many programmers are unsure and just add some parenthesis to make sure. This is fine I think and also makes the code easier to read without having to memorize the whole precedence table.

Maybe we could add a warning in the compiler when this happens?

[chqrlie]

Again, this was discussed a Lot. I think the C system is a big cause for bugs. Many programmers are unsure and just add some parenthesis to make sure. This is fine I think and also makes the code easier to read without having to memorize the whole precedence table.

Yes, and mandating these parentheses seems to me a better fix than changing the precedence rules.