Associativity of matmul unclear from grammar for m_expr

Documentation

6.7. Binary arithmetic operations shows the grammar rule

m_expr ::=  u_expr | m_expr "*" u_expr | m_expr "@" m_expr |
            m_expr "//" u_expr | m_expr "/" u_expr |
            m_expr "%" u_expr

in which all operators except @ for matmul are left-associative (meaning, a*b*c parses as m_expr a*b and u_expr c ) while @ is ambiguous: Does a@b@c parse as m_expr a@b and m_expr c or as m_expr a and m_expr b@c?

According to PEP 465, @ should have the same associativity as *. Should this grammar rule should reflect that by saying m_expr "@" u_expr, that is, with the right-hand side changed from m_expr to u_expr?

Linked PRs

• gh-138847
• gh-145636
• gh-149281

Comienza con la sección «6.7 Binary arithmetic operations» enlazada en el issue y compara su gramática m_expr con la asociatividad especificada en PEP 465. Revisa los PR enlazados gh-138847, gh-145636 y gh-149281; se considera terminado cuando la documentación hace inequívoca la asociatividad de @ y coincide con la especificación.