Currently, the grammar defines three rules related to assignment:
variable_declaration: used in const expressions and let bindings
named_field: used in named tuples and named arguments
assignment_expression: used everywhere else
variable_declaration doesn't actually do anything right now. const and let rules could be defined in terms of assignment_expression. The only place variable_declaration might be useful is in global/local declarations, but those are not defined in the grammar yet.
assignment_expression is not defined as an _expression to avoid conflicts. This is only a problem with simple assignments, but compound assignments don't have that problem. In fact, it's necessary to allow compound assignments in more places to allow things like xs[i+=1].
A viable solution might be:
Define a rule assignment_statement (that only uses =)
Redefine const, let, etc. to use assignment_statement
Redefine assignment_expression to only use compound assignments
Currently, the grammar defines three rules related to assignment:
variable_declaration: used in const expressions and let bindingsnamed_field: used in named tuples and named argumentsassignment_expression: used everywhere elsevariable_declarationdoesn't actually do anything right now.constandletrules could be defined in terms ofassignment_expression. The only placevariable_declarationmight be useful is in global/local declarations, but those are not defined in the grammar yet.assignment_expressionis not defined as an_expressionto avoid conflicts. This is only a problem with simple assignments, but compound assignments don't have that problem. In fact, it's necessary to allow compound assignments in more places to allow things likexs[i+=1].A viable solution might be:
assignment_statement(that only uses=)const,let, etc. to useassignment_statementassignment_expressionto only use compound assignmentsFixing quote expressions is necessary to add local/global scope qualifiers to avoid test regressions. see https://github.com/tree-sitter/tree-sitter-julia/pull/6#issuecomment-663881117 (that's why this is an issue and not a PR).