Manual has incorrect statement of what lvalues are

[ezyang]

Currently, the manual states that only "Path, field and index expressions are lvalues." However, this seems inaccurate; for example, *x should be considered an lvalue.

expr_kind in src/librustc/middle/ty.rs seems to have the definitive say on what is an lvalue:

    match expr.node {
        ast::ExprPath(..) | ast::ExprSelf => {
            match resolve_expr(tcx, expr) {
                ast::DefVariant(..) | ast::DefStruct(..) => RvalueDpsExpr,

                // Fn pointers are just scalar values.
                ast::DefFn(..) | ast::DefStaticMethod(..) => RvalueDatumExpr,

                // Note: there is actually a good case to be made that
                // def_args, particularly those of immediate type, ought to
                // considered rvalues.
                ast::DefStatic(..) |
                ast::DefBinding(..) |
                ast::DefUpvar(..) |
                ast::DefArg(..) |
                ast::DefLocal(..) |
                ast::DefSelf(..) => LvalueExpr,

                def => {
                    tcx.sess.span_bug(expr.span, format!(
                        "Uncategorized def for expr {:?}: {:?}",
                        expr.id, def));
                }
            }
        }

        ast::ExprUnary(_, ast::UnDeref, _) |
        ast::ExprField(..) |
        ast::ExprIndex(..) => {
            LvalueExpr
        }
[snip]

So, I suggest changing this line to something like, "Paths to memory locations, fields, index expressions and dereferences are lvalues." I can submit a pull request if people agree.

[nikomatsakis]

I agree. I would define lvalues like:

"An lvalue is an expression that represents a memory location. These expressions are paths to local variables, function and method arguments (including self), and static variables as well as dereferences (*expr), indexing expressions (expr[expr]), and field references (expr.f)."