The EBNF specifies the syntax, not the behavior. EBNF does not indicate the order *of *evaluation of source code, only the order of characters *in *the source code.
On Wednesday, July 19, 2017 at 6:59:29 PM UTC-4, Ian Lance Taylor wrote: > > On Wed, Jul 19, 2017 at 3:28 PM, <adrian...@gmail.com <javascript:>> > wrote: > > > > A question on Stack Overflow led me to carefully examine the spec and I > feel like there may be some detail that's missing - the behavior is easy > enough to work with, but it's effectively undefined according to the > language spec. Specifically, with a variable x of type *[]string for > example, *x[0] will not work because it is evaluated as *(x[0]), not as > (*x)[0]. This is unexpected based on the spec because the only > specifications that could apply are the general order of evaluation, which > is left to right (not the case here), and operator precedence which states > that pointer dereference is a unary operator and unary operators have > highest precedence (again clearly not what's happening). > > > > The closest it comes to explaining this behavior is in the section on > address operators, which implies that the address operator & applies to the > entire slice expression next to it (or struct field selector, etc). This > leaves one to assume the same implication applies to the pointer > dereference operator as well. > > > > Is there something covering this that I glossed over reading the spec? > If it's not just something I missed, is this worth clarifying in the spec, > without changing the behavior (purely a documentation change)? > > This is expressed in the EBNF grammar in the language spec. x[0] is a > PrimaryExpr. *x[0] is a unary_op applied to a PrimaryExpr. > > Ian > -- You received this message because you are subscribed to the Google Groups "golang-nuts" group. To unsubscribe from this group and stop receiving emails from it, send an email to golang-nuts+unsubscr...@googlegroups.com. For more options, visit https://groups.google.com/d/optout.
