Different "?" syntax required when indexing maps with sumtypes and without

[jdonnerstag]

V version: V 0.2.4 359b674.66a67de
OS: windows, Microsoft Windows 10 Enterprise v19042 64-bit

What did you do?
I updates a module of mine vlang-yaml to test it against the latest V version

What did you expect to see?
Run my test cases like v -stats test ., and using -stats to also see all the compiler warnings.

What did you see instead?
The error message is clear: Append ? like so x1.obj["national"]?. See below

What is wrong: Only when indexing a map with a sum type, you need to handle the option (error). If not a sum type, it is not necessary. That is inconsistent. Either the errors must be handle when indexing a map or not. My preference:

x := mymap["a"]
x := mymap["a"] or { mydefault }
x := mymap["a"] or { return error("..") }

step3_reader_test.v:58:15: warning: `or {}` block required when indexing a map with sum type value
   56 |             assert x2.ar[2] == YamlValue("New York Yankees")
   57 |         }
   58 |         x3 := x1.obj["national"]
      |                     ~~~~~~~~~~~~
   59 |         assert x3 is YamlListValue
   60 |         if x3 is YamlListValue {

[dumblob]

Yep, this is one of the few exceptions. Allegedly this exception is there to make use of maps easier at the cost of pushing compile-time safety to run-time panic (i.e. also a kind of "safety").

But I don't like this strawman reasoning as we can easily propagate the error with ? (or in the future with ! which though has a different meaning as of now) as in x := mymap["a"]?["b"]?["c"]? which if required by compiler is just one character more than x := mymap["a"]["b"]["c"] and makes it explicit and safe before even executing the resulting app.

We shouldn't also forget about the case if it's an l-value. But that's an old discussion and V lang differs now in a way we could easily use ? or ! as noted above.

Feel free to push harder for this than I'm able to 😉!

Another slightly related topic: #10332 .