V: Assigned map value is lost

[mvlootman]

V version: 0.1.17
OS: MacOS Mojave

What did you do?
Assign a value to map and after reading it back it was not set to the expected value.

See code below, not sure if this is a bug or I should code it different.
How should the struct field be initialised? Should I create a method to return a new instance of the struct initialised to a empty map?

sample code:

struct MyStruct{
mut:
        visited  map[string]bool // map[string]bool{} not allowed => struct_map.v:4 expected `.name` but got `{`
}

const (
        key = 'key'
)

fn main(){

        fails()
        succeeds()
}

fn fails(){
        println('\nfails()')

        mut ms := MyStruct{}
        println(ms.visited[key].str()) // false (=OK)
        ms.visited[key] = true // assign true value to map
        println(ms.visited[key].str())  // assigned true so should print true (but prints false!)  <------
}

 fn succeeds(){
        println('\nsucceeds()')

        mut ms := MyStruct{}
        ms.visited =  map[string]bool{} // caller creates empty map

        println(ms.visited[key].str()) // false (=OK)
        ms.visited[key] = true // assign true value to map
        println(ms.visited[key].str())  // true (=OK)_
 }

What did you expect to see?
The succeeds() output for both, or a compiler error stating the map was not initialised. Since now the set value is lost without warnings.

What did you see instead?
fails()
false
false

succeeds()
false
true

[joe-conigliaro]

@mvlootman this is expected behavior as in the first example you did not initialize the map, you can do it when you create a new instance of MyStruct like this:

mut ms := MyStruct{
    visited: map[string]bool{}
}

[mvlootman]

@joe-conigliaro I understand the issue and the solution. But shouldn't this generate a compile/runtime error/panic when assigning to an uninitialised map? Instead of failing silently.

[joe-conigliaro]

@mvlootman I guess It will eventually be implemented to throw an error when trying to assign to an uninitialized map. I might add it even.

[mvlootman]

@joe-conigliaro Is there a reason why arrays are initialised and maps are not?

[joe-conigliaro]

I'm not sure of the reasoning

[medvednikov]

It's simply not implemented yet.

[joe-conigliaro]

In that case ill try implement, thought you had it like that for a reason

[ArcDrake]

Will this be implemented in a map?

    // Iterate just over the keys
    for key in phone_book {
        println($key)
    }

    // Iterate just over the values
    for value in phone_book {
        println($value)
    }

It throws: cannot range over type map_int

but

    // Iterate over all key:value pairs
    for key, value in phone_book {
        println('$key: $value')
    }

works fine.

Or will..

keys := phone_book.keys()

for key in keys {
   print('$key: ')
}

..be the way to go?
(I'm working on Rosetta Code examples for maps)
edit: Not the official Rosetta homepage, I'm not confident enough for that. Only a small github repo I like. Btw on the official homepage already exists a language called V and V is called Vlang :/

[ntrel]

As maps need to be initialised before use, this is the same case as pointers. V doesn't allow null initialisation for pointers, so likewise V should require that all maps are initialised before use. If the user doesn't want to initialise the map when it comes into scope, they should use an optional map.

[emily33901]

@ArcDrake fairly certain you can do for k, v in map

@ntrel why not just init to empty map

[joe-conigliaro]

This PR adds auto initialization for struct map fields: https://github.com/vlang/v/pull/1606

[mvlootman]

fixed by #1606

[ntrel]

why not just init to empty map

If it doesn't allocate that's ok, assuming it gets done in all situations. It looks like it just sets element_size. So after map is made generic, initialization might not be needed.