Hi There,

I am trying to switch between two structs as I am using the csvReader on a raw string. The pseudo-code below throws a "cannot implicitly convert" error due to difference between struct_type1 and struct_type2. I must be doing something wrong or have a wrong understanding of how this function works. Could someone give a good suggestion on handling this?

Thanks!

Best,
Selim

auto records = rawtext.csvReader!struct_type1(';');
if(aControlCondition) {
records = rawtext.csvReader!struct_type2(';');
}

// Iterate through each data row.		
foreach (record; records) {
writeln(record);
}

On Thursday, 5 November 2020 at 21:18:52 UTC, Selim Ozel wrote:
> auto records = rawtext.csvReader!struct_type1(';');

D is statically typed and `auto` means "deduce this type for me based on this one function's return value". It is not like JavaScript's `var` whose type may change.

If I'm not mistaken the `csvReader` function returns a range struct, and the full type is something long and unwieldy like `CsvReader!(struct_type1, cast(Malformed)1, string, dchar, string[])`. So just think of `records` as being that.

(You can tell what type it is at compilation with `pragma(msg, typeof(records).stringof)`.)

> if(aControlCondition) {
> records = rawtext.csvReader!struct_type2(';');

Here `csvReader!struct_type2(';')` returns a value of type `CsvReader!(struct_type2, ...)`, which is a different type from that of `records` (and they're not implicitly convertible). So the error message is right. If `auto` worked like `var` your code would work, but it doesn't.

You need two different variables and two different `foreach`es. For the same code to work on both types, the easy solution is templates. Perhaps make the `foreach` part after the reads a templated function that accepts any type passed to it?

On Thursday, 5 November 2020 at 22:36:36 UTC, Anonymouse wrote:
> If I'm not mistaken the `csvReader` function returns a range struct, and the full type is something long and unwieldy like `CsvReader!(struct_type1, cast(Malformed)1, string, dchar, string[])`. So just think of `records` as being that.

I actually first going this route but couldn't figure out the correct name for that data type. It is quite long.

> You need two different variables and two different `foreach`es. For the same code to work on both types, the easy solution is templates. Perhaps make the `foreach` part after the reads a templated function that accepts any type passed to it?

Embedding the foreach loop inside a template function and deciding on the data type at the higher level function solved my issue. Thanks for the pointer!

Best,
Selim

On Fri, Nov 06, 2020 at 07:17:53PM +0000, Selim Ozel via Digitalmars-d-learn wrote:
> On Thursday, 5 November 2020 at 22:36:36 UTC, Anonymouse wrote:
> > If I'm not mistaken the `csvReader` function returns a range struct, and the full type is something long and unwieldy like `CsvReader!(struct_type1, cast(Malformed)1, string, dchar, string[])`.  So just think of `records` as being that.
> 
> I actually first going this route but couldn't figure out the correct name for that data type. It is quite long.
[...]

You can use the typeof() operator to capture the type of a long, unwieldy type in an alias. This is useful if you ever need to store such a return type somewhere, e.g.:

	alias T = typeof(csvReader(...));

	struct MyStorage {
		T result;
	}

	MyStorage s;
	s.result = csvReader(...);

Let the compiler figure out the type for you. :-)


T

-- 
Guns don't kill people. Bullets do.

On Friday, 6 November 2020 at 19:35:47 UTC, H. S. Teoh wrote:
> You can use the typeof() operator to capture the type of a long, unwieldy type in an alias. This is useful if you ever need to store such a return type somewhere, e.g.:
>
> 	alias T = typeof(csvReader(...));
>
> 	struct MyStorage {
> 		T result;
> 	}
>
> 	MyStorage s;
> 	s.result = csvReader(...);
>
> Let the compiler figure out the type for you. :-)
>
>
> T

That's great, thanks!

S