On 11/26/24 5:12 PM, Andy Valencia wrote:
> Again and again for testing I run into how nice it would be to have an open "File" which has its contents set by the unit test code:
> 
>> auto f = StringFile("my test data...", "r");
> 
> I've searched high and low without discovering something which would fit the bill.
> 
> Thanks in advance,
> Andy
> 

An earlier post by H. S. Teoh may be helpful:

 https://forum.dlang.org/post/mailman.10192.1685482943.31357.digitalmars-d@puremagic.com

Ali

On Wednesday, 27 November 2024 at 02:43:21 UTC, Ali Çehreli wrote:
> On 11/26/24 5:12 PM, Andy Valencia wrote:
>> Again and again for testing I run into how nice it would be to have an open "File" which has its contents set by the unit test code:
>> 
>>> auto f = StringFile("my test data...", "r");
> An earlier post by H. S. Teoh may be helpful:
> https://forum.dlang.org/post/mailman.10192.1685482943.31357.digitalmars-d@puremagic.com

Thank you.  It's somewhat bemusing to, in essence, wrap the entirety of an implementation around a #define just to slot in an alternative to File.  With my OO programming goggles on, I can't help but notice that if Phobos had made File a class--or an interface--then it would be possible through simple type compatibility to have a FileFile instance which would satisfy the File API using the filesystem, and as no special case have a StringFile which did the same from the body of a string.

But I used your suggested approach, and it does indeed work nicely.  Thank you again.

Andy

On Wednesday, 27 November 2024 at 14:37:03 UTC, Andy Valencia wrote:
> With my OO programming goggles on, I can't help but notice that if Phobos had made File a class--or an interface--

oo doesn't own polymorphism you could do 99% of the value with

```d
struct dummyfile{
  auto byLineCopy()=>//string.spiltter('\n')....
}

auto myfunc(F)(F file){
  foreach(s;file.byLineCopy){
    ...
  }
}
unittest{ assert(
  dummyfile(q{
    ...
  }).myfunc
  == ...);
}
```

On Wednesday, 27 November 2024 at 20:38:22 UTC, monkyyy wrote:
> On Wednesday, 27 November 2024 at 14:37:03 UTC, Andy Valencia wrote:
>> With my OO programming goggles on, I can't help but notice that if Phobos had made File a class--or an interface--
>
> oo doesn't own polymorphism you could do 99% of the value with
>
> ```d
> struct dummyfile{
>   auto byLineCopy()=>//string.spiltter('\n')....
> }
>
> auto myfunc(F)(F file){
>   foreach(s;file.byLineCopy){
>     ...
>   }
> }

Yes, the page pointed to by Ali Çehreli up above shows this, thus my comment about how OO techniques avoid the "big hammer" of macro expansion (well, template expansion...).

Andy