morestl: add examples for std::optional, any, variant

[theanalyst]

Draft WIP #69 PR ...don't have a full tex system currently so will be (ab)using the CI for a bit while I still make up other examples...

[theanalyst]

Added slides on std::variant and std::any now; also a brief one on std::string_view. I think this is ready for review now, happy to change anything.

[sponce]

Now that the different slides are written, I feel like the original one should be dropped and the content distributed into the different sections. It feels odd to start with it, then switch to string_view and comes back to the 3 items, one by one.

[theanalyst]

Now that the different slides are written, I feel like the original one should be dropped and the content distributed into the different sections. It feels odd to start with it, then switch to string_view and comes back to the 3 items, one by one.

This makes sense, also makes it easy slide estate planning wise so I could mostly go for one slide explaining the item and the next one with the program example

[sponce]

@theanalyst This is almost ready to be merged, but if you do not mind, I would go for dropping the introduction slide and merge it in the others. I can do it myself if you agree

[hageboeck]

I have one formatting thing and one optional line break. Otherwise, the output of the program would look a bit complicated.

[theanalyst]

@theanalyst This is almost ready to be merged, but if you do not mind, I would go for dropping the introduction slide and merge it in the others. I can do it myself if you agree

Yes sure go ahead, sorry things got a bit too busy after the vacation

[sponce]

Ok, will do. I was also busy until now, so starting only today. One question I have though. The code presented for std::any looks wrong to me :

 if (Monster *m = std::any_cast<Monster>(&e)) {
          m->take_hit(this->damage);
        } else if (Orc *o = std::any_cast<Orc>(&e)) {
          o->take_hit(this->damage*0.5);
        } else if (Cat *c = std::any_cast<Cat>(&e)) {//No!

it assumes std::any_cast will return nullptr, but it will throw std::bad_any_cast, not return

[hageboeck]

it assumes std::any_cast will return nullptr, but it will throw std::bad_any_cast, not return

Actually, only the reference and universal reference versions will throw @sponce. The pointer versions do return nullptr. cppreference.
But what I find very confusing on first look is that the template argument is the type <Orc> whereas the argument of the cast is a pointer (&e).
This is different from what you normally have, e.g. static_cast<Type*>(&e). This is what we should discuss in detail. If we both get confused on first look (in slightly different directions, admittedly), learners might be confused as well.

[sponce]

On now I've seen the little "(1-3)" in the cpp reference ! Thanks @hageboeck for being more awake than I was.
I find this extremely confusing...

I anyway would not take an example using pointers. After all we want to teach people to drop them everywhere.

[sponce]

So I've created a new version of this one in #89 (could not push to @theanalyst' s branch). I've changed the examples for any and if constexpr. @hageboeck if you could review, would be appreciated

[theanalyst]

it assumes std::any_cast will return nullptr, but it will throw std::bad_any_cast, not return

Actually, only the reference and universal reference versions will throw @sponce. The pointer versions do return nullptr. cppreference. But what I find very confusing on first look is that the template argument is the type <Orc> whereas the argument of the cast is a pointer (&e). This is different from what you normally have, e.g. static_cast<Type*>(&e). This is what we should discuss in detail. If we both get confused on first look (in slightly different directions, admittedly), learners might be confused as well.

Yes indeed, that was a mistake, should've been Type*<&e> good catch. Also agree teaching value/references are easier, lesser chances to footgun ;)