Lazy any #4343
Lazy any #4343
Conversation
|
@karlnapf I'm in progress yet, just to let you know something is going on ;) |
| { | ||
| auto v = 9; | ||
| auto any = make_any<int>([=]() { return v; }); | ||
| EXPECT_EQ(any.as<int32_t>(), v); |
There was a problem hiding this comment.
@karlnapf this thing works now, but it definitely needs more thorough testing
There was a problem hiding this comment.
shall we try to test a slightly more elaborate example of registering a member function that returns a constant or something?
|
@karlnapf an interesting question is whether we want to support the following assignments:
|
|
idk, imo we only need the functions to returns stuff. But you are the guru here :D |
| any = make_any<int>([=]() { return v; }), TypeMismatchException); | ||
| } | ||
|
|
||
| TEST(Any, lazy_member_function) |
There was a problem hiding this comment.
What do you think about a test case that asserts laziness?
There was a problem hiding this comment.
Also, we need to keep in mind equals/serialization should exclude lazy computations.
There was a problem hiding this comment.
i.e. we need some kind of markup. Otherwise these things will be equaled/cloned/serialized as well....or is that intended? Costs computation and I am not sure it is worth it
There was a problem hiding this comment.
ah, clone_from will be a problem as it attempts to write into the member
There was a problem hiding this comment.
That's a bunch of questions :)
- Serialization is tricky. It seems the 'serializator' itself should know
- We can explicitly throw something when trying to do equals
- Clone is cheap but it won't work correctly. It would use the original object
There was a problem hiding this comment.
I think then in clone and equals, I would just skip all registered function anys
| } | ||
|
|
||
| template <class T> | ||
| struct has_result_type |
|
|
||
| int computed_member() const | ||
| { | ||
| return 90210; |
|
All fixed? :D |
|
Not yet, there is a bug in has_type_fallback. And some lazyness flag visible outside is due. |
|
add the lazyness flag you lazy flagger |
|
I am lazy to add the lazyness |
| any = make_any<int>([=]() { return 222; }); | ||
| EXPECT_FALSE(any.cloneable()); | ||
| EXPECT_FALSE(any.visitable()); | ||
| EXPECT_THROW(any.visit(nullptr), std::logic_error); |
| { | ||
| SG_SDEBUG( | ||
| "Skipping clone of %s::%s of type %s.\n", this->get_name(), | ||
| tag.name().c_str(), own.type().c_str()); |
There was a problem hiding this comment.
actually, why not move the check after the debug msg below, then you dont need to print the name and stuff again, but can just say "Skipping as not cloneable"
There was a problem hiding this comment.
Yeah but since it is here such thing might be useful anyway
| "Skipping comparison of %s::%s of type %s as it is " | ||
| "non-visitable.\n", | ||
| this->get_name(), tag.name().c_str(), own.type().c_str()); | ||
| continue; |
There was a problem hiding this comment.
same here, i would move this down and only say "skipping" to avoid duplicate code
* Add simple draft of lazy any * Do not deref null pointer in any * Specialize for std::function * Add assignment tests * Use is_functional to actually call functions * Fix style * Add get_value specialization for Empty * Make has_type_fallback support std::function * Add cloneable/visitable to filter out some Anys * Skip non-visitable and non-cloneable anys * Add missed continue
No description provided.