Refactor generic_data_handle.
#2467
Conversation
The commit changes the `generic_data_handle` as follows:
* Common sanitizing logic is consolidated in two methods `assert_*`.
* The logic explicitly deals with the 'literal' and 'data_handle'
cases separately.
* Hides the `type_name` method, by making it private.
* Adds assignment from a `data_handle`.
* Improves documentation.
This comment has been minimized.
This comment has been minimized.
|
✔️ 5dfa16c -> Azure artifacts URL |
Codecov Report
@@ Coverage Diff @@
## master #2467 +/- ##
=======================================
Coverage 60.41% 60.41%
=======================================
Files 626 626
Lines 120812 120823 +11
=======================================
+ Hits 72990 73001 +11
Misses 47822 47822
... and 1 file with indirect coverage changes 📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more |
|
✔️ 6025a9b -> Azure artifacts URL |
| * * small (`sizeof(T) <= sizeof(void*)`), trivial types such as `int` or `float`, | ||
| * * and raw pointers. | ||
| * | ||
| * In this context raw pointer is a regular C pointer, i.e. just and address. |
There was a problem hiding this comment.
| * In this context raw pointer is a regular C pointer, i.e. just and address. | |
| * In this context raw pointer is a regular C pointer, i.e. just an address. |
| template <typename T> | ||
| static constexpr bool can_be_stored_literally_v = | ||
| std::is_trivial_v<T> && !std::is_pointer_v<T> && sizeof(T) <= sizeof(void*); | ||
| (std::is_trivial_v<T> && sizeof(T) <= sizeof(void*)) || std::is_pointer_v<T>; |
There was a problem hiding this comment.
Pointer types are trivial, so this is equivalent
| (std::is_trivial_v<T> && sizeof(T) <= sizeof(void*)) || std::is_pointer_v<T>; | |
| std::is_trivial_v<T> && sizeof(T) <= sizeof(void*); |
but I am not totally convinced by the change to make can_be_stored_literally_v<double*> be true.
Storing small, trivial non-pointer values is an unfortunate feature supported by generic_data_handle that goes beyond it being a type-erased data_handle<T>. I hope that feature can be removed fairly promptly.
In contrast, storing a raw (in the sense that it is a not a permutation-stable handle to an entry in an soa container) T* is something that data_handle<T> does support, and which will not go away any time soon.
Given this, it seems preferable to keep as much of a distinction as possible between these two cases.
The commit changes the
generic_data_handleas follows:assert_*.type_namemethod, by making it private.data_handle.can_be_stored_literallyallows pointers.generic_data_handlein a type unsafe manner.