Dictionary integrity protection #3032
Conversation
This one is debatable. It follows modern C++.
This one prohibits annoying casts. It is also important for future context injection and follows STL practice.
Returning const object does not prevent from using mutable by triggering copy constructor.
This one is more stylistic but is consistent with practice of using pre- operations for iterators.
This is a small compilation time optimization by reducing total number of generated types. Should not change runtime behavior.
mtszkarbowiak
force-pushed
the
fix/dictionary-integrity
branch
2 times, most recently
from
af99ed5 to
89a0410
Compare
mtszkarbowiak
force-pushed
the
fix/dictionary-integrity
branch
from
89a0410 to
1d56220
Compare
|
Can someone help me with changes in the bindings generator? |
You can take a look into generated code here: |
|
This refactor introduced very large change and potential perf overhead (such as assert on bucket state and function call). Thus I have 2 ideas that maybe could solve this problem but with different approache: I think more important would be addition of What do you think? |
|
I like the idea a lot. I will try it. |
2 participants
Currently
Dictionaryuses buckets to expose data. This means that the key can be changed, potentially ruining the state of the collection.This is an API breaking change.
HashSetis to be revised in the future.