protocolbuffers / protocolbuffers/protocolbuffers.github.io
repeated_ptr_field.h -- document lifetime of pointers and references
@jguamie is already working on this.
Since Oct 17, 2025.
- Dominant language
- HTML
- Stars
- 69
- Forks
- 179
- PR merge metrics
- No merged PRs in 30d
Description
In STL containers that store value elements (e.g. vector<Bar> as opposed to vector<Bar*>), element pointers are invalidated by operations such as adding or removing elements from the container.
The generated C++ code for repeated message fields returns a container of value elements:
https://protobuf.dev/reference/cpp/cpp-generated/#repeatedmessage
const RepeatedPtrField<Bar>& bar()
RepeatedPtrField<Bar>* mutable_bar()
However, based on the "Ptr" in the name, and looking at the current implementation, it seems that the intention is for RepeatedPtrField to behave more like vector<Foo*>, where it is safe to hold on to pointers to the underlying messages.
For example, is this safe?:
message Bar {
string baz = 1;
}
message Foo {
repeated Bar bar = 1;
}
...
std::string do_something(Foo *foo) {
// Get a reference to first element of bar
const Bar& bar = *foo->bar()->begin();
// Add a bar to foo
foo->add_bar();
// Do something with the first bar reference
return bar.baz();
}
I'm not finding clear documentation on this behavior though. Assuming it is meant to be safe to hold on to these references/pointers, could this please be documented?
Related stack overflow making the same assumption:
https://stackoverflow.com/questions/33219022/do-pointers-to-items-of-a-repeated-gpb-field-stay-valid-if-the-field-is-modified
Contributor guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Assessment
This issue has not been assessed yet.