This memory management should be handled by Rust.
I've confirmed this works by running the new included tests with msan.
The sanitizer is necessary to detect an incorrect copy_from impl
that uses-after-free from the upb arena.
PiperOrigin-RevId: 604689154
A public "raw" field in a safe wrapper is guaranteed unsound!
This makes repeated and map inner access consistent and
avoids exposing raw internals.
It also provides the accessors necessary for implementing map
access for external types.
PiperOrigin-RevId: 604405543
Part 3 of 4 (added a stage).
The getter and mut_getter bifurcate based on the kernel.
upb returns Option<RawMessage>, while cpp returns the RawMessage.
This means that we can't have a unified MessageVTable in vtable.rs, so we've split these out in {upb.rs and cpp.rs}.
$field$_entry is now prepped and populated for the $field$_mut swappage in the following CL.
PiperOrigin-RevId: 601230880
We want to return $pb$::FieldEntry<'_, $msg_type$> for msg_mut accessors as opposed to the current state (returning $Msg$Mut directly).
In this CL, we pave the way to implementing field entry returns.
We introduce { MessagePresentMutData, MessageAbsentMutData } and impl { ProxiedWithRawVTable, ProxiedWithRawOptionalVTable }. I initially tried a blanket impl approach, but it collided with the already existing PrimitiveVTable constructs; perhaps worth revisiting post 0.6.
In a followup, we'll flesh out the bodies. Lastly, we'll perform the swapover by
replacing $field$_mut with $field$_entry, updating all related tests.
PiperOrigin-RevId: 599282850
We now support fields with bytes as map values e.g. map<i32, bytes>. The implementation for the C++ runtime was straightforward. The majority of the changes in this CL are about the UPB runtime. In UPB, when we insert Rust bytes/string into the map we need to first copy the bytes onto the maps arena. To support this I have rewritten the macro that implements the ProxiedInMapValue types. I refactored the functionality to convert between UPB and Rust types into the 'UpbTypeConversions' trait. This trait has a function 'to_message_value_if_required' which does the copying for bytes and strings.
PiperOrigin-RevId: 599118416
This adds private methods of:
-- .raw_msg() to Msg+MsgMut+MsgView
-- .raw_arena() to Msg+MsgMut [upb kernel only]
And updates the accessors to use the self.raw_msg() / self.raw_arena().
A couple more things will need to be changed before the accessors can be verbatim reused in Msg/MsgView/MsgMut which will be mailed separately.
PiperOrigin-RevId: 598869392
- ProxiedInMapValue is defined in maps.rs, and no longer in the runtime files {upb, cpp}.rs.
- ProxiedInMapValue's methods accept and return Proxied types.
- InnerMapMut no longer has any generic type parameters.
- Through this refactoring the Map type is no longer a ZST. Creating a new map is now as simple as `Map::new()`.
PiperOrigin-RevId: 597765165
This change is a pure refactoring and simplification of the code. We replace all MapsWith<TYPE>KeyOps traits through a single generic ProxiedInMapValue<K> trait. Through connecting the runtime maps implementation with Proxied the code gets a lot simpler e.g. we can use View<T> instead of hardcoding the concrete type behind it.
I also expect this change to be beneficial for the gencode. In a subsequent CL we'll implement message values for maps. After this change we'll only have to implement a single trait, while before we had to implement num(key types) many traits.
PiperOrigin-RevId: 596562909
For the cpp runtime, call the `Message::CopyFrom` method.
For the upb runtime, expose the message `MiniTable` and call `upb_Message_DeepCopy`.
PiperOrigin-RevId: 595166276
- Rename most usage of `'a` to `'msg`
- Remove a no-op unused lifetime param for `remove` in maps
- Elide lifetimes recommended by clippy
PiperOrigin-RevId: 589878364
This change implements maps with keys and values of type string e.g. Map<ProtoStr, i32> and Map<ProtoStr, ProtoStr>.
Implementing the Map type for ProtoStr has been different from scalar types because ProtoStr is an unsized type i.e. its size is not known at compile time. The existing Map implementation assumed sized types in many places. To make unsized types fit into the existing code architecture I have added an associated type 'Value' to the MapWith*KeyOps traits. The associated type needs to be sized and is the type returned by the Map::get(self, key) method e.g. for aProtoStr, the `type Value = &ProtoStr`.
PiperOrigin-RevId: 588783751
This change implements the Proxied trait for the Map type.
It leaves a TODO to implement SettableValue. I haven't implemented SettableValue yet because I have not yet been able to verify that we get the right copy_from semantics for both kernels. I'll implement set_on in a follow up.
PiperOrigin-RevId: 584285061
Before this change the runtimes export a Map and MapInner. This change merges the Map and MapInner types into a single MapInner type, removing the Map type from the runtime (upb.rs, cpp.rs).
The motivation for this change is twofold:
1) A separate Map type is not strictly needed by the runtime. I hope this reduces some complexity.
2) After this change we can introduce a runtime-agnostic protobuf::Map type that implements Proxied.
PiperOrigin-RevId: 582978008
This CL implements Maps for scalar types for the C++ runtime. It's orthogonal to cl/580453646. This CL is constrained by having to force template instantiation of proto2::Map<K, V>. Put differently, a Rust protobuf::Map<K, V> implementation needs to call 'extern "C"' functions with both key and value type in the function name (e.g. __pb_rust_Map_i32_f64_get()). We use macros to generate a Map implementation for every (K,V)-pair. An alternative would have been to use vtables.
Luckily a key in a protobuf map can only be integer types, bool and string. So the number of key types is bounded by the specification, while the number of value types is not i.e. any protobuf message can be a value in a map. Given these constraints we introduce one 'MapKeyOps' trait per key type e.g. MapKeyBOOLOps or MapKeyI32Ops. These traits need to be implemented for every value type e.g. 'impl MapKeyBOOLOps for i32' will implement 'Map::<bool, i32>'. In particular the MapKeyOps traits can also be implemented for generated messages without violating the orphan rule.
This CL also contains significant changes to the UPB runtime so that both upb.rs and cpp.rs export a similar interface to simplify the implementation in map.rs and the generated code.
This CL does not yet implement the Proxied trait.
PiperOrigin-RevId: 582951914
We've had access to views for submessages for a while:
If you hit some_message.submsg().some_int(), you'll get a view for that int.
Until now, there hasn't been a way to get some_message.submsg_mut(), so we introduce the mutational pathway here.
We haven't added fully-functioning mutation, but this is a step towards that goal.
subview was inaccurate, so I've refactored and renamed: { accessor_fns_for_views, accessor_fns_for_muts }.
PiperOrigin-RevId: 581984371
We shouldn't give easy access to the zeroed block, so let's pass in pbi::private to deter misuse.
Also saw that arena wasn't used mutably in a repeatedfield test, so let mut -> let;
PiperOrigin-RevId: 574896460
We had previously commented out the upb portion of simple_nested_test.
This is because nonmutable getters have submessages being NULL by default.
This means that trying to fetch anything, like a simple scalar from that nested message would segfault.
This CL makes the externC return an Option<RawMessage> since we've discovered that upb can return NULL. This way, we can check for `None` and handle the NULL case appropriately.
We know that the NULL pathway can only come from terra upb, since
cpp automagically constructs submsgs if they don't exist.
We've augmented upb.rs to contain a scratch space that allocates a zeroed-out contiguous chunk of memory @64KB. Since a block of zeroed-out memory is a legit message from upb's point of view, we can provide $pbr$::ScratchSpace::zeroed_block() to upb in order to get the default submessage behavior we want from upb.
This block is lazily allocated upon first request. This means that a consumer of the cpp kernel will not incur an additional cost.
PiperOrigin-RevId: 573840755
This makes a few changes:
- It changes generated messages to reference message innards as a type in `__runtime` instead of branching on what fields should be there. That results in much less bifurcation in gencode and lets runtime-agnostic code reference raw message innards.
- It adds a generic mechanism for creating vtable-based mutators. These vtables point to thunks generated for interacting with C++ or upb fields. Right now, the design results in 2-word (msg+vtable) mutators for C++ and 3-word mutators (msg+arena+vtable) for UPB. See upb.rs for an explanation of the design options. I chose the `RawMessage+&Arena` design for mutator data as opposed to a `&MessageInner` design because it did not result in extra-indirection layout changes for message mutators. We could revisit this in the future with performance data, since this results in all field mutators being 3 words large instead of the register-friendly 2 words.
- And lastly, as a nearby change that touches on many of the same topics, it adds some extra SAFETY comments for Send/Sync in message gencode.
PiperOrigin-RevId: 559483437
These are more type safe, and more clearly distinguish between a
raw message and serialized data.
This also defines a macro to create new opaque pointer types, and
switches `RawArena` to using it.
PiperOrigin-RevId: 552957136
This CL sets up the basic plumbing end-to-end for singular message fields.
We add skeletonized support for `Proxied` messages. This is done
by creating structs for $Msg$View and $Msg$Mut, and providing
stubbed impls.
PiperOrigin-RevId: 552609955
This adds `#![deny(unsafe_op_in_unsafe_fn)]` which removes the
implicit `unsafe` block that `unsafe fn` does.
It also adds many more `SAFETY` docs, corrects some incomplete
ones, and catches a null pointer returned by `upb_Arena_New`.
PiperOrigin-RevId: 549067106
No need to say "represents" when describing a type (all types represent
something in real world, they are not the real thing), and "ABI-compatible"
needs a dash.
PiperOrigin-RevId: 546813197
Protobuf Team Bot
Marcel Hlopko
Alyssa Haroldsen
Hong Shin
Kevin King
Jakob Buchgraber
Sandy Zhang
Joshua Haberman
Adrian Sadłocha
Dmitri Gribenko