Implement support in the protobuf runtime to better understand
message types that are not generated by the official generator.
In particular:
* Add a best-effort implementation of protobuf reflection for
"non-nullable" fields which are supposed to be represented by *T,
but are instead represented by a T. "Non-nullable" message fields
report presence based on whether the message is the zero Go value.
* We do NOT implement support for "non-nullable" fields in the
table-driven implementation since we assume that the aberrant messages
that we care about have a Marshal and Unmarshal method.
* We better handle custom messages that implement Marshal and Unmarshal,
but do NOT implement Merge. In that case, we implement merge in terms of
a back-to-back marshal and unmarshal.
* We better tolerate the situations where a protobuf message field
cannot be mapped to a Go struct field since the latter is missing.
In such cases, reflection treats the field as if it were unpopulated.
Setting such fields will panic.
This change allows the runtime to handle all message types declared
in the "go.etcd.io/etcd" and "k8s.io" modules where protobuf reflection,
Marshal, Unmarshal, Reset, Merge, and Equal all work.
The only types that still do not fully work are:
* "k8s.io/api/authentication/v1".ExtraValue
* "k8s.io/api/authentication/v1beta1".ExtraValue
* "k8s.io/api/authorization/v1".ExtraValue
* "k8s.io/api/authorization/v1beta1".ExtraValue
* "k8s.io/api/certificates/v1".ExtraValue
* "k8s.io/api/certificates/v1beta1".ExtraValue
* "k8s.io/apimachinery/pkg/apis/meta/v1".MicroTime
* "k8s.io/apimachinery/pkg/apis/meta/v1".Time
* "k8s.io/apimachinery/pkg/apis/meta/v1".Verbs
While Marshal, Unmarshal, Reset, and Merge continue to work,
protobuf reflection and any functionality that depends on it
(e.g., prototext, protojson, Equal, etc.) will not work.
Change-Id: I67a9d2f1bec35248045ad0c16220d02fc2e0e172
Reviewed-on: https://go-review.googlesource.com/c/protobuf/+/300869
Trust: Joe Tsai <joetsai@digital-static.net>
Trust: Joe Tsai <thebrokentoaster@gmail.com>
Reviewed-by: Damien Neil <dneil@google.com>
This CL adds runtime support for unknown fields to be represented
as *[]byte in addition to the current representation as []byte.
This CL does not change generated code to use *[]byte.
Comparison between using *[]byte and []byte:
• Every message supports unknown fields, so use of []byte
expands a message size by 24B (for 64-bit systems).
In contrast, *[]byte only expands a message by 8B.
This has significant memory implications for small messages.
• If unknown fields are encountered, *[]byte has extra overhead
allocating the 24B slice header. However, it is assumed
that messages rarely see any unknown fields at runtime,
or generally do so for a temporary period of time.
Change-Id: I81935e4ea7394166e61ff4579f76f59fa792dfc9
Reviewed-on: https://go-review.googlesource.com/c/protobuf/+/244937
Reviewed-by: Damien Neil <dneil@google.com>
The order package replaces the mapsort and fieldsort packages.
It presents a common API for ordered iteration over message fields
and map fields.
It has a number of pre-defined orderings.
Change-Id: Ie6cd423da30b4757864c352cb04454f21fe07ee2
Reviewed-on: https://go-review.googlesource.com/c/protobuf/+/239837
Reviewed-by: Herbie Ong <herbie@google.com>
Any attempt at guessing the size for a fixed coderFieldInfo array
will always get it wrong in some cases, either by under-estimating
or over-estimating the count. The former causes worse caching behavior,
while the latter causes memory waste.
As a middle ground, just pre-allocate a slice of the exact length.
Each element will have memory locality with each other, but not
be guaranteed to have memory locality with the parent coderMessageInfo.
name old time/op new time/op delta
EmptyMessage/Wire/Marshal-8 43.1ns ±11% 42.6ns ± 8% -1.32% (p=0.036 n=50+49)
EmptyMessage/Wire/Unmarshal-8 18.6ns ±10% 18.9ns ±12% ~ (p=0.054 n=50+50)
EmptyMessage/Wire/Validate-8 15.0ns ± 9% 14.7ns ±10% -2.44% (p=0.002 n=50+45)
EmptyMessage/Clone-8 163ns ±20% 149ns ±19% -8.58% (p=0.000 n=48+53)
RepeatedInt32/Wire/Marshal-8 4.27µs ±12% 4.24µs ±13% ~ (p=0.612 n=48+52)
RepeatedInt32/Wire/Unmarshal-8 3.47µs ±14% 3.50µs ±11% ~ (p=0.217 n=50+53)
RepeatedInt32/Wire/Validate-8 2.12µs ±12% 2.09µs ± 9% ~ (p=0.121 n=50+51)
RepeatedInt32/Clone-8 3.04µs ±18% 2.98µs ±36% ~ (p=0.289 n=51+54)
Required/Wire/Marshal-8 281ns ±14% 276ns ±11% ~ (p=0.059 n=48+55)
Required/Wire/Unmarshal-8 117ns ±14% 118ns ±11% ~ (p=0.358 n=49+53)
Required/Wire/Validate-8 87.6ns ± 9% 88.0ns ±12% ~ (p=0.373 n=48+53)
Required/Clone-8 533ns ±12% 507ns ±15% -4.71% (p=0.000 n=49+54)
Change-Id: I4cf3134e424130bee728b7591127e5c80f07e2db
Reviewed-on: https://go-review.googlesource.com/c/protobuf/+/232937
Reviewed-by: Damien Neil <dneil@google.com>
In the upcoming 3.12.x release of protoc, the proto3 language will be
amended to support true presence for scalars. This CL adds support
to both the generator and runtime to support these semantics.
Newly added public API:
protogen.Plugin.SupportedFeatures
protoreflect.FieldDescriptor.HasPresence
protoreflect.FieldDescriptor.HasOptionalKeyword
protoreflect.OneofDescriptor.IsSynthetic
Change-Id: I7c86bf66d0ae56642109beb5f2132184593747ad
Reviewed-on: https://go-review.googlesource.com/c/protobuf/+/230698
Reviewed-by: Damien Neil <dneil@google.com>
This particular off-by-one can never happen in practice. We will crash
if a field has a number one greater than the maximum field number in the
dense array, but if field number N is in the array we will always put
N+1 in it as well.
Fix the off-by-one anyway.
Change-Id: I8c1304f2fc0d7b91036bde3f7ddb7115c21781ca
Reviewed-on: https://go-review.googlesource.com/c/protobuf/+/229278
Reviewed-by: Joe Tsai <joetsai@google.com>
Microbenchmarks are inconclusive/noisy, but shows a small but noticeable
improvement on internal benchmarks.
Change-Id: Ic46c6aac8a42c4dc749c4f3583d8c8c95e9548b7
Reviewed-on: https://go-review.googlesource.com/c/protobuf/+/229277
Reviewed-by: Joe Tsai <joetsai@google.com>
An Is prefix implies it returns a boolean.
A Check prefix better suggests that it could return an error.
Change-Id: I6ffcb32099a944c656c07654c294a0980efb2d0e
Reviewed-on: https://go-review.googlesource.com/c/protobuf/+/220338
Reviewed-by: Damien Neil <dneil@google.com>
Refactor the fast-path size, marshal, unmarshal, and isinit functions to
take the *coderFieldInfo for the field as input.
This replaces a number of closures capturing field-specific information
with functions taking that information as an explicit parameter.
Change-Id: I8cb39701265edb7b673f6f04a0152d5f4dbb4d5d
Reviewed-on: https://go-review.googlesource.com/c/protobuf/+/218937
Reviewed-by: Joe Tsai <thebrokentoaster@gmail.com>
We may want to make changes to the inputs and outputs of the fast-path
functions in the future. For example, we likely want to add the ability
for the fast-path unmarshal to report back whether the unmarshaled
message is known to be initialized.
Change the signatures of these functions to take in and return struct
types which can be extended with whatever fields we want in the future.
Change-Id: Idead360785df730283a4630ea405265b72482e62
Reviewed-on: https://go-review.googlesource.com/c/protobuf/+/215719
Reviewed-by: Joe Tsai <joetsai@google.com>
This adds a experimental function to the internal/impl package which
validates a wire-format message against a message type. The validator
reports whether the message can be successfully unmarshaled, and whether
the result is initialized (all required fields are set). In some cases,
the validator returns ambiguous results when full validation would be
expensive.
The validator is unused outside of tests. In the future, it may be used
to permit lazy unmarshaling of some data. It is being added now for
testing; in particular, the wire fuzzer now checks the validator output
for consistency with the unmarshaler.
The validator adds a small amount of unused per-MessageType state. If
this becomes a concern, we could conditionalize it with a build tag.
Change-Id: I4216ef81d6a9ed975302eed189b02d08608858b4
Reviewed-on: https://go-review.googlesource.com/c/protobuf/+/212302
Reviewed-by: Joe Tsai <thebrokentoaster@gmail.com>
Change size, marshal, and isinit operations on oneofs to look up the
currently-set oneof type in a map rather than testing for each possible
oneof field in turn.
Significantly improves oneof encoding speed for oneofs with a
substantial number of fields:
go test ./proto -bench=./oneof.*string.*test.TestAll -benchmem -count=8 -cpu=1
name old time/op new time/op delta
Encode/oneof_(string)_(*test.TestAllTypes) 911ns ± 1% 397ns ± 3% -56.45% (p=0.000 n=8+7)
Decode/oneof_(string)_(*test.TestAllTypes) 899ns ± 1% 922ns ± 1% +2.49% (p=0.001 n=7+7)
Fixesgolang/protobuf#950
Change-Id: I9393a87975ce09011d885a8af4a63a639ea8452f
Reviewed-on: https://go-review.googlesource.com/c/protobuf/+/210281
Reviewed-by: Joe Tsai <thebrokentoaster@gmail.com>
Stash fast-path information for extensions on the ExtensionInfo. In
the usual case where an ExtensionType's underlying implementation is
an *ExtensionInfo, fetching the fast-path information becomes a type
assertion rather than a mutex-guarded map access.
Maintain a global sync.Map for the case where an ExtensionType isn't an
*ExtensionInfo.
Substantially improves performance for fast-path operations on
extensions:
Encode/MessageSet_type_id_before_message_content-12 267ns ± 1% 185ns ± 1% -30.44% (p=0.001 n=7+7)
Encode/basic_scalar_types_(*test.TestAllExtensions)-12 1.94µs ± 1% 0.40µs ± 1% -79.32% (p=0.000 n=8+7)
Change-Id: If048b521deb3665a090ea3d0a178c61691d4201e
Reviewed-on: https://go-review.googlesource.com/c/protobuf/+/210540
Reviewed-by: Joe Tsai <joetsai@google.com>
The v1 wire marshaler sorts fields as follows:
- All extensions, sorted by field number.
- All non-oneof fields, sorted by field number.
- All oneof fields, in indeterminate order.
We already make some steps toward supporting this ordering: The
fast path encoder places extensions in sorted order at the start
of the message.
This commit moves oneof fields to the end of the message, makes the
reflection-based encoder use this ordering when deterministic marshaling
is enabled, and adds a test to catch unintentional changes to the
ordering.
Users SHOULD NOT depend on stability of the marshal output. It is
subject to change over time. Without deterministic marshaling enabled,
it is subject to change over calls to Marshal.
Change-Id: I6cfd89090d790a3bb50785f32b94d2781d7d08db
Reviewed-on: https://go-review.googlesource.com/c/protobuf/+/206800
Reviewed-by: Joe Tsai <thebrokentoaster@gmail.com>
In the v1 implementation, unknown MessageSet items are stored in a
message's unknown fields section in non-MessageSet format. For example,
consider a MessageSet containing an item with type_id T and value V.
If the type_id is not resolvable, the item will be placed in the unknown
fields as a bytes-valued field with number T and contents V. This
conversion is then reversed when marshaling a MessageSet containing
unknown fields.
Preserve this behavior in v2.
One consequence of this change is that actual unknown fields in a
MessageSet (any field other than 1) are now discarded. This matches
the previous behavior.
Change-Id: I3d913613f84e0ae82481078dbc91cb25628651cc
Reviewed-on: https://go-review.googlesource.com/c/protobuf/+/205697
Reviewed-by: Joe Tsai <thebrokentoaster@gmail.com>
Call the Marshal or Unmarshal method on legacy messages implementing
protoV1.Marshaler or protoV2.Unmarshaler.
We do this in the impl package by creating an appropriate function in
the protoiface.Methods struct for legacy messages.
In proto.MarshalAppend, return the bytes provided by the fast-path
marshal function even when the returned error is non-nil.
Fixesgolang/protobuf#955
Change-Id: I36924af9ff959a946c43f2295ef3202216e81b32
Reviewed-on: https://go-review.googlesource.com/c/protobuf/+/197357
Reviewed-by: Joe Tsai <joetsai@google.com>
Changes made:
* Move MessageInfo.{methods,extensionFieldInfosMu,extensionFieldInfos}
to the coderMessageInfo type since they are fields all related to the
fast-path implementation, which is what the coderMessageInfo is for.
* Rename message_field.go -> message_reflect_field.go to make it obvious
from the file name that this only deals with the reflection implementation.
* Rename message_test.go -> message_reflect_test.go.
* Move reflection-specific implementation functions from message.go
to message_reflect.go. The intention is to make it such that message.go
is the entry point to message implementations and is agnostic towards
whether we are implementing reflection or the table-driven fast path.
Overall, there is no semantic changes. Just code movement.
Change-Id: I7743c39ba84dc63cd2a02934c3319285e16d6b1c
Reviewed-on: https://go-review.googlesource.com/c/protobuf/+/190100
Reviewed-by: Herbie Ong <herbie@google.com>
Remove the remaining uses of the prototype package.
The most significant change is to impl.MessageInfo, which now directly
implements the MessageType interface. This involves two notable changes
to exported fields of MessageInfo:
- PBType is now Desc.
- GoType is now GoReflectType. (Name changed to avoid a conflict with
the GoType method of the MessageType interface.)
Fixesgolang/protobuf#911
Change-Id: Ie2aa4766d6887ceaa9cf06b1f109aa6e6e2a208f
Reviewed-on: https://go-review.googlesource.com/c/protobuf/+/189340
Reviewed-by: Joe Tsai <thebrokentoaster@gmail.com>
Change protoiface.ExtensionDescV1 to implement protoreflect.ExtensionType.
ExtensionDescV1's Name field conflicts with the Descriptor Name method,
so change the protoreflect.{Message,Enum,Extension}Type types to no
longer implement the corresponding Descriptor interface. This also leads
to a clearer distinction between the two types.
Introduce a protoreflect.ExtensionTypeDescriptor type which bridges
between ExtensionType and ExtensionDescriptor.
Add extension accessor functions to the proto package:
proto.{Has,Clear,Get,Set}Extension. These functions take a
protoreflect.ExtensionType parameter, which allows writing the
same function call using either the old or new API:
proto.GetExtension(message, somepb.E_ExtensionFoo)
Fixesgolang/protobuf#908
Change-Id: Ibc65d12a46666297849114fd3aefbc4a597d9f08
Reviewed-on: https://go-review.googlesource.com/c/protobuf/+/189199
Reviewed-by: Joe Tsai <thebrokentoaster@gmail.com>
MessageSets are a deprecated proto1 feature, long since superseded by
extensions. Add disabled-by-default support behind flags.Proto1Legacy.
Change-Id: I7d3ace07f3b0efd59673034f3dc633b908345a88
Reviewed-on: https://go-review.googlesource.com/c/protobuf/+/185538
Reviewed-by: Joe Tsai <thebrokentoaster@gmail.com>
CL/174938 removed these methods in favor of a method that returned
only the descriptors. This CL adds back in the Type methods alongside
the Descriptor methods.
In a vast majority of protobuf usages, only the descriptor information
is needed. However, there is a small percentage that legitimately needs
the Go type information. We should provide both, but document that the
descriptor-only information is preferred.
Change-Id: Ia0a098997fb1bd009994940ae8ea5257ccd87cae
Reviewed-on: https://go-review.googlesource.com/c/protobuf/+/184578
Reviewed-by: Damien Neil <dneil@google.com>
The following adjustments were made:
* The pragma.NoUnkeyedLiterals is moved to be the first field.
This is done to keep the options struct smaller. Even if the last
field is zero-length, Go GC implementation details forces the struct
to be padded at the end.
* Methods are documented as always treating AllowPartial as true.
* Added a support flag for UnmarshalOptions.DiscardUnknown.
Change-Id: I1f75d226542ab2bb0123d9cea143c7060df226d8
Reviewed-on: https://go-review.googlesource.com/c/protobuf/+/185998
Reviewed-by: Damien Neil <dneil@google.com>
We modify protoc-gen-go to stop generating exported XXX fields.
The unsafe implementation is unaffected by this change since unsafe
can access fields regardless of visibility. However, for the purego
implementation, we need to respect Go visibility rules as enforced
by the reflect package.
We work around this by generating a exporter function that given
a reference to the message and the field to export, returns a reference
to the unexported field value. This exporter function is protected by
a constant such that it is not linked into the final binary in non-purego
build environment.
Updates golang/protobuf#276
Change-Id: Idf5c1f158973fa1c61187ff41440acb21c5dac94
Reviewed-on: https://go-review.googlesource.com/c/protobuf/+/185141
Reviewed-by: Damien Neil <dneil@google.com>
Move data used by the fast-path implementations into a substructure of
MessageInfo and initialize it separately.
Change-Id: Ib855ee8ea5cb0379528b52ba0e191319aa5e2dff
Reviewed-on: https://go-review.googlesource.com/c/protobuf/+/184077
Reviewed-by: Joe Tsai <thebrokentoaster@gmail.com>
