Usage of these is pervasive in code which works with proto2, and proto2
will be with us for a long, long time to come. Move them to the proto
package.
Change-Id: I1b2e57429fd5a8f107a848a4492d20c27f304bd7
Reviewed-on: https://go-review.googlesource.com/c/protobuf/+/185543
Reviewed-by: Joe Tsai <thebrokentoaster@gmail.com>
The invalidExtensions flag no longer seems necessary. Tests pass without it.
Change-Id: Ieb35e26912b047718ccbfcdc926625aec1cd8c87
Reviewed-on: https://go-review.googlesource.com/c/protobuf/+/185937
Reviewed-by: Damien Neil <dneil@google.com>
The proper semantics for a message field within a oneof
when unmarshaling is to merge into an existing message,
rather than replacing it.
Change-Id: I7c08f6e4fa958c6ee6241e9083f7311515a97e15
Reviewed-on: https://go-review.googlesource.com/c/protobuf/+/185957
Reviewed-by: Damien Neil <dneil@google.com>
Move the Go cache from .cache/gocache to .gocache.
This moves it out of the .cache directory so that Travis-CI will
not cache the Go cache.
Unfortunately, the Go toolchain's caching algorithm is not
aggressive enough in evicting old entries, causing Travis-CI
to keep caching an ever growing Go cache.
We're at the point where network IO moving a massive Go cache
is more costly than not having it at all.
Change-Id: I3104efcdb8fa81a550900e8d06299e50296386f3
Reviewed-on: https://go-review.googlesource.com/c/protobuf/+/185838
Reviewed-by: Herbie Ong <herbie@google.com>
When setting a proto3 bytes field to an empty, non-nil bytes slice,
just store a nil slice in the underderlying storage.
This is done to avoid presenting the illusion to the user that
presence is preserved for proto3.
Updates golang/protobuf#896
Change-Id: I1b97bedd547d336863c65d9418d8f07edf69ccd6
Reviewed-on: https://go-review.googlesource.com/c/protobuf/+/185577
Reviewed-by: Herbie Ong <herbie@google.com>
We define MessageState, which is essentially an atomically set *MessageInfo.
By nesting this as the first field in every generated message, we can
implement the reflective methods on a *MessageState when obtained by
unsafe casting a concrete message pointer as a *MessageState.
The MessageInfo held by MessageState provides additional Go type information
to interpret the memory that comes after the contents of the MessageState.
Since we are nesting a MessageState in every message,
the memory use of every message instance grows by 8B.
On average, the body of ProtoReflect grows from 133B to 202B (+50%).
However, this is offset by XXX_Methods, which is 108B and
will be removed in a future CL. Taking into account the eventual removal
of XXX_Methods, this is a net reduction of 25%.
name old time/op new time/op delta
Name/Value-4 70.3ns ± 2% 17.5ns ± 6% -75.08% (p=0.000 n=10+10)
Name/Nil-4 70.6ns ± 3% 33.4ns ± 2% -52.66% (p=0.000 n=10+10)
name old alloc/op new alloc/op delta
Name/Value-4 16.0B ± 0% 0.0B -100.00% (p=0.000 n=10+10)
Name/Nil-4 16.0B ± 0% 0.0B -100.00% (p=0.000 n=10+10)
name old allocs/op new allocs/op delta
Name/Value-4 1.00 ± 0% 0.00 -100.00% (p=0.000 n=10+10)
Name/Nil-4 1.00 ± 0% 0.00 -100.00% (p=0.000 n=10+10)
Change-Id: I92bd58dc681c57c92612fd5ba7fc066aea34e95a
Reviewed-on: https://go-review.googlesource.com/c/protobuf/+/185460
Reviewed-by: Damien Neil <dneil@google.com>
The strs.UnsafeString casts a []byte as a string.
This allows us to avoid duplicated functionality.
Change-Id: I9930b94bae35eac0f98c0fa62963b300bc8d7e49
Reviewed-on: https://go-review.googlesource.com/c/protobuf/+/185459
Reviewed-by: Herbie Ong <herbie@google.com>
We already support unmarshaling weak fields, support the other direction.
Change-Id: I514e6b7b18cf9a3567fb1775a1e389fe64f22d22
Reviewed-on: https://go-review.googlesource.com/c/protobuf/+/185340
Reviewed-by: Herbie Ong <herbie@google.com>
When decoding, only treat the name as being explicitly set if the
name differs from what the JSON name would have been had it been
automatically derived from the protobuf field name.
Change-Id: Ida9256eafe7af20c7a06be2d4fb298be44276104
Reviewed-on: https://go-review.googlesource.com/c/protobuf/+/185398
Reviewed-by: Herbie Ong <herbie@google.com>
Create a new internal/strs package that unifies common functionality:
* Since protobuf itself pseudo-specifies at least 4 different camel-case
and snake-case conversion functions, we define all variants in one place.
* We move the internal/filedesc.nameBuilder function to this package.
We simplify its implementation to not depend on a strings.Builder fork
under the hood since the semantics we desire is simpler than what
strings.Builder provides.
* We use strs.Builder in reflect/protodesc in its construction of all
the full names. This is perfect use case of strs.Builder since all
full names within a file descriptor share the same lifetime.
* Add an UnsafeString and UnsafeBytes cast function that will be useful
in the near future for optimizing encoding/prototext and encoding/protojson.
Change-Id: I2cf07cbaf6f72e5f9fd6ae3d37b0d46f6af2ad59
Reviewed-on: https://go-review.googlesource.com/c/protobuf/+/185198
Reviewed-by: Damien Neil <dneil@google.com>
Make the fast-path and slow-path versions of IsInitialized report
exactly the same errors: An errors.RequiredNotSet containing the
full name of one of the unset required fields.
Bugfix: Fast-path IsInitialized on a nil message reports an error only
when the message directly contains required fields.
Bugfix: Include fast-path IsInitialized in legacy messageIfaceWrapper.
Fixesgolang/protobuf#887
Change-Id: Ia5e4b386f8c23f6f855d995f4a098b1338acbae3
Reviewed-on: https://go-review.googlesource.com/c/protobuf/+/185397
Reviewed-by: Joe Tsai <thebrokentoaster@gmail.com>
Don't fail on test cases with partial messages.
Create a new message on each decode cycle.
Change-Id: I7c1d91a2853e340fc0bae05a238bb28eb682e6ce
Reviewed-on: https://go-review.googlesource.com/c/protobuf/+/185377
Reviewed-by: Joe Tsai <thebrokentoaster@gmail.com>
The logic for this will be removed entirely in a future CL.
Change-Id: I3a20fb2a207fc5bedc550b3f6deeab7f63eaa2c4
Reviewed-on: https://go-review.googlesource.com/c/protobuf/+/185243
Reviewed-by: Herbie Ong <herbie@google.com>
The encoding/prototext and encoding/protojson are implemented entirely
in terms of protobuf reflection, which side-steps this information.
Remove the hacks in the generator to special-case MessageSet.
Change-Id: I708c4636b77672545a103b7ab686f103b9dfc514
Reviewed-on: https://go-review.googlesource.com/c/protobuf/+/185240
Reviewed-by: Herbie Ong <herbie@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>
Associate the oneof wrapper types with a message by conveying that
information to the associated MessageInfo.
Change-Id: Iabfca593850e1d6a89498a37eacbf22dbb73bd20
Reviewed-on: https://go-review.googlesource.com/c/protobuf/+/185239
Reviewed-by: Damien Neil <dneil@google.com>
This test verifies that the inialization logic does not accidentally
execute the lazy initialization logic that is intended to only evaluate
upon first use.
Change-Id: I5e9dea17ce88081c78195af0e86dd38223689f63
Reviewed-on: https://go-review.googlesource.com/c/protobuf/+/184259
Reviewed-by: Damien Neil <dneil@google.com>
The code organization is simpler if we keep the functions encoding and
decoding a particular type (e.g., maps) together rather than split
across files.
This rename is happening in a separate CL from cl/185241 to preserve
rename history. (Git gets confused when you rename a->b and b->c in the
same commit.)
Change-Id: Idfbb3ff8cf0db149c68d650f89ff3fb8ac833322
Reviewed-on: https://go-review.googlesource.com/c/protobuf/+/184942
Reviewed-by: Joe Tsai <thebrokentoaster@gmail.com>
The code organization is simpler if we keep the functions encoding and
decoding a particular type (e.g., maps) together rather than split
across files. Rename various "encode" files to "codec" in preparation
for adding fast-path decoding.
Change-Id: If1e271da99d31533ffefc19b1fc847936fa9484a
Reviewed-on: https://go-review.googlesource.com/c/protobuf/+/185241
Reviewed-by: Joe Tsai <thebrokentoaster@gmail.com>
The protoreflect.Descriptor.Options method is currently documented as
returning a reference to the options, where the user must not mutate
the returned message. This changes internal/filedesc to avoid returning
a copy of the options by caching the first unmarshal.
See golang/protobuf#877
Change-Id: I15701d33fbda7535b21b2add72628b02992c373f
Reviewed-on: https://go-review.googlesource.com/c/protobuf/+/185197
Reviewed-by: Benny Siegert <bsiegert@gmail.com>
Hyrum's Law dictates that if we do not prevent naughty behavior,
people will rely on it. If we do not validate that the provided
file descriptor is correct today, it will be near impossible
to add proper validation checks later on.
The logic added validates that the provided file descriptor is
correct according to the same semantics as protoc,
which was reversed engineered to derive the set of rules implemented here.
The rules are unfortunately complicated because protobuf is a language
full of many non-orthogonal features. While our logic is complicated,
it is still 1/7th the size of the equivalent C++ code!
Change-Id: I6acc5dc3bd2e4c6bea6cd9e81214f8104402602a
Reviewed-on: https://go-review.googlesource.com/c/protobuf/+/184837
Reviewed-by: Damien Neil <dneil@google.com>
Overview of changes:
* Add an option that specifies whether to replace unresolvable references
with a placeholder instead of producing an error. Since the prior behavior
produced placeholders (not always), we default to that behavior for now,
but will enable strict resolving in a future CL.
* The option is not yet exported because there is concern about what the
public API should look like. This will be exposed in a future CL.
* Unlike before, we now permit placeholders for unresolvable enum values.
* We implement relative name resolution logic.
* We handle the case where the type is unknown, but type_name is specified.
In such a case, we populate both FieldDescriptor.{Enum,Message} and leave
the FieldDescriptor.Kind with the zero value. If the type_name happened
to resolve, we use that to determine the type.
* If a placeholder is used to represent a relative name,
the FullName reports an invalid full name with a "*." prefix.
Change-Id: Ifa8c750423c488fb9324eec4d033a2f251505fda
Reviewed-on: https://go-review.googlesource.com/c/protobuf/+/184317
Reviewed-by: Damien Neil <dneil@google.com>
There is little performance benefit to aliasing the input since we copy
every field except the options. Thus, just go all the way and copy the
options as well and document this as such.
Change-Id: If6ca5ce0ee03c9f76e528023b6056ad99d3ca209
Reviewed-on: https://go-review.googlesource.com/c/protobuf/+/184879
Reviewed-by: Damien Neil <dneil@google.com>
This does not remove all dependencies,
but all of the cases where it can now be implemented in terms of v2.
Change-Id: Idc5b0273f0d35c284bf2141eb9cce998692ceb15
Reviewed-on: https://go-review.googlesource.com/c/protobuf/+/184878
Reviewed-by: Herbie Ong <herbie@google.com>
Previously, when aberrantLoadMessageDesc returned it was guaranteed
to have initialized the current message through the use of the done signal.
However, this does not guarantee that the descriptor for a cylic reference
has also finished initialization.
Rather than add more complicated logic to wait until all cyclic references
have finished initializing, just add a global lock for the entire
aberrantLoadMessageDesc function.
This slows down performance, but is easier to reason about.
Change-Id: I4cdae8b955f71ee40fa6979f5a8d548d9749042c
Reviewed-on: https://go-review.googlesource.com/c/protobuf/+/184657
Reviewed-by: Damien Neil <dneil@google.com>
This fixes a bug introduced by CL/182360.
Overview of the problem:
* CL/182360 removes the internal/prototype package, such that
protodesc was re-implemented using internal/filedesc.
* As a result of that change, resolving internal dependencies became
the responsibility of protodesc.
* Dependency resolution used the following two-pass algorithm:
1) first pass derives the full name of all declarations
2) second pass fully initializes each descriptor declaration,
now being able to resolve local dependencies from the previous step.
* When the second pass looks up a local dependency, it is guaranteed to
find it, but it is not guaranteed that the dependency has been initialized
(since it may appear later on). This is problematic for default enum values
since it implies that the enum dependency may not be sufficiently
initialized to be able to query its set of values, leading to panics.
* CL/182360 recognized the problem and attempted to enforce an initialization
ordering where nested enums were always initialized before the body of the
message declaration itself.
* However, that ordering fails to enforce that that enum declarations outside
the parent tree are initialized beforehand. For example, referring to an
enum value that is declared within a sibling of the parent message.
* This CL fixes the problem with a three-pass algorithm:
1) first pass derives the full name *and* fully initialize the
entire descriptor *except* for dependency references (i.e., type_name).
2) second pass only resolves dependency references,
where we do not need to worry about initialization ordering.
3) third pass validates the descriptors are well-formed.
This can now depend on all information being fully initialized.
* While a lot of code moves, this change is actually very mechanical.
Other than split things apart, no new logic is introduced nor removed.
Change-Id: Ia91d4aade8f6187c19d704d43ae96b3b9d276792
Reviewed-on: https://go-review.googlesource.com/c/protobuf/+/184297
Reviewed-by: Damien Neil <dneil@google.com>
The aberrant support logic only has access to the Go type
information, and not a concrete value. However, the XXX_MessageName
method exists on some hacky dynamic proto implementations where
it is only valid to call on a concrete value, not just newly created
instance of the given type.
However, from the perspective of the support logic, it is impossible
to distinguish between dynamic messages and hand-crafted custom messages.
Thus, just drop support for XXX_MessageName. We won't get the full name
of the message right, but oh well, what can we do.
Change-Id: Icc272861e11a355639fb82a991ca2854a9edc0c7
Reviewed-on: https://go-review.googlesource.com/c/protobuf/+/184557
Reviewed-by: Damien Neil <dneil@google.com>
Aberrant messages are hand-crafted messages that happen to work because
they use the same struct tags that generated code emits.
This happens to work in v1, but is unspecified behavior and entirely outside
the compatibility promise.
Support for this was added early on in the history of the v2 implementation,
but entirely untested. It was removed in CL/182360 to reduce the
technical debt of the legacy implementation. Unfortunately, sufficient number
of targets do rely on this aberrant support, so it is being added back.
The logic being added is essentially the same thing as the previous logic,
but ported to use internal/filedesc instead of the now deleted
internal/prototype package.
Change-Id: Ib5cab3e90480825b9615db358044ce05a14b05bd
Reviewed-on: https://go-review.googlesource.com/c/protobuf/+/184517
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>
Len looks like it should be O(1), but the need to check for
non-zero-length repeated fields makes it at minimum O(n) where n is
the number of repeated fields. In practice, it's O(n) where n is the
number of fields altogether.
The Len function is not especially useful, easily duplicated with Range
and a counter, and can be surprisingly inefficient. Drop it.
Change-Id: I24b27433217e131e842bd18dd58475bcdf62ef97
Reviewed-on: https://go-review.googlesource.com/c/protobuf/+/183678
Reviewed-by: Joe Tsai <thebrokentoaster@gmail.com>
Previous code tries to do all-lowercase match all non-extension field
names to match group field names, which is incorrect. Fix to check for
only group field type and make sure that the format is correct as well.
Fixesgolang/protobuf#878.
Fix typo in text proto string in internal/impl/message_test.go that
wasn't caught before due to above issue.
Change-Id: Ief952907306435ed76a095e96e29fcc9c0027b73
Reviewed-on: https://go-review.googlesource.com/c/protobuf/+/183737
Reviewed-by: Joe Tsai <thebrokentoaster@gmail.com>
This is a breaking change.
The replacement is the Files.FindDescriptorByName method,
which is more flexible as it handles all descriptor types.
Change-Id: I2ccd544a7630396a2428b1d41f836c5246070912
Reviewed-on: https://go-review.googlesource.com/c/protobuf/+/183700
Reviewed-by: Damien Neil <dneil@google.com>
Preserving the unknown enum in the String method helps errors
produced by reflect/protodesc be more informative.
Change-Id: I8efb09cb3c744bf4483b310053df7686da540387
Reviewed-on: https://go-review.googlesource.com/c/protobuf/+/183699
Reviewed-by: Damien Neil <dneil@google.com>
This change makes it such that Files now functionally registers all
descriptors in a file (not just enums, messages, extensions, and services),
but also including enum values, messages fields/oneofs, and service methods.
The ability to look up any descriptor by full name is needed to:
1) properly detect namespace conflicts on enum values
2) properly implement the relative name lookup logic in reflect/protodesc
The approach taken:
1) Assumes that a FileDescriptor has no internal name conflicts.
This will (in a future CL) be guaranteed by reflect/protodesc and
is guaranteed today by protoc for generated descriptors.
2) Observes that the only declarations that can possibly conflict
with another file are top-level declarations (i.e., enums, enum values,
messages, extensions, and services). Enum values are annoying
since they live in the same scope as the parent enum, rather than
being under the enum.
For the internal data structure of Files, we only register the top-level
declarations. This is the bare minimum needed to detect whether the file
being registered has any namespace conflicts with previously registered files.
We shift the effort to lookups, where we now need to peel off the end fragments
of a full name until we find a match in the internal registry. If a match
is found, we may need to descend into that declaration to find a nested
declaration by name.
For initialization, we modify internal/filedesc to initialize the
enum values for all top-level enums. This performance cost is offsetted
by the fact that Files.Register now avoids internally registering
nested enums, messages, and extensions.
For lookup, the cost has shifted from O(1) to O(N),
where N is the number of segments in the full name.
Top-level descriptors still have O(1) lookup times.
Nested descriptors have O(M) lookup times,
where M is the level of nesting within a single file.
Change-Id: I950163423431f04a503b6201ddcc20a62ccba017
Reviewed-on: https://go-review.googlesource.com/c/protobuf/+/183697
Reviewed-by: Damien Neil <dneil@google.com>
The dynamicpb package permits creating Message values from a
MessageDescriptor.
Change-Id: Ice429ae45a0835dffb5a7ec8c0bd2c1df7aac8a2
Reviewed-on: https://go-review.googlesource.com/c/protobuf/+/174960
Reviewed-by: Joe Tsai <thebrokentoaster@gmail.com>
