Implement #712: idiomatic protobuf map<K,V> support (read + write) - #719
Merged
Conversation
Maps are encoded on the wire exactly like a `repeated` entry sub-message
(key = tag 1, value = tag 2). This exposes them idiomatically as JSON Objects
in both directions, at streaming and databind levels.
Schema (TypeResolver / ProtobufField): a `map<K,V>` field (or, from a protoc
descriptor set, a `repeated` field whose entry message carries the `map_entry`
option) is resolved into a synthetic entry message plus a repeated, map-flagged
field. Key types are restricted to integral / bool / string.
Generator: a map field is written as a sequence of length-delimited entry
sub-messages; each writeFieldName opens an entry (key rendered from the JSON
field name per the key type) and primes the value field.
Parser: a map surfaces as START_OBJECT + FIELD_NAME(key)/value pairs +
END_OBJECT, mirroring the unpacked-repeated-message machinery (with an entry
context so end-offset tracking survives buffer reloads). Absent key/value decode
to proto3 defaults; a value that precedes the key is reported clearly.
Descriptor path (FileDescriptorSet): propagate the `map_entry` message option so
.desc-loaded maps get the same treatment as .proto ones. NOTE: this changes the
shape of .desc-loaded maps, which previously surfaced as an array of
{key,value} entries.
Also flips the two #708 tests that asserted maps throw, and updates the now-stale
notes in ProtobufSchemaPreprocessor. Schema generation from POJO Map fields
(ProtoBufSchemaVisitor.expectMapFormat) remains unsupported; left for a follow-up.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
While entries are iterated `_currentMessage` points at the synthetic entry type. The map exit taken when a different field tag follows restored `_parsingContext` and `_currentField` but not `_currentMessage`, so the replayed tag was resolved against the entry type whenever the `next`-field chain missed -- which it does for any tag gap after the map. proto3 omits unset fields, so such a gap is ordinary, and a document written by this module could fail to read back with "Undefined property (id 3, wire type 2) for message type map<m>". Also fixes a related case where entries split by another field decoded trailing entry bytes as a bogus `key` field. Restores it on the EOF exit too, matching what `_checkEnd()` already did. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Existing message-valued map coverage used flat value messages (scalar fields only). A value message with nesting of its own pushes further context levels through the same save/restore path the map exit has to unwind -- which is what a4d66df got wrong, and its regression test covers only the scalar shape. Six tests, all passing as written; regression guards, not fixes: - value message holding another message - value message holding a `repeated` field - value message declaring its own `map` (one synthetic entry type current while another is entered) - three entries of nested-message + repeated, then a field after the map - empty value message: pins the wire bytes, showing an empty value still writes an entry unlike an empty map - the a4d66df tag-gap shape with a message-valued map, one level deeper Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
A `null` map value has no protobuf encoding, so it writes as a key-only entry; read back, such an entry yields `null` rather than the empty message protobuf would consider it. An entry must produce some value token -- unlike a plain message field, which simply stays absent -- and the choice round-trips, but nothing pinned it, so either direction could be flipped unnoticed. Two tests: the message-valued case (both directions, plus the contrast with an empty value message, which does write `12 00`), and the scalar case, whose read direction `testAbsentKeyAndValueDefaults` already covers. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Maps are encoded on the wire exactly like a
repeatedentry sub-message (key = tag 1, value = tag 2). This exposes them idiomatically as JSON Objects in both directions, at streaming and databind levels.Schema (TypeResolver / ProtobufField): a
map<K,V>field (or, from a protoc descriptor set, arepeatedfield whose entry message carries themap_entryoption) is resolved into a synthetic entry message plus a repeated, map-flagged field. Key types are restricted to integral / bool / string.Generator: a map field is written as a sequence of length-delimited entry sub-messages; each writeFieldName opens an entry (key rendered from the JSON field name per the key type) and primes the value field.
Parser: a map surfaces as START_OBJECT + FIELD_NAME(key)/value pairs + END_OBJECT, mirroring the unpacked-repeated-message machinery (with an entry context so end-offset tracking survives buffer reloads). Absent key/value decode to proto3 defaults; a value that precedes the key is reported clearly.
Descriptor path (FileDescriptorSet): propagate the
map_entrymessage option so .desc-loaded maps get the same treatment as .proto ones. NOTE: this changes the shape of .desc-loaded maps, which previously surfaced as an array of {key,value} entries.Also flips the two #708 tests that asserted maps throw, and updates the now-stale notes in ProtobufSchemaPreprocessor. Schema generation from POJO Map fields (ProtoBufSchemaVisitor.expectMapFormat) remains unsupported; left for a follow-up.