adityamparikh opened a new pull request, #292:
URL: https://github.com/apache/solr-mcp/pull/292
## Summary
`get-schema` and `solr://{collection}/schema` now return a
`CollectionSchema` record instead of SolrJ's `SchemaRepresentation`.
```diff
getSchema(collection)
- schemaRequest.process(solrClient, collection) # SolrJ parses into
SchemaRepresentation
- .getSchemaRepresentation() # FieldTypeDefinition:
{attributes:{name,class,...},
- # analyzer:null,
multiTermAnalyzer:null, similarity:null}
+ solrClient.request(new SchemaRequest(), collection) # Schema API
JSON, objects as SimpleOrderedMap (a Map)
+ objectMapper.convertValue(response.get("schema"), CollectionSchema.class)
```
The records type what clients read and keep everything else:
```text
CollectionSchema name, version, uniqueKey, fieldTypes, fields,
dynamicFields, copyFields, other
Field name, type, indexed, stored, docValues, multiValued,
required, other
FieldType name, className ("class"), analyzer, indexAnalyzer,
queryAnalyzer, multiTermAnalyzer, other
Analyzer charFilters, tokenizer, filters, other
CopyField source, dest, maxChars, other
```
`other` is a catch-all marked `@JsonAnySetter @JsonAnyGetter`. Jackson
collects each property the record does not declare into it
(`positionIncrementGap`, `vectorDimension`, `useDocValuesAsStored`,
`similarity`, …) and writes it back out at the same level, so what a client
receives is the Schema API's own JSON, and a property a later Solr version adds
is passed on instead of dropped.
The new records are registered in `SolrNativeHints.MCP_TOOL_RECORDS`; the
`SchemaRepresentation` hint is removed because nothing serializes it any more.
## Evidence
- **Before:** `get-schema` on `_default` returns 32,169 bytes: each field
type wrapped as `{"attributes":{"name":…,"class":…}, "analyzer":null,
"multiTermAnalyzer":null, "similarity":null, …}`.
**After:** 23,261 bytes, byte-for-byte the size of `curl …/schema`'s
`schema` section:
`{"name":"text_general","class":"solr.TextField","positionIncrementGap":"100",
"indexAnalyzer":{…}, …}`.
- `SchemaServiceIntegrationTest.getSchema_serializesToTheSchemaApiResponse`:
the serialized record equals Solr's Schema API response, compared as parsed
JSON.
- `SchemaServiceTest.getSchema_keepsPropertiesTheRecordsDoNotDeclare`: an
unknown field property and an unknown top-level section reach the JSON at the
level Solr sent them, with no `other` key.
- `McpClientIntegrationTestBase.getSchema` (through Spring AI's
serialization): `class` sits at the top level of a field type, there is no
`attributes` wrapper and no `other` key, and the arrays are present.
- `./gradlew build`: 460 tests, 0 failures, 0 skipped.
- `./gradlew nativeTest -Pnative`: 303 successful, 0 failed; the round-trip
test and both MCP `getSchema` tests (HTTP and STDIO) pass in the native image.
The 154 skipped are the `@DisabledInNativeImage` Mockito tests.
## Merge Danger
**Door:** two-way
Reverting restores the SolrJ type.
**Blast Radius:** schema-consumers
The JSON shape of `get-schema` and `solr://{collection}/schema` changes:
field-type properties move out of `attributes` to the top level, as the Schema
API sends them, and the null analyzer slots disappear. A client that parsed the
SolrJ shape has to read the Schema API shape instead. Java callers of
`SchemaService.getSchema` get `CollectionSchema` instead of
`SchemaRepresentation`.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]