serializers: skip_supertypes filter in RocksDbSerializer streaming write

Plumbed through to the Python convert_path_to_rocksdb wrapper.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

src/ifcopenshell-python/ifcopenshell/__init__.py

@@ -378,15 +378,24 @@ def stream2_from_string(data: str) -> Generator[dict]:
             yield inst
 
 
-def convert_path_to_rocksdb(ifcspf_path: Union[Path, str], rocksdb_path: Union[Path, str]) -> None:
+def convert_path_to_rocksdb(
+    ifcspf_path: Union[Path, str],
+    rocksdb_path: Union[Path, str],
+    skip_supertypes: Optional[list[str]] = None,
+) -> None:
     """Converts an IFC-SPF file on disk to the IfcOpenShell-specific
     RocksDB encoding. RocksDB is an embedded key-value store that allows
     partial reads and is therefore more memory efficient with larger files.
 
     :param ifcspf_path: Input file path - needs to exist
     :param rocksdb_path: RocksDB file path (directory) - may exist, but result may then be invalid
+    :param skip_supertypes: Optional list of entity type names. Any instance
+        whose declaration is-a (or subclasses) one of these names will be
+        omitted from the resulting RocksDB. Note: references to skipped
+        instances become dangling - reads that walk into them will fail.
     """
-    ser = ifcopenshell_wrapper.RocksDbSerializer(str(ifcspf_path), str(rocksdb_path), True)
+    skip = list(skip_supertypes) if skip_supertypes else []
+    ser = ifcopenshell_wrapper.RocksDbSerializer(str(ifcspf_path), str(rocksdb_path), True, skip)
     ser.finalize()
 
 

src/serializers/RocksDbSerializer.cpp

@@ -23,9 +23,10 @@ RocksDbSerializer::RocksDbSerializer(ifcopenshell::file* file, const std::string
 	output_file_->calculate_unit_factors = false;
 }
 
-RocksDbSerializer::RocksDbSerializer(const std::string& input_filename, const std::string& rocksdb_filename, bool stream)
+RocksDbSerializer::RocksDbSerializer(const std::string& input_filename, const std::string& rocksdb_filename, bool stream, const std::vector<std::string>& skip_supertypes)
 	: file_(input_filename)
 	, rocksdb_filename_(rocksdb_filename)
+	, skip_supertypes_(skip_supertypes)
 {
 }
 
@@ -144,6 +145,21 @@ void RocksDbSerializer::write_streaming_() {
 
 			const bool is_header = decl->schema() == &Header_section_schema::get_schema();
 
+			if (!is_header && decl->as_entity() && !skip_supertypes_.empty()) {
+				bool skip = false;
+				for (const auto& super : skip_supertypes_) {
+					if (decl->is(super)) {
+						skip = true;
+						break;
+					}
+				}
+				if (skip) {
+					streamer.references().clear();
+					streamer.inverses().clear();
+					continue;
+				}
+			}
+
 			std::vector<express::Base> simple_type_instances;
 
 			for (size_t i = 0; i < data->storage_->size(); i++) {

src/serializers/RocksDbSerializer.h

@@ -8,17 +8,21 @@
 
 #include <rocksdb/db.h>
 
+#include <string>
+#include <vector>
+
 class SERIALIZERS_API RocksDbSerializer : public Serializer {
 private:
 	rocksdb::DB* db_;
 	std::string rocksdb_filename_;
 	std::variant<ifcopenshell::file*, std::string> file_;
 	ifcopenshell::file* output_file_;
+	std::vector<std::string> skip_supertypes_;
 
 	void write_streaming_();
 public:
 	RocksDbSerializer(ifcopenshell::file* file, const std::string& rocksdb_filename);
-	RocksDbSerializer(const std::string& input_filename, const std::string& rocksdb_filename, bool stream);
+	RocksDbSerializer(const std::string& input_filename, const std::string& rocksdb_filename, bool stream, const std::vector<std::string>& skip_supertypes = {});
 
 	virtual ~RocksDbSerializer() {}