Make sure the stream ID reported when reading records using a mulitireader is the virtual stream ID

Summary: MultiRecordFileReader must use virtualized Stream IDs, to guarantee that stream IDs are unique, even when files read together have streams with identical stream IDs. This change ensures that when reading records through MultiRecordFileReader, we use the virtual stream ID, not the file level stream ID.

Reviewed By: liorda

Differential Revision: D85787188

fbshipit-source-id: b1606f5268bcac33c9c802b79a955fa306809d0e

Author: Georges Berenger

csrc/reader/MultiVRSReader.cpp

@@ -666,6 +666,8 @@ py::object OssMultiVRSReader::readNextRecordInternal() {
   }
 
   auto r = PyRecord(record, nextRecordIndex_, lastRecord_);
+  // the stream ID in record is the file's stream ID, not the unique stream ID
+  r.streamId = reader_.getUniqueStreamId(&record).getNumericName();
   nextRecordIndex_++;
   return py::cast(r);
 }