Specifically, after running code like the example code in the data conversion documentation to create parquet data (I assume the same is true for sqlite but have not checked), there is no single file that holds both pulse and truth information for all events, which is the only example given in the dataset documentation. To align with the data conversion documentation, the dataset documentation should include an example where path in the ParquetDataset constructor points to the directory containing the directories containing merged truth and pulse info.