pub fn read_edges(
dir: &Path,
rel_name: &str,
mode: OntologyMode,
) -> Result<Vec<RecordBatch>, DataFusionError>Expand description
Read all edge rows for relation rel_name from the project at dir.
The on-disk layout mirrors GraphWriter:
- Strict / Advisory — typed edges in
topology/edges/<rel_name>.parquet(TYPED_EDGE_SCHEMA). - Exploratory — all edges in
topology/edges/_exploratory.parquet(EXPLORATORY_EDGE_SCHEMA, carrying arel_type_namecolumn). The returned batch is not filtered byrel_name; callers that need a single relation must filter onrel_type_namethemselves.
Returns a single (possibly empty) RecordBatch with the mode-appropriate
schema; a missing file yields an empty batch rather than an error.
§Errors
Returns DataFusionError::Execution if rel_name is not a plain file
stem (contains path separators or ..), and propagates Parquet / Arrow
errors encountered while reading.