Format support, precisely
Opening a file
is only the first step.
Capabilities belong to the reader and workflow. Check the detected source before assuming every operation is available.
| Source | Available workflow | Boundary |
|---|---|---|
| CSV / TSV / DSV / TXT / LOG | Delimited table inspection and supported data operations | Delimiter, encoding and query setup matter. |
| JSON / JSONL / NDJSON | Structured records and nested-value inspection | Capabilities depend on source shape. |
| Parquet / Arrow IPC / Feather | Columnar table pages and supported query/copy paths | Do not infer identical operations from the extension. |
| JSONL / Parquet / Arrow folders | Bounded catalog, resumable cross-shard scans, projection and supported Hive partition pruning | No claim of complete global schema, sort or row count. |
| SQLite / DuckDB | Read-only catalog, explicit table selection, projected and parameterized pages | Stable database ordering needs suitable keys. |
| GZIP / Zstandard / BZIP2 | Bounded preview or explicit full disk-backed preparation | Full preparation needs time, disk and an output budget. |
| XLSX / XLS | Spreadsheet readers | Check workbook and reader capabilities in setup. |
| ZIP / TAR / WebDataset | Bounded archive-member catalog | WebDataset sample grouping and media browsing pending. |
| NPY / NPZ / Zarr / Safetensors | Array/tensor metadata and inventories | Typed value slicing is not yet a complete workflow. |
| TFRecord | Bounded framing and CRC inspection | Example / SequenceExample payload decoding pending. |
| HDF5 / Avro / ORC / MessagePack | File summaries | Not full record/group decoders. |
| ODS / XLSB / XML | Archive members or bounded text, as appropriate | Not complete workbook-cell or XML-tree decoding. |
Opening executable serialized objects such as pickle/joblib is not part of automatic data inspection. Inspect untrusted formats with the appropriate isolated tools.
Open a source and inspect its capabilities