Supported export formats

Diafunc tables export to 21 formats spanning six categories: tabular, columnar / structured, semi-structured, archive (database files), scientific, and statistical-software. Any registered format can be wrapped with one of four compression algorithms (gzip / bzip2 / xz / zstd) and / or a single-entry archive (zip / tar) to produce e.g. .parquet.gz or .csv.tar.xz.

The live catalogue is served at GET /api/v1/storage/table-exports/supported-formats — the same source of truth the Lab format picker uses for its dropdown and the CLI uses for diafunc table export-formats.

Need to import data into Diafunc? See supported file formats.

Row × column tables. CSV / TSV stringify every cell; Excel promotes numeric and boolean columns to native cell types when the column type is declared.

Format	Extension	MIME	Notes
CSV	.csv	text/csv	Defaults: , delimiter, " quote, ? null token. Override via ?delimiter= / ?quote= / ?nullValue= / ?quoteAll=.
TSV	.tsv	text/tab-separated-values	Same as CSV with tab delimiter.
Excel	.xlsx	application/vnd.openxmlformats…	Single sheet; numeric / boolean columns get native cell types.

Self-describing typed-column formats. Per-column ColumnSpec drives the on-disk type — Parquet / Arrow / ORC / Avro all emit long / double / boolean / string directly rather than stringifying.

Format	Extension	MIME	Notes
Parquet	.parquet	application/vnd.apache.parquet	Snappy-compressed single file. Single-file output; no partitioning.
Apache Arrow IPC / Feather	.arrow .feather	application/vnd.apache.arrow.file	One record batch. Requires runtime --add-opens for java.nio.
Apache ORC	.orc	application/vnd.apache.orc	VectorizedRowBatch writer.
Avro container	.avro	application/avro	In-memory DataFileWriter; no temp file.

Document-shaped exports. JSON / JSONL / YAML emit typed scalar values; XML and HTML stringify into element text; PDF renders a paginated visual artifact (not round-trippable as data).

Format	Extension	MIME	Notes
JSON	.json	application/json	Top-level array of objects.
JSON Lines / NDJSON	.jsonl	application/x-ndjson	One JSON object per line.
YAML	.yaml	application/x-yaml	Sequence of mappings.
XML	.xml	application/xml	<rows><row><col>…</col></row></rows>.
HTML	.html	text/html	Single <table> document.
PDF	.pdf	application/pdf	Paginated landscape render. Visual artifact — not round-trippable.

Single-table database files. The exported table lands as one user table inside the file; queries are immediately runnable with sqlite3 or the DuckDB CLI.

Format	Extension	MIME	Notes
SQLite	.sqlite	application/vnd.sqlite3	Column types: INTEGER / REAL / BOOLEAN / TEXT.
DuckDB	.duckdb	application/vnd.duckdb	Column types: BIGINT / DOUBLE / BOOLEAN / VARCHAR.

Hierarchical scientific-data formats. HDF5 writes a rank-2 /data dataset; NetCDF-4 is HDF5 with the _NCProperties attribute (NetCDF-4 IS HDF5 on disk). Classic CDF-1 / CDF-2 not produced.

Format	Extension	MIME	Notes
HDF5	.h5	application/x-hdf5	Rank-2 dataset at /data. Row / column name datasets when set.
NetCDF-4	.nc4	application/x-netcdf	HDF5-delegated; NetCDF-4 readers (Python netCDF4, NCO) read it natively.

Hand-rolled writers mirroring the import-side readers. Round-trip-tested via the matching reader.

Format	Extension	MIME	Notes
SAS	.sas7bdat	application/x-sas-data	Hand-rolled sas7bdat writer; round-trips via the parso reader.
SPSS	.sav	application/x-spss-sav	Uncompressed $FL2; short-string columns only (≤8 bytes).
Stata	.dta	application/x-stata-dta	Release 118 (Stata 14+). Numeric→double; strings→str244.
R serialized data	.rds	application/x-rds	Format version 2 XDR, gzip envelope. data.frame with logical / integer / double / character columns.

Layer on top of any format. Order is format → archive → compression so ?archive=tar&compression=gzip produces a real gzipped tarball (.tar.gz), not a zip-of-gz.

Wrapper	Values	Notes
Compression	gzip · bzip2 · xz · zstd	Same four algorithms as on the import side. Adds .gz / .bz2 / .xz / .zst to the filename.
Archive	zip · tar	Single-entry archive. zip carries its own per-entry deflate so ?compression= is ignored when ?archive=zip.

Optional half-open row / column ranges: ?fromRow=&toRow=&fromColumn=&toColumn=. Out-of-range values clamp to the actual table extent; inverted ranges return 400.

CSV / TSV exports accept dialect overrides: ?delimiter=&quote=&nullValue=&quoteAll=. delimiter and quote are single-character; quoteAll is true by default.