`extract`

Extracts tables or text from PDF/DOC/DOCX/XLS/XLSX files and outputs CSV, JSON, NDJSON, Parquet, or a Frictionless Data Package. PDF extraction supports table, text, or OCR modes.

pip install "undatum[extract]"

# PDF tables to CSV
undatum extract report.pdf --format-out csv --output report.csv

# Extract tables from multiple files
undatum extract data/*.pdf --format-out parquet --output-dir out/

# PDF text extraction for specific pages
undatum extract report.pdf --method text --pages 1-3 --format-out ndjson --output report.ndjson

Optional extra: pip install "undatum[extract]" installs:

pdfplumber — PDF tables/text
pdf2image + pytesseract — OCR (--method ocr; system Tesseract still required)
textract — legacy .doc

Reference

Reads: PDF, DOCX, DOC, HTML and images (OCR) · Writes: CSV, JSON, JSON Lines, Parquet or a data package · Memory: one document at a time · Engines: python

undatum extract [OPTIONS] INPUT_FILES...

Argument	Description
`INPUT_FILES...`	Input file(s) to extract from. (required)

Option	Description	Default
`-O`, `--format-out` TEXT	Output format: csv, json, ndjson, parquet, datapackage.	`csv`
`-o`, `--output` TEXT	Output file path (single table only).
`--output-dir` TEXT	Output directory for multiple tables.
`--method` TEXT	Extraction method: tables, text, ocr.
`--pages` TEXT	PDF pages to extract (e.g., 1-3,7,10-12).
`--ocr` / `--no-ocr`	Enable OCR for scanned PDFs.	`--no-ocr`
`--flatten` / `--no-flatten`	Flatten multiple tables into one output table.	`--no-flatten`
`--verbose` / `--no-verbose`	Enable verbose logging output.	`--no-verbose`

Deprecated spellings (removed in 2.0): --output-format → --format-out.

Reference​

Reference