Document Read Blob
Description
Section titled “Description”The Document: Read Blob step reads whole documents into a base64 text column named Blob, with one row per document. Read one document you choose, or one document for each row of a table of paths. Later steps, such as Image OCR and Text NLP AI, read the content.
Source & Options
Section titled “Source & Options”- Table of document paths (optional) — set this to read one document for each row. Leave it blank to read the one document you choose.
- Path column — the column of that table that holds each document’s path.
- Document — the document account and path to read when there is no table.
- Fail when a document is missing — off by default. Off, a missing document is skipped with a warning that names its path, and a missing single document writes an empty table. On, the step fails.
- Largest document (bytes) (optional) — a larger document fails rather than being split.
- Output Table — where the rows are written.
Columns from the table of paths pass through to the output.
The More Options group sets these keys:
- Keep the row of a missing document, with an empty Blob (
keep_missing) — keep the row of a missing document, with an emptyBlob, instead of skipping it. - Read the document as an archive, one row per entry (
unpack_enabled) — with one chosen document and no table of paths, read the document as an archive (zip, jar, tar, tar.gz, tar.bz2, cpio, ar, or a singlegzorbz2file) and write one row per entry:Path(below the Entry Folder,unpack_destination),Directory,Status(ExtractedorFolder), andBlob(none for a folder). Entry Name Encoding (unpack_charset, defaultcp437) is the encoding of zip entry names that are not UTF-8. A missing document gives no rows.
To write blobs back to documents, use Document: Write Blob.