PDF_TO_IMAGES - Realtime Compute for Apache Flink - Alibaba Cloud Documentation Center

Splits a PDF file into per-page images and returns each page as a row.

Syntax

PDF_TO_IMAGES(content [, image_format] [, dpi] [, start_page] [, pages])

Parameters

Parameter	Type	Required	Description
`content`	VARBINARY	Yes	The binary content of the PDF file. Use `FETCH_CONTENT` to retrieve content from a remote file.
`image_format`	STRING	No	The output image format. Supported values: `'jpg'`, `'png'`. Default: `'jpg'`.
`dpi`	INT	No	The rendering resolution in dots per inch (DPI), which controls image sharpness. Default: `200`.
`start_page`	INT	No	The first page to process. Page numbers are 0-indexed. Default: `0`.
`pages`	INT	No	The number of pages to process, starting from `start_page`. Must be used with `start_page` and cannot be used alone. The function processes pages in the range `[start_page, start_page + pages)`. Default: all pages from `start_page` to the end of the document.

Return parameters

The function returns one row per page, with the following columns:

Parameter	Type	Description
`mime_type`	STRING	The MIME type of the output image, such as `image/jpeg`.
`page_no`	INT	The PDF page number, 0-indexed.
`image_content`	VARBINARY	The binary content of the page image.

Example

The following query fetches a PDF from a URL and converts each page to a JPEG image at 150 DPI. The LATERAL TABLE syntax calls PDF_TO_IMAGES as a table-valued function and joins its output rows with the input.

SELECT
    p.mime_type AS mime_type,
    p.page_no AS page_no
FROM (
    SELECT FETCH_CONTENT(pdf_url) AS pdf_content
    FROM (
        VALUES ('https://example.com/sample.pdf')
    ) T (pdf_url)
) AS t1,
LATERAL TABLE(PDF_TO_IMAGES(t1.pdf_content, 'jpg', 150)) AS p(mime_type, page_no, image_content);

Sample output:

mime_type(STRING)	page_no(INT)
image/jpeg	0
image/jpeg	1
image/jpeg	2
image/jpeg	3