extract-text

PDF to text

Take the words out of a PDF, in the order they are read.

Reads: a PDF — application/pdf
Writes: one file — text files or JSON
Tool page: /pdf-to-text

Example

Read the words off a document so a script can search them.

Given

report.pdf — a PDF of written text

From code

            import { loadDocument, runOperation } from '@pagecraft/core';

const report = await loadDocument({ bytes, name: 'report.pdf' });

const [output] = await runOperation('extract-text', [report], {
  format: 'plain',
});

From a shell

            pagecraft extract-text report.pdf --format plain

What comes back

One text file, containing “Quarterly report”.

This example is run by the test suite. If it stopped being true, the build would fail.

Options

`pages--pages`

string, a page selection: '1-3, 7, 9-', 'even', 'last'

Pages

Which pages to take the text from.

Default: 'all'

`format--format`

string, one of

Default: 'plain'

plain — The words, page by page.
json — Every run of text with the rectangle it occupies, for a tool to work with.

`dehyphenate--dehyphenate`

boolean

Rejoin split words

Put a word back together when the line it was set on broke it in half.

Default: true

What it cannot do

A scanned page holds a picture of words, not words. Run OCR over it first, and this will read the layer OCR adds.
Reading order is worked out from where the text sits on the page. It is right for columns, headings and tables, and it can be wrong on a page laid out like a poster.
Columns are recognised from three lines or more of them. Two short columns side by side are read across, the way a table is.
Text drawn sideways — a rotated watermark, a turned table header — is kept whole and read after the upright text, rather than being folded into a line it crosses.