office-text-extractor-browser/source/parsers/pdf.ts

Fork of office-text-extractor with unreleased changes that include browser support

// source/parsers/pdf.ts
// The text extracter for PDF files.

import { type Buffer } from 'buffer/'
// @ts-expect-error There are no types for this package.
import parsePdf from 'pdf-parse/lib/pdf-parse.js'

import type { TextExtractionMethod } from '../lib.js'

export class PdfExtractor implements TextExtractionMethod {
	/**
	 * The type(s) of input acceptable to this method.
	 */
	mimes = ['application/pdf']

	/**
	 * Extract text from a PDF file if possible.
	 *
	 * @param payload The input and its type.
	 * @returns The text extracted from the input.
	 */
	apply = async (input: Buffer): Promise<string> => {
		// Convert the PDF to text and return the text.
		const parsedPdf = (await parsePdf(input, {
			pagerender: renderPage,
		})) as { text: string }
		return parsedPdf.text
	}
}

/**
 * We have to redefine this function to ensure that there are spaces between
 * words in the output text.
 *
 * @param data The data stored in the PDF about the page.
 * @returns The text content on the page
 */
const renderPage = async (data: unknown): Promise<string> => {
	const options = {
		normalizeWhitespace: false,
		disableCombineTextItems: false,
	}

	// @ts-expect-error todo: figure out the types
	return data.getTextContent(options).then((textContent: unknown) => {
		let lastY = ''
		let text = ''

		// @ts-expect-error todo: figure out the types
		for (const item of textContent.items) {
			if (!(lastY === item.transform[5] || !lastY)) text += '\n'
			// The word + a space
			text += (item.str as string) + ' '

			lastY = item.transform[5] as string
		}

		return text
	})
}