Extra the letters of a PDF document and their positions

pdfjs_letter_position_extraction.js

/*
When loading the PDF document, you need to set the `fontExtraProperties` parameter to `true`. e.g.
getDocument({
  ...,
  fontExtraProperties: true
})
*/

var generateId = () => {
    return (
        Date.now().toString(36) +
        Math.floor(
            Math.pow(10, 12) + Math.random() * 9 * Math.pow(10, 12)
        ).toString(36)
    )
}

function itemToRect(item, viewport, heightScale=1) {
	const transformed = Util.transform(
		viewport.transform,
		item.transform
	)

	let [left, top] = Util.applyTransform(
		[0, 1],
		transformed
	)

	let width = Util.applyTransform(
		[item.width, 0],
		viewport.transform
	)[0]

	let height = Util.applyTransform(
		[item.height, 0],
		viewport.transform
	)[0]

	left /= viewport.width
	width /= viewport.width

	top /= viewport.height
	height /= viewport.height

	height *= heightScale

	return { left, top, width, height }
}

function* itemToRects(item, page, heightScale) {
	const { viewport } = viewer.getPageView(page.pageNumber - 1)

	const bbox = {
		...itemToRect(item, viewport, heightScale),
		page: page.pageNumber
	}

	const itemFont = page.commonObjs.get(item.fontName)

	const letterItems = []
	let itemFontWidth = 0

	let itemText = item.str

	if (item.hasEOL) {
		itemText += '\n'
	}

	for (const letter of itemText) {
		const fontWidth = itemFont.widths[
			letter.charCodeAt()
		] || 0

		letterItems.push({
			letter,
			fontWidth,
			leftOffset: itemFontWidth
		})

		itemFontWidth += fontWidth
	}

	for (const letterItem of letterItems) {
		const { letter, fontWidth, leftOffset } = letterItem

		let left = 0
		let width = 0

		if (itemFontWidth > 0) {
			left = leftOffset / itemFontWidth * bbox.width + bbox.left
			width = fontWidth / itemFontWidth * bbox.width
		}

		yield {
			letter,
			left,
			width,
			page: bbox.page,
			top: bbox.top,
			height: bbox.height,
		}
	}
}

async function getLetterRects(page) {
	const { items } = await page.getTextContent()
	const formattedLetterRects = []

	for (const item of items) {
		const letterRects = itemToRects(item, page, heightScale=1.25)

		for (const letterRect of letterRects) {
			const { letter, ...rect } = letterRect

			formattedLetterRects.push({
				id: generateId(),
				text: letter,
				position: {
					page: page.pageNumber,
					rects: [rect]
				}
			})
		}
	}

	return formattedLetterRects
}