let db = require('@config/database')

let SiteScrape = require('@helpers/SiteScrape')
const cs = require('@helpers/CryptoString')

let Attachment = module.exports = {}

const cheerio = require('cheerio')
const fs = require('fs')
const path = require('path')
const crypto = require('crypto')

const gm = require('gm')

const tesseract = require("node-tesseract-ocr")

// Left as-is deliberately: the upload / OCR / thumbnail code below resolves
// against it, and it is correct under pm2 (cwd = server/). Not touching it
// keeps the upload path byte-identical.
const filePath = '../staticFiles/'

// Absolute equivalent, used by the scrape path only. filePath is CWD-relative,
// so under Jest (cwd = repo root) scrape downloads landed in a sibling of the
// repo rather than in staticFiles/. This resolves to the same directory as
// filePath under pm2.
const scrapeFilePath = path.join(__dirname, '..', '..', 'staticFiles') + path.sep

const config = require('@config/scrape')
const UrlSafety = require('@helpers/UrlSafety')
const HttpFetch = require('@helpers/HttpFetch')
const Realtime = require('@helpers/Realtime')

// Attachment.migrateOld

Attachment.textSearch = (userId, searchTerm) => {
	return new Promise((resolve, reject) => {

		const front = 5
		const tail = 150

		const query = `
			SELECT 
				*,
				substring(
					text,
			        IF(LOCATE(?, text) > ${tail}, LOCATE(?, text) - ${front}, 1),
			        ${tail} + LENGTH(?) + ${front}
				) as snippet
			FROM attachment 
			WHERE user_id = ?
			AND visible != 0
			AND MATCH(text)
			AGAINST(? IN NATURAL LANGUAGE MODE)
			LIMIT 1000`

		db.promise()
			.query(query, [searchTerm, searchTerm, searchTerm, userId, searchTerm])
			.then((rows, fields) => {
				resolve(rows[0]) //Return all attachments found by query
			})
		.catch(console.log)
	})
}

Attachment.search = (userId, noteId, attachmentType, offset, setSize, includeShared) => {
	console.log([userId, noteId, attachmentType, offset, setSize, includeShared])
	return new Promise((resolve, reject) => {

		let params = [userId]
		let query = `
			SELECT attachment.*, note.share_user_id FROM attachment 
			LEFT JOIN note ON (attachment.note_id = note.id)
			WHERE attachment.user_id = ? AND visible = 1 
			`

		if(noteId && noteId > 0){
			//
			// Show everything if note ID is present
			//
			query += 'AND attachment.note_id = ? '
			params.push(noteId)

		} else {
			//
			// Other filters if NO note id
			//

			if(attachmentType == 'links'){
				query += 'AND attachment_type = 1 '
			}
			if(attachmentType == 'files'){
				query += 'AND attachment_type > 1 '
			}

			query += `AND note.archived = ${ attachmentType == 'archived' ? '1':'0' } `
			query += `AND note.trashed = ${ attachmentType == 'trashed' ? '1':'0' } `

			if(!attachmentType){
				// Null note ID means it was pushed by bookmarklet
				query += 'OR attachment.note_id IS NULL '
			}
		}
		

		if(!noteId){
			const sharedOrNot = includeShared ? ' NOT ':' ' 
			query += `AND note.share_user_id IS${sharedOrNot}NULL `
		}


		query += 'ORDER BY last_indexed DESC '

		const limitOffset = parseInt(offset, 10) || 0 //Either parse int, or use zero
		const parsedSetSize = parseInt(setSize, 10) || 20
		query += ` LIMIT ${limitOffset}, ${parsedSetSize}`

		console.log(query)

		db.promise()
			.query(query, params)
			.then((rows, fields) => {
				resolve(rows[0]) //Return all attachments found by query
			})
		.catch(console.log)
	})
}

Attachment.urlForNote = (userId, noteId) => {
	return new Promise((resolve, reject) => {
		db.promise()
			.query(`SELECT * FROM attachment WHERE user_id = ? AND note_id = ? AND attachment_type = 1 ORDER BY last_indexed DESC;`, [userId, noteId])
			.then((rows, fields) => {
				resolve(rows[0]) //Return all attachments found by query
			})
		.catch(console.log)
	})
}

//Update attachment in database
Attachment.update = (userId, attachmentId, updatedText, noteId) => {
	return new Promise((resolve, reject) => {
		db.promise()
			.query(`UPDATE attachment SET text = ? WHERE id = ? AND user_id = ?`, 
				[updatedText, attachmentId, userId])
			.then((rows, fields) => {
				resolve(true)
			})
		.catch(console.log)
	})
}

Attachment.delete = (userId, attachmentId, urlDelete = false) => {

	let attachment = null
	let noteExists = true

	return new Promise((resolve, reject) => {
		db.promise()
			.query('SELECT * FROM attachment WHERE id = ? AND user_id = ? LIMIT 1', [attachmentId, userId])
			.then((rows, fields) => {

				//Attachment doesn't exist, return done
				if(rows[0].length == 0){
					return resolve(true)
				}

				attachment = rows[0][0]

				return db.promise().query('SELECT count(id) as `exists` FROM note WHERE id = ?', [attachment.note_id])

			})

			.then((rows, fields) => {

				noteExists = (rows[0][0]['exists'] > 0)

				let url = attachment.url
				const noteId = attachment.note_id

				//Try to delete file and thumbnail
				try { 
					fs.unlinkSync(filePath+attachment.file_location) 
				} catch(err) { console.error('File Does not exist') }
				try { 
					fs.unlinkSync(filePath+'thumb_'+attachment.file_location)
				} catch(err) { console.error('Thumbnail Does not exist') }

				//Do not delete link attachments, just hide them. They will be deleted if removed from note or if note is deleted
				if(attachment.attachment_type == 1 && !urlDelete && noteExists){
					db.promise()
						.query(`UPDATE attachment SET visible = 0 WHERE id = ?`, [attachmentId])
						.then((rows, fields) => resolve(true))
						.catch(console.log)

					return resolve(true)
				} else {
					db.promise()
						.query(`DELETE FROM attachment WHERE id = ?`, [attachmentId])
						.then((rows, fields) => resolve(true))
						.catch(console.log)
				}
			})
			.catch(console.log)
	})
}

Attachment.processUploadedFile = (userId, noteId, fileObject) => {
	return new Promise((resolve, reject) => {

		const rawFilename = fileObject.filename
		const extension = '.'+fileObject.originalname.split('.').pop()
		const goodFileName = rawFilename+extension
		const fileName = fileObject.originalname //Actual name of the file, dog.jpg

		//Rename random file name to one with an extension
		fs.rename(filePath+rawFilename, filePath+goodFileName, (err) => {

			const created = Math.round((+new Date)/1000)
			
			db.promise()
			.query(`
				INSERT INTO attachment 
					(note_id, user_id, attachment_type, \`text\`, last_indexed, file_location) 
				VALUES 
					(?, ?, ?, ?, ?, ?)
			`, [noteId, userId, 2, 'Add a description to -> '+fileName, created, goodFileName])
			.then((rows, fields) => {

				Attachment.generateThumbnail(goodFileName)

				//If its an image, scrape text
				if(true){

					// https://github.com/tesseract-ocr/tesseract/wiki/ImproveQuality
					//psm 3 - default, 11 - as much text as possible
					const config = { lang: "eng", oem: 1, psm: 3 }

					tesseract.recognize(filePath+goodFileName, config)
					.then(text => {

						text = text.slice(0, -1).trim()

						if(text.length > 5){
							console.log('Inserting text')
							db.promise().query(
								`UPDATE attachment SET text = ? WHERE id = ? AND user_id = ? LIMIT 1`,
								[text, rows[0].insertId, userId]
							).then(results => {
								resolve({ fileName, goodFileName })
							})
						} else {
							return resolve({ fileName, goodFileName })
						}

					})
					.catch(error => {
						console.log(error.message)
					})

				} else {
					resolve({ fileName, goodFileName })
				}
			})
			.catch(console.log)

		})
	})
}

Attachment.generateThumbnail = (fileName) => {
	return new Promise((resolve, reject) => {
		gm(filePath+fileName)
		.resize(550) //Resize to width of 550 px 
		.quality(75) //compression level 0 - 100 (best)
		.write(filePath + 'thumb_'+fileName, function (err) {
			resolve(fileName)
		})
	})
}

//
// Diff the URLs currently in a note against the attachments already stored for
// it, soft-delete the ones that vanished, and enqueue the new ones.
//
// Returns immediately with the text already known from cache; freshly scraped
// text arrives asynchronously via the worker. Nothing here waits on the network.
//
Attachment.scanTextForWebsites = (userId, noteId, noteText) => {

	const ScrapeQueue = require('@workers/ScrapeQueue')
	const Schema = require('@helpers/Schema')

	// Fixed: the old version called resolve() without returning, so an empty
	// note fell through and ran the DB lookup anyway.
	if (!noteText || noteText.length == 0) { return Promise.resolve('') }

	// When the scrape schema is absent the new columns do not exist, so
	// creating a pending row would fail on every save. The diff and soft-delete
	// below still run - they touch no new columns - so removing a link from a
	// note behaves exactly as before.
	const scrapeAvailable = Schema.isScrapeSchemaReady()

	let knownText = ''

	return Attachment.urlForNote(userId, noteId)
		.then(attachments => {

			const foundUrls = SiteScrape.getCleanUrls(noteText)
			const remaining = foundUrls.slice()

			attachments.forEach(attachment => {

				// Compared on normalised keys rather than raw strings, so an
				// http/https or www. variant is not treated as a different
				// resource and needlessly deleted and re-scraped.
				const index = remaining.findIndex(url =>
					url === attachment.url || UrlSafety.sameResource(url, attachment.url))

				if (index !== -1) {
					knownText += attachment.text || ''
					remaining.splice(index, 1)
				} else {
					// Unchanged behaviour: link attachments are hidden, not deleted.
					Attachment.delete(userId, attachment.id, true)
				}
			})

			if (!remaining.length) { return knownText }

			// No schema, no previews. Return what is already cached rather than
			// attempting inserts that cannot succeed.
			if (!scrapeAvailable) { return knownText }

			// Create a shell row per new URL, then hand off to the worker.
			return Promise.all(remaining.map(url =>
				Attachment.createPendingUrlAttachment(userId, noteId, url)
					.then(attachmentId => ({ userId, noteId, attachmentId, url }))
					.catch(error => {
						console.log('[scrape] could not create attachment row', url, error.message)
						return null
					})
			))
			.then(jobs => {
				const valid = jobs.filter(Boolean)

				// Fire and forget - the note save must not wait on this.
				ScrapeQueue.enqueueUrls(valid)
					.catch(error => console.log('[scrape] enqueue failed', error.message))

				return knownText
			})
		})
		// The old version had no catch here at all.
		.catch(error => {
			console.log('[scrape] scanTextForWebsites failed', error.message)
			return knownText
		})
}

//
// Create the shell attachment row for a URL and return its id.
//
// Returns the id BEFORE any network work begins. The old code armed a 5s
// setTimeout calling request.cancel() synchronously, before this INSERT had
// resolved - so a slow INSERT meant the timer fired against a null reference
// and threw inside a bare timer callback, taking the process down.
//
Attachment.createPendingUrlAttachment = (userId, noteId, url) => {
	const created = Math.round((+new Date) / 1000)

	return db.promise()
		.query(
			`INSERT INTO attachment
				(note_id, user_id, attachment_type, text, url, last_indexed, file_location, scrape_state, enqueued_at)
			 VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?)`,
			[noteId, userId, 1, url, url, created, null, 0, created]
		)
		.then(([result]) => result.insertId)
}

//
// Scrape a set of URLs for a note and resolve with their combined text.
//
// Kept for the bookmarklet path. The old implementation resolved only when a
// completion counter reached foundUrls.length, so a single rejection meant the
// promise never settled and the caller hung forever. allSettled inside the
// worker removes the counter entirely.
//
Attachment.scrapeUrlsCreateAttachments = (userId, noteId, foundUrls) => {

	const ScrapeQueue = require('@workers/ScrapeQueue')
	const Schema = require('@helpers/Schema')

	if (!foundUrls || !foundUrls.length) { return Promise.resolve('') }
	if (!Schema.isScrapeSchemaReady()) { return Promise.resolve('') }

	return Promise.all(foundUrls.map(url =>
		Attachment.createPendingUrlAttachment(userId, noteId, url)
			.then(attachmentId => ({ userId, noteId, attachmentId, url }))
			.catch(error => {
				console.log('[scrape] could not create attachment row', url, error.message)
				return null
			})
	))
	.then(jobs => ScrapeQueue.enqueueUrls(jobs.filter(Boolean)))
	.then(texts => texts.join(''))
}

//
// Download and store a preview image.
//
// The old downloadFileFromUrl had no timeout, no size limit, no User-Agent and
// no status check: it piped whatever came back straight into a write stream, so
// a 404 HTML error page was written to disk and handed to gm. It also wrote the
// file before validating anything, which is why staticFiles/ accumulated
// zero-byte orphans.
//
Attachment.downloadThumbnail = (imageUrl, pageUrl) => {

	if (!imageUrl) { return Promise.resolve(null) }

	// crypto.randomBytes rather than two concatenated Math.random() calls.
	const fileName = crypto.randomBytes(16).toString('hex') + '_scrape'
	const fullPath = scrapeFilePath + fileName
	const thumbPath = scrapeFilePath + 'thumb_' + fileName
	const tempPath = fullPath + '.tmp'

	return HttpFetch.fetchBinary(imageUrl, {
		pageUrl,
		timeoutMs: config.imageTimeoutMs,
		maxBytes: config.maxImageBytes,
	})
	.then(result => {

		// Status and content-type were both gated inside fetchBinary, so by
		// here we are holding bytes that really are an image.
		if (!result.buffer || !result.buffer.length) { return null }
		if (!magicBytesMatch(result.buffer, result.mime)) {
			console.log('[scrape] image bytes do not match declared type', result.mime, imageUrl)
			return null
		}

		// Write to a temp path, then rename, so a failure cannot leave a
		// half-written file behind for the client to serve.
		return fs.promises.writeFile(tempPath, result.buffer)
			.then(() => fs.promises.rename(tempPath, fullPath))
			.then(() => resizeToThumbnail(fullPath, thumbPath))
			.then(() => fileName)
			.catch(error => {
				console.log('[scrape] could not store image', imageUrl, error.message)
				return cleanupFiles([tempPath, fullPath, thumbPath]).then(() => null)
			})
	})
	.catch(error => {
		console.log('[scrape] image fetch failed', imageUrl, error.kind || error.message)
		return null
	})
}

//
// Resize via gm. Degrades rather than failing: gm (and tesseract) are not in
// shell.nix buildInputs, so the binary is absent in dev - the old code returned
// null here and left the downloaded file orphaned on disk. Copying the
// full-size image to the thumbnail path keeps the preview working.
//
const resizeToThumbnail = (fullPath, thumbPath) => {
	return new Promise(resolve => {
		gm(fullPath)
			.resize(550)   //Resize to width of 550 px
			.quality(85)   //compression level 0 - 100 (best)
			.write(thumbPath, error => {
				if (!error) { return resolve(true) }

				console.log('[scrape] gm unavailable or failed, using full size as thumbnail')
				fs.promises.copyFile(fullPath, thumbPath)
					.then(() => resolve(true))
					.catch(() => resolve(false))
			})
	})
}

const cleanupFiles = (paths) => {
	return Promise.all(paths.map(target =>
		fs.promises.unlink(target).catch(() => null)
	))
}

//
// Sanity-check the bytes against the declared content type. A server can claim
// image/png and send anything.
//
const IMAGE_MAGIC = [
	{ mime: 'image/png', bytes: [0x89, 0x50, 0x4e, 0x47] },
	{ mime: 'image/jpeg', bytes: [0xff, 0xd8, 0xff] },
	{ mime: 'image/gif', bytes: [0x47, 0x49, 0x46, 0x38] },
	{ mime: 'image/bmp', bytes: [0x42, 0x4d] },
]

const magicBytesMatch = (buffer, mime) => {
	const expected = IMAGE_MAGIC.find(entry => entry.mime === mime)

	// webp/avif/ico and friends have container-dependent headers - accept them
	// on the content-type alone rather than guessing wrong.
	if (!expected) { return true }
	if (buffer.length < expected.bytes.length) { return false }

	return expected.bytes.every((byte, index) => buffer[index] === byte)
}

Attachment.generatePushKey = (userId) => {
	return new Promise((resolve, reject) => {

		db.promise()
		.query("SELECT pushkey FROM user WHERE id = ? LIMIT 1", [userId])
		.then((rows, fields) => {

			const pushKey = rows[0][0].pushkey
			
			// push key exists
			if(pushKey && pushKey.length > 0){

				return resolve(pushKey)

			} else {

				// generate and save a new key
				const newPushKey = cs.createSmallSalt()
				db.promise()
				.query('UPDATE user SET pushkey = ? WHERE id = ? LIMIT 1', [newPushKey,userId])
				.then((rows, fields) => {

					return resolve(newPushKey)
				})
			}
			
		})
	})
}

Attachment.deletePushKey = (userId) => {
	return new Promise((resolve, reject) => {

		db.promise()
		.query('UPDATE user SET pushkey = null WHERE id = ? LIMIT 1', [userId])
		.then((rows, fields) => {

			return resolve(rows[0].affectedRows == 1)
		})
	})
}

Attachment.getPushkeyBookmarklet = (userId) => {
	return new Promise((resolve, reject) => {

		Attachment.generatePushKey(userId)
		.then( pushKey => {

			let bookmarklet = Attachment.generateBookmarkletText(pushKey)
			return resolve(bookmarklet)

		})
	})
}

//
// Bookmarklet entry point: GET /api/public/pushmebaby?pushkey=...&url=...
//
// The URL arrives straight off a query string, so it is validated here - the
// old code passed it through completely unvalidated, bypassing even the https
// normalisation that note text goes through.
//
// Response timing is preserved: the bookmarklet still answers after the scrape,
// but now with a hard cap instead of the previous unbounded hang.
//
Attachment.pushUrl = (pushkey, url) => {

	const ScrapeQueue = require('@workers/ScrapeQueue')
	const Schema = require('@helpers/Schema')

	// Preserve the bookmarklet contract - resolve truthy, leaking nothing -
	// rather than failing an insert against columns that do not exist.
	if (!Schema.isScrapeSchemaReady()) { return Promise.resolve(true) }

	let userId = null
	pushkey = String(pushkey || '').replace(/ /g, '+')

	return db.promise()
		.query("SELECT id FROM user WHERE pushkey = ? LIMIT 1", [pushkey])
		.then(([rows]) => {

			// Unchanged: an unknown pushkey resolves truthy and leaks nothing
			// about whether the key exists.
			if (rows.length == 0) { return null }

			userId = rows[0].id

			return UrlSafety.assertFetchable(url)
				.then(() => Attachment.createPendingUrlAttachment(userId, null, url))
				.then(attachmentId => ScrapeQueue.awaitUrl(
					ScrapeQueue.enqueueUrl({ userId, noteId: null, attachmentId, url }),
					config.totalBudgetMs
				))
				.catch(error => {
					console.log('[scrape] pushUrl refused', url, error.reason || error.message)
					return null
				})
		})
		.then(() => {
			if (userId) {
				Realtime.emitToUser(userId, 'update_counts')
				Realtime.emitToUser(userId, 'update_note_attachments')
			}
			return true
		})
		.catch(error => {
			console.log('[scrape] pushUrl failed', error.message)
			return true
		})
}

Attachment.generateBookmarkletText = (pushKey) => {

	const endpoint = '/api/public/pushmebaby'
	let url = 'https://www.solidscribe.com' + endpoint
	if(process.env.NODE_ENV === 'development'){
		// url = 'https://192.168.1.164' + endpoint
	}

	// Terminate each line with a semi-colon, super important, since spaces are removed.
	
	// document.getElementById(id).remove();
	url += '?pushkey='+encodeURIComponent(pushKey)
	const bookmarkletV3 = `
		javascript: (() => {
			var p = encodeURIComponent(window.location.href);
			var n = "`+url+`&url="+p;
			window.open(n, '_blank', 'noopener=noopener');
			window.focus();

			var k = document.createElement("div");
			k.setAttribute("style", "position:fixed;right:10px;top:10px;z-index:222222;border-radius:4px;font-size:1.3em;padding:20px 15px;background: #8f51be;color:white;");
			k.innerHTML = "Posted URL to your Solid Scribe account";

			document.body.appendChild(k);

			setTimeout(()=>{
				k.remove();
			},5000);

		})();
	`

	return bookmarkletV3
		.replace(/\t|\r|\n/gm, "") // Remove tabs, new lines, returns
		.replace(/\s+/g, ' ') // remove double spaces
		.trim()
}