wiki: Let search function return a list of pages instead of a generator

[ipfire.org.git] / src / backend / wiki.py

#!/usr/bin/python3

import PIL
import PIL.ImageFilter
import difflib
import io
import logging
import os.path
import re
import tornado.gen
import urllib.parse

from . import misc
from . import util
from .decorators import *

class Wiki(misc.Object):
	def _get_pages(self, query, *args):
		res = self.db.query(query, *args)

		for row in res:
			yield Page(self.backend, row.id, data=row)

	def _get_page(self, query, *args):
		res = self.db.get(query, *args)

		if res:
			return Page(self.backend, res.id, data=res)

	def get_page_title(self, page, default=None):
		doc = self.get_page(page)
		if doc:
			return doc.title

		return default or os.path.basename(page)

	def get_page(self, page, revision=None):
		page = Page.sanitise_page_name(page)
		assert page

		if revision:
			return self._get_page("SELECT * FROM wiki WHERE page = %s \
				AND timestamp = %s", page, revision)
		else:
			return self._get_page("SELECT * FROM wiki WHERE page = %s \
				ORDER BY timestamp DESC LIMIT 1", page)

	def get_recent_changes(self, account, limit=None):
		pages = self._get_pages("SELECT * FROM wiki \
			WHERE timestamp >= NOW() - INTERVAL '4 weeks' \
			ORDER BY timestamp DESC")

		for page in pages:
			if not page.check_acl(account):
				continue

			yield page

			limit -= 1
			if not limit:
				break

	def create_page(self, page, author, content, changes=None, address=None):
		page = Page.sanitise_page_name(page)

		# Write page to the database
		page = self._get_page("INSERT INTO wiki(page, author_uid, markdown, changes, address) \
			VALUES(%s, %s, %s, %s, %s) RETURNING *", page, author.uid, content or None, changes, address)

		# Send email to all watchers
		page._send_watcher_emails(excludes=[author])

		return page

	def delete_page(self, page, author, **kwargs):
		# Do nothing if the page does not exist
		if not self.get_page(page):
			return

		# Just creates a blank last version of the page
		self.create_page(page, author=author, content=None, **kwargs)

	def make_breadcrumbs(self, url):
		# Split and strip all empty elements (double slashes)
		parts = list(e for e in url.split("/") if e)

		ret = []
		for part in ("/".join(parts[:i]) for i in range(1, len(parts))):
			ret.append(("/%s" % part, self.get_page_title(part, os.path.basename(part))))

		return ret

	def search(self, query, account=None, limit=None):
		query = util.parse_search_query(query)

		res = self._get_pages("SELECT wiki.* FROM wiki_search_index search_index \
			LEFT JOIN wiki ON search_index.wiki_id = wiki.id \
			WHERE search_index.document @@ to_tsquery('english', %s) \
				ORDER BY ts_rank(search_index.document, to_tsquery('english', %s)) DESC",
			query, query)

		pages = []
		for page in res:
			# Skip any pages the user doesn't have permission for
			if not page.check_acl(account):
				continue

			# Return any other pages
			pages.append(page)

			# Break when we have found enough pages
			if limit and len(pages) >= limit:
				break

		return pages

	def refresh(self):
		"""
			Needs to be called after a page has been changed
		"""
		self.db.execute("REFRESH MATERIALIZED VIEW wiki_search_index")

	# ACL

	def check_acl(self, page, account):
		res = self.db.query("SELECT * FROM wiki_acls \
			WHERE %s ILIKE (path || '%%') ORDER BY LENGTH(path) DESC LIMIT 1", page)

		for row in res:
			# Access not permitted when user is not logged in
			if not account:
				return False

			# If user is in a matching group, we grant permission
			for group in row.groups:
				if group in account.groups:
					return True

			# Otherwise access is not permitted
			return False

		# If no ACLs are found, we permit access
		return True

	# Files

	def _get_files(self, query, *args):
		res = self.db.query(query, *args)

		for row in res:
			yield File(self.backend, row.id, data=row)

	def _get_file(self, query, *args):
		res = self.db.get(query, *args)

		if res:
			return File(self.backend, res.id, data=res)

	def get_files(self, path):
		files = self._get_files("SELECT * FROM wiki_files \
			WHERE path = %s AND deleted_at IS NULL ORDER BY filename", path)

		return list(files)

	def get_file_by_path(self, path):
		path, filename = os.path.dirname(path), os.path.basename(path)

		return self._get_file("SELECT * FROM wiki_files \
			WHERE path = %s AND filename = %s AND deleted_at IS NULL", path, filename)

	def upload(self, path, filename, data, mimetype, author, address):
		# Upload the blob first
		blob = self.db.get("INSERT INTO wiki_blobs(data) VALUES(%s) RETURNING id", data)

		# Create entry for file
		return self._get_file("INSERT INTO wiki_files(path, filename, author_uid, address, \
			mimetype, blob_id, size) VALUES(%s,  %s, %s, %s, %s, %s, %s) RETURNING *", path,
			filename, author.uid, address, mimetype, blob.id, len(data))

	def find_image(self, path, filename):
		for p in (path, os.path.dirname(path)):
			file = self.get_file_by_path(os.path.join(p, filename))

			if file and file.is_image():
				return file


class Page(misc.Object):
	def init(self, id, data=None):
		self.id = id
		self.data = data

	def __repr__(self):
		return "<%s %s %s>" % (self.__class__.__name__, self.page, self.timestamp)

	def __eq__(self, other):
		if isinstance(other, self.__class__):
			return self.id == other.id

	def __lt__(self, other):
		if isinstance(other, self.__class__):
			if self.page == other.page:
				return self.timestamp < other.timestamp

			return self.page < other.page

	@staticmethod
	def sanitise_page_name(page):
		if not page:
			return "/"

		# Make sure that the page name does NOT end with a /
		if page.endswith("/"):
			page = page[:-1]

		# Make sure the page name starts with a /
		if not page.startswith("/"):
			page = "/%s" % page

		# Remove any double slashes
		page = page.replace("//", "/")

		return page

	@property
	def url(self):
		return self.page

	@property
	def full_url(self):
		return "https://wiki.ipfire.org%s" % self.url

	@property
	def page(self):
		return self.data.page

	@property
	def title(self):
		return self._title or os.path.basename(self.page[1:])

	@property
	def _title(self):
		if not self.markdown:
			return

		# Find first H1 headline in markdown
		markdown = self.markdown.splitlines()

		m = re.match(r"^# (.*)( #)?$", markdown[0])
		if m:
			return m.group(1)

	@lazy_property
	def author(self):
		if self.data.author_uid:
			return self.backend.accounts.get_by_uid(self.data.author_uid)

	def _render(self, text):
		logging.debug("Rendering %s" % self)

		# Link images
		replacements = []
		for match in re.finditer(r"!\[(.*?)\]\((.*?)\)", text):
			alt_text, url = match.groups()

			# Skip any absolute and external URLs
			if url.startswith("/") or url.startswith("https://") or url.startswith("http://"):
				continue

			# Try to split query string
			url, delimiter, qs = url.partition("?")

			# Parse query arguments
			args = urllib.parse.parse_qs(qs)

			# Find image
			file = self.backend.wiki.find_image(self.page, url)
			if not file:
				continue

			# Scale down the image if not already done
			if not "s" in args:
				args["s"] = "768"

			# Format URL
			url = "%s?%s" % (file.url, urllib.parse.urlencode(args))

			replacements.append((match.span(), file, alt_text, url))

		# Apply all replacements
		for (start, end), file, alt_text, url in reversed(replacements):
			text = text[:start] + "[![%s](%s)](%s?action=detail)" % (alt_text, url, file.url) + text[end:]

		# Add wiki links
		patterns = (
			(r"\[\[([\w\d\/\-\.]+)(?:\|(.+?))\]\]", r"\1", r"\2", None, True),
			(r"\[\[([\w\d\/\-\.]+)\]\]", r"\1", r"\1", self.backend.wiki.get_page_title, True),

			# External links
			(r"\[\[((?:ftp|git|https?|rsync|sftp|ssh|webcal)\:\/\/.+?)(?:\|(.+?))\]\]",
				r"\1", r"\2", None, False),

			# Mail
			(r"\[\[([a-zA-Z0-9_.+-]+@[a-zA-Z0-9-]+\.[a-zA-Z0-9-.]+)\]\]",
				r"\1", r"\1", None, False),
		)

		for pattern, link, title, repl, internal in patterns:
			replacements = []

			for match in re.finditer(pattern, text):
				l = match.expand(link)
				t = match.expand(title)

				if internal:
					# Allow relative links
					if not l.startswith("/"):
						l = os.path.join(self.page, l)

					# Normalise links
					l = os.path.normpath(l)

				if callable(repl):
					t = repl(l) or t

				replacements.append((match.span(), t or l, l))

			# Apply all replacements
			for (start, end), t, l in reversed(replacements):
				text = text[:start] + "[%s](%s)" % (t, l) + text[end:]

		# Borrow this from the blog
		return self.backend.blog._render_text(text, lang="markdown")

	@property
	def markdown(self):
		return self.data.markdown or ""

	@property
	def html(self):
		return self._render(self.markdown)

	@property
	def timestamp(self):
		return self.data.timestamp

	def was_deleted(self):
		return self.markdown is None

	@lazy_property
	def breadcrumbs(self):
		return self.backend.wiki.make_breadcrumbs(self.page)

	def get_latest_revision(self):
		revisions = self.get_revisions()

		# Return first object
		for rev in revisions:
			return rev

	def get_revisions(self):
		return self.backend.wiki._get_pages("SELECT * FROM wiki \
			WHERE page = %s ORDER BY timestamp DESC", self.page)

	@lazy_property
	def previous_revision(self):
		return self.backend.wiki._get_page("SELECT * FROM wiki \
			WHERE page = %s AND timestamp < %s ORDER BY timestamp DESC \
			LIMIT 1", self.page, self.timestamp)

	@property
	def changes(self):
		return self.data.changes

	# ACL

	def check_acl(self, account):
		return self.backend.wiki.check_acl(self.page, account)

	# Sidebar

	@lazy_property
	def sidebar(self):
		parts = self.page.split("/")

		while parts:
			sidebar = self.backend.wiki.get_page("%s/sidebar" % os.path.join(*parts))
			if sidebar:
				return sidebar

			parts.pop()

	# Watchers

	@lazy_property
	def diff(self):
		if self.previous_revision:
			diff = difflib.unified_diff(
				self.previous_revision.markdown.splitlines(),
				self.markdown.splitlines(),
			)

			return "\n".join(diff)

	@property
	def watchers(self):
		res = self.db.query("SELECT uid FROM wiki_watchlist \
			WHERE page = %s", self.page)

		for row in res:
			# Search for account by UID and skip if none was found
			account = self.backend.accounts.get_by_uid(row.uid)
			if not account:
				continue

			# Return the account
			yield account

	def is_watched_by(self, account):
		res = self.db.get("SELECT 1 FROM wiki_watchlist \
			WHERE page = %s AND uid = %s", self.page, account.uid)

		if res:
			return True

		return False

	def add_watcher(self, account):
		if self.is_watched_by(account):
			return

		self.db.execute("INSERT INTO wiki_watchlist(page, uid) \
			VALUES(%s, %s)", self.page, account.uid)

	def remove_watcher(self, account):
		self.db.execute("DELETE FROM wiki_watchlist \
			WHERE page = %s AND uid = %s", self.page, account.uid)

	def _send_watcher_emails(self, excludes=[]):
		# Nothing to do if there was no previous revision
		if not self.previous_revision:
			return

		for watcher in self.watchers:
			# Skip everyone who is excluded
			if watcher in excludes:
				logging.debug("Excluding %s" % watcher)
				continue

			logging.debug("Sending watcher email to %s" % watcher)

			# Compose message
			self.backend.messages.send_template("wiki/messages/page-changed",
				recipients=[watcher], page=self, priority=-10)


class File(misc.Object):
	def init(self, id, data):
		self.id   = id
		self.data = data

	@property
	def url(self):
		return os.path.join(self.path, self.filename)

	@property
	def path(self):
		return self.data.path

	@property
	def filename(self):
		return self.data.filename

	@property
	def mimetype(self):
		return self.data.mimetype

	@property
	def size(self):
		return self.data.size

	@lazy_property
	def author(self):
		if self.data.author_uid:
			return self.backend.accounts.get_by_uid(self.data.author_uid)

	@property
	def created_at(self):
		return self.data.created_at

	def is_pdf(self):
		return self.mimetype in ("application/pdf", "application/x-pdf")

	def is_image(self):
		return self.mimetype.startswith("image/")

	@lazy_property
	def blob(self):
		res = self.db.get("SELECT data FROM wiki_blobs \
			WHERE id = %s", self.data.blob_id)

		if res:
			return bytes(res.data)

	def get_thumbnail(self, size):
		cache_key = "-".join((self.path, util.normalize(self.filename), self.created_at.isoformat(), "%spx" % size))

		# Try to fetch the data from the cache
		thumbnail = self.memcache.get(cache_key)
		if thumbnail:
			return thumbnail

		# Generate the thumbnail
		thumbnail = self._generate_thumbnail(size)

		# Put it into the cache for forever
		self.memcache.set(cache_key, thumbnail)

		return thumbnail

	def _generate_thumbnail(self, size, **args):
		image = PIL.Image.open(io.BytesIO(self.blob))

		# Remove any alpha-channels
		if image.format == "JPEG" and not image.mode == "RGB":
			# Make a white background
			background = PIL.Image.new("RGBA", image.size, (255,255,255))

			# Flatten both images together
			flattened_image = PIL.Image.alpha_composite(background, image)

			# Remove the alpha channel
			image = flattened_image.convert("RGB")

		# Resize the image to the desired resolution
		image.thumbnail((size, size), PIL.Image.LANCZOS)

		if image.format == "JPEG":
			# Apply a gaussian blur to make compression easier
			image = image.filter(PIL.ImageFilter.GaussianBlur(radius=0.05))

			# Arguments to optimise the compression
			args.update({
				"subsampling" : "4:2:0",
				"quality"     : 70,
			})

		with io.BytesIO() as f:
			# If writing out the image does not work with optimization,
			# we try to write it out without any optimization.
			try:
				image.save(f, image.format, optimize=True, **args)
			except:
				image.save(f, image.format, **args)

			return f.getvalue()