search: Use PostgreSQL's websearch_to_tsquery()

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

#!/usr/bin/python3

import difflib
import logging
import os.path
import re
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 make_path(self, page, path):
		# Nothing to do for absolute links
		if path.startswith("/"):
			pass

		# Relative links (one-level down)
		elif path.startswith("./"):
			path = os.path.join(page, path)

		# All other relative links
		else:
			p = os.path.dirname(page)
			path = os.path.join(p, path)

		# Normalise links
		return os.path.normpath(path)

	def page_exists(self, path):
		page = self.get_page(path)

		# Page must have been found and not deleted
		return page and not page.was_deleted()

	def get_page_title(self, page, default=None):
		# Try to retrieve title from cache
		title = self.memcache.get("wiki:title:%s" % page)
		if title:
			return title

		# If the title has not been in the cache, we will
		# have to look it up
		doc = self.get_page(page)
		if doc:
			title = doc.title
		else:
			title = os.path.basename(page)

		# Save in cache for forever
		self.memcache.set("wiki:title:%s" % page, title)

		return title

	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 \
			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)

		# Update cache
		self.memcache.set("wiki:title:%s" % page.page, page.title)

		# 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):
		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 @@ websearch_to_tsquery('english', %s) \
				ORDER BY ts_rank(search_index.document, websearch_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")

	def get_watchlist(self, account):
		pages = self._get_pages(
			"WITH pages AS (SELECT * FROM wiki_current \
					LEFT JOIN wiki ON wiki_current.id = wiki.id) \
			SELECT * FROM wiki_watchlist watchlist \
				LEFT JOIN pages ON watchlist.page = pages.page \
				WHERE watchlist.uid = %s",
			account.uid,
		)

		return sorted(pages)

	# 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 account.is_member_of_group(group):
					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, revision=None):
		path, filename = os.path.dirname(path), os.path.basename(path)

		if revision:
			# Fetch a specific revision
			return self._get_file("SELECT * FROM wiki_files \
				WHERE path = %s AND filename = %s AND created_at <= %s \
				ORDER BY created_at DESC LIMIT 1", path, filename, revision)

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

	def get_file_by_path_and_filename(self, path, filename):
		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):
		# Replace any existing files
		file = self.get_file_by_path_and_filename(path, filename)
		if file:
			file.delete(author)

		# Upload the blob first
		blob = self.db.get("INSERT INTO wiki_blobs(data) VALUES(%s) \
			ON CONFLICT (digest(data, %s)) DO UPDATE SET data = EXCLUDED.data \
			RETURNING id", data, "MD5")

		# 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 render(self, path, text):
		r = WikiRenderer(self.backend, path)

		return r.render(text)


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"^#\s*(.*)( #)?$", 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)

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

	@property
	def html(self):
		return self.backend.wiki.render(self.page, self.markdown)

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

	def was_deleted(self):
		return not self.markdown

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

	def is_latest_revision(self):
		return self.get_latest_revision() == self

	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

			# Check permissions
			if not self.backend.wiki.check_acl(self.page, watcher):
				logging.debug("Watcher %s does not have permissions" % watcher)
				continue

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

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

	def restore(self, author, address):
		changes = "Restore to revision from %s" % self.timestamp.isoformat()

		return self.backend.wiki.create_page(self.page,
			author, self.markdown, changes=changes, address=address)


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

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

	@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 delete(self, author=None):
		self.db.execute("UPDATE wiki_files SET deleted_at = NOW(), deleted_by = %s \
			WHERE id = %s", author.uid if author else None, self.id)

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

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

		# Return first object
		for rev in revisions:
			return rev

	def get_revisions(self):
		revisions = self.backend.wiki._get_files("SELECT * FROM wiki_files \
			WHERE path = %s ORDER BY created_at DESC", self.path)

		return list(revisions)

	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 = util.generate_thumbnail(self.blob, size)

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

		return thumbnail


class WikiRenderer(misc.Object):
	schemas = (
		"ftp://",
		"git://",
		"http://",
		"https://",
		"rsync://",
		"sftp://",
		"ssh://",
		"webcal://",
	)

	# Links
	links = re.compile(r"<a href=\"(.*?)\">(.*?)</a>")

	# Images
	images = re.compile(r"<img alt(?:=\"(.*?)\")? src=\"(.*?)\" (?:title=\"(.*?)\" )?/>")

	def init(self, path):
		self.path = path

	def _render_link(self, m):
		url, text = m.groups()

		# Emails
		if "@" in url:
			# Strip mailto:
			if url.startswith("mailto:"):
				url = url[7:]

			return """<a class="link-external" href="mailto:%s">%s</a>""" % \
				(url, text or url)

		# External Links
		for schema in self.schemas:
			if url.startswith(schema):
				return """<a class="link-external" href="%s">%s</a>""" % \
					(url, text or url)

		# Everything else must be an internal link
		path = self.backend.wiki.make_path(self.path, url)

		return """<a href="%s">%s</a>""" % \
			(path, text or self.backend.wiki.get_page_title(path))

	def _render_image(self, m):
		alt_text, url, caption = m.groups()

		# Skip any absolute and external URLs
		if url.startswith("/") or url.startswith("https://") or url.startswith("http://"):
			return """<figure class="figure"><img src="%s" class="figure-img img-fluid rounded" alt="%s">
				<figcaption class="figure-caption">%s</figcaption></figure>
			""" % (url, alt_text, caption or "")

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

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

		# Build absolute path
		url = self.backend.wiki.make_path(self.path, url)

		# Find image
		file = self.backend.wiki.get_file_by_path(url)
		if not file or not file.is_image():
			return "<!-- Could not find image %s in %s -->" % (url, self.path)

		# Scale down the image if not already done
		if not "s" in args:
			args["s"] = "920"

		return """<figure class="figure"><img src="%s?%s" class="figure-img img-fluid rounded" alt="%s">
		<figcaption class="figure-caption">%s</figcaption></figure>
		""" % (url, urllib.parse.urlencode(args), caption, caption or "")

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

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

		# Postprocess links
		text = self.links.sub(self._render_link, text)

		# Postprocess images to <figure>
		text = self.images.sub(self._render_image, text)

		return text