X-Git-Url: https://git.mdrn.pl/wolnelektury.git/blobdiff_plain/67df302167539474edd02b57c2e8bd30349d7625..d90edc4826d6c3d3a40f52fa5c88cf4b1741ab63:/apps/catalogue/utils.py?ds=inline diff --git a/apps/catalogue/utils.py b/apps/catalogue/utils.py index e44b3792f..a48ec0370 100644 --- a/apps/catalogue/utils.py +++ b/apps/catalogue/utils.py @@ -2,13 +2,27 @@ # This file is part of Wolnelektury, licensed under GNU Affero GPLv3 or later. # Copyright © Fundacja Nowoczesna Polska. See NOTICE for more information. # +from __future__ import with_statement + import random import time from base64 import urlsafe_b64encode +from django.http import HttpResponse, HttpResponseRedirect, Http404, HttpResponsePermanentRedirect +from django.core.files.uploadedfile import UploadedFile +from django.core.files.base import File +from django.core.files.storage import DefaultStorage from django.utils.hashcompat import sha_constructor from django.conf import settings +from celery.task import task +from os import mkdir, path, unlink +from errno import EEXIST, ENOENT +from fcntl import flock, LOCK_EX +from zipfile import ZipFile +from reporting.utils import read_chunks +from celery.task import task +import catalogue.models # Use the system (hardware-based) random number generator if it exists. if hasattr(random, 'SystemRandom'): @@ -31,3 +45,141 @@ def split_tags(tags): result.setdefault(tag.category, []).append(tag) return result + +class ExistingFile(UploadedFile): + + def __init__(self, path, *args, **kwargs): + self.path = path + return super(ExistingFile, self).__init__(*args, **kwargs) + + def temporary_file_path(self): + return self.path + + def close(self): + pass + + +class LockFile(object): + """ + A file lock monitor class; createas an ${objname}.lock + file in directory dir, and locks it exclusively. + To be used in 'with' construct. + """ + def __init__(self, dir, objname): + self.lockname = path.join(dir, objname + ".lock") + + def __enter__(self): + self.lock = open(self.lockname, 'w') + flock(self.lock, LOCK_EX) + + def __exit__(self, *err): + try: + unlink(self.lockname) + except OSError as oe: + if oe.errno != oe.EEXIST: + raise oe + self.lock.close() + + +@task +def create_zip(paths, zip_slug): + """ + Creates a zip in MEDIA_ROOT/zip directory containing files from path. + Resulting archive filename is ${zip_slug}.zip + Returns it's path relative to MEDIA_ROOT (no initial slash) + """ + # directory to store zip files + zip_path = path.join(settings.MEDIA_ROOT, 'zip') + + try: + mkdir(zip_path) + except OSError as oe: + if oe.errno != EEXIST: + raise oe + zip_filename = zip_slug + ".zip" + + with LockFile(zip_path, zip_slug): + if not path.exists(path.join(zip_path, zip_filename)): + zipf = ZipFile(path.join(zip_path, zip_filename), 'w') + try: + for arcname, p in paths: + if arcname is None: + arcname = path.basename(p) + zipf.write(p, arcname) + finally: + zipf.close() + + return 'zip/' + zip_filename + + +def remove_zip(zip_slug): + """ + removes the ${zip_slug}.zip file from zip store. + """ + zip_file = path.join(settings.MEDIA_ROOT, 'zip', zip_slug + '.zip') + try: + unlink(zip_file) + except OSError as oe: + if oe.errno != ENOENT: + raise oe + + +class AttachmentHttpResponse(HttpResponse): + """Response serving a file to be downloaded. + """ + def __init__ (self, file_path, file_name, mimetype): + super(AttachmentHttpResponse, self).__init__(mimetype=mimetype) + self['Content-Disposition'] = 'attachment; filename=%s' % file_name + self.file_path = file_path + self.file_name = file_name + + with open(DefaultStorage().path(self.file_path)) as f: + for chunk in read_chunks(f): + self.write(chunk) + +@task +def async_build_pdf(book_id, customizations, file_name): + """ + A celery task to generate pdf files. + Accepts the same args as Book.build_pdf, but with book id as first parameter + instead of Book instance + """ + book = catalogue.models.Book.objects.get(id=book_id) + print "will gen %s" % DefaultStorage().path(file_name) + if not DefaultStorage().exists(file_name): + book.build_pdf(customizations=customizations, file_name=file_name) + print "done." + + +class MultiQuerySet(object): + def __init__(self, *args, **kwargs): + self.querysets = args + self._count = None + + def count(self): + if not self._count: + self._count = sum(len(qs) for qs in self.querysets) + return self._count + + def __len__(self): + return self.count() + + def __getitem__(self, item): + try: + indices = (offset, stop, step) = item.indices(self.count()) + except AttributeError: + # it's not a slice - make it one + return self[item : item + 1][0] + items = [] + total_len = stop - offset + for qs in self.querysets: + if len(qs) < offset: + offset -= len(qs) + else: + items += list(qs[offset:stop]) + if len(items) >= total_len: + return items + else: + offset = 0 + stop = total_len - len(items) + continue \ No newline at end of file