class Deduplicator(multiprocessing.Process):
def __init__(self, settings_json, stop_event, db_lock):
"""
Create a Hasher Process, which will be bound to the stop_event, performing post-processing on downloaded Files.
"""
super().__init__()
self._settings = settings_json
self._stop_event = stop_event
self._lock = db_lock
self.progress = DownloaderProgress()
self.progress.clear(status="Starting up...")
self._session = None
self.daemon = True
def run(self):
""" Threaded loading of elements. """
settings.from_json(self._settings)
sql.init_from_settings()
try:
self._session = sql.session()
self.progress.clear(status="Starting up...")
self.progress.set_running(True)
while not self._stop_event.is_set():
self._dedupe()
self.progress.set_status("Ready for new files...")
self._stop_event.wait(2)
self._dedupe() # Run one final pass after downloading stops.
self.progress.clear(status="Finished.", running=False)
except Exception as ex:
print('Deduplication Process Error:', ex)
self.progress.set_error(ex)
self.progress.set_running(False)
traceback.print_exc()
finally:
sql.close()
def _dedupe(self):
unfinished = self._session\
.query(File) \
.options(joinedload(File.urls))\
.filter(File.hash == None)\
.filter(File.downloaded == True)\
.all()
unfinished = list(
filter(lambda _f: not any(u.album_id for u in _f.urls),
unfinished)) # Filter out albums.
if not unfinished:
return
for idx, f in enumerate(unfinished):
self.progress.set_status("Deduplicating (%s) files..." %
(len(unfinished) - idx))
path = SanitizedRelFile(base=settings.get("output.base_dir"),
file_path=f.path)
is_album = any(u.album_id for u in f.urls)
if not path.is_file() or is_album:
continue
new_hash = FileHasher.get_best_hash(path.absolute())
# print('New hash for File:', f.id, '::', new_hash)
matches = self._find_matching_files(new_hash, ignore_id=f.id)
# print('\tActual matches:', matches)
with self._lock:
f.hash = Hash.make_hash(f, new_hash)
if len(matches):
# print("Found duplicate files: ", new_hash, "::", [(m.id, m.path) for m in matches])
best, others = self._choose_best_file(matches + [f])
# print('Chose best File:', best.id)
for o in others:
self._upgrade_file(new_file=best, old_file=o)
self._session.commit()
self._prune()
def _find_matching_files(self, search_hash, ignore_id):
sp = Hash.split_hash(search_hash)
all_hashes = self._session \
.query(File) \
.join(Hash, File.hash) \
.filter(
(Hash.full_hash == search_hash) |
(Hash.p1 == sp[0]) |
(Hash.p2 == sp[1]) |
(Hash.p3 == sp[2]) |
(Hash.p4 == sp[3])
).all()
# print(sp)
# print('Potential matches:', len(all_hashes), all_hashes)
return list(
filter(lambda f: self._check_hash_match(f, search_hash),
all_hashes))
def _check_hash_match(self, file, search_hash):
""" Compare the given hash against the given SQL File.
Returns invalid if the target File has albums, or is not fully processed.
"""
if not file.hash or any(u.album_id or not u.processed
for u in file.urls):
return False
if FileHasher.hamming_distance(search_hash, file.hash.full_hash) >= 4:
return False
return True
def _choose_best_file(self, files):
files = sorted(
files,
key=lambda f: SanitizedRelFile(
base=settings.get("output.base_dir"), file_path=f.path).size(),
reverse=True)
return files[0], files[1:]
def _upgrade_file(self, new_file, old_file):
# print('Upgrading old file:', old_file.id, old_file.path, ' -> ', new_file.id, new_file.path)
self._session.query(URL). \
filter(URL.file_id == old_file.id). \
update({URL.file_id: new_file.id})
file = SanitizedRelFile(base=settings.get("output.base_dir"),
file_path=old_file.path)
if file.is_file():
file.delete_file()
def _prune(self):
with self._lock:
orphans = self._session.query(File).filter(
~File.urls.any()).delete(synchronize_session='fetch')
self._session.commit()
class Downloader(multiprocessing.Process):
def __init__(self, reader, ack_queue, settings_json, db_lock):
"""
Create a Downloader Process, which will be bound to the queue given, listening for URLs to download.
"""
super().__init__()
self._reader = reader
self._settings = settings_json
self.progress = DownloaderProgress()
self._session = None
self._db_lock = db_lock
self._ack_queue = ack_queue
self.daemon = True
def run(self):
""" Threaded loading of elements. """
settings.from_json(self._settings)
sql.init_from_settings()
self._session = sql.session()
self.progress.clear(status="Starting up...", running=True)
failed = False
for nxt_id in self._reader:
try:
url = self._session.query(
sql.URL).filter(sql.URL.id == nxt_id).first()
if not url:
raise Exception("Unknown URL ID provided: (%s}" % nxt_id)
file = url.file
path = SanitizedRelFile(base=settings.get("output.base_dir"),
file_path=str(file.path))
self.progress.set_file(path.relative())
self.progress.set_status("Attempting to Handle URL...")
self.progress.set_running(True)
task = handlers.HandlerTask(url=url.address, file_obj=path)
resp = handlers.handle(task, self.progress)
is_album_parent = False
with self._db_lock:
if resp.album_urls:
if url.album_id:
resp.album_urls = [
] # Ignore nested Albums to avoid recursion.
else:
url.album_id = str(uuid.uuid4())
is_album_parent = True
else:
resp.album_urls = []
url.failed = not resp.success
url.failure_reason = resp.failure_reason
url.last_handler = resp.handler
url.album_is_parent = is_album_parent
if resp.rel_file:
file.downloaded = True
file.path = resp.rel_file.relative()
file.hash = None
utime(resp.rel_file.absolute(), times=(time(), time()))
self._session.commit()
# Once *all* processing is completed on this URL, the Downloader needs to ACK it.
# If any additional Album URLS were located, they should be sent before the ACK.
self._ack_queue.put(
AckPacket(url_id=nxt_id, extra_urls=resp.album_urls))
self.progress.clear(status="Waiting for URL...")
except Exception as ex:
failed = str(ex)
self._ack_queue.put(AckPacket(url_id=nxt_id, extra_urls=[]))
print(ex)
traceback.print_exc()
self.progress.set_error("Exited with error: {%s}" % failed)
break
sql.close()
self.progress.clear(
"Finished." if not failed else "Exited with error: %s" % failed,
running=False)
class Deduplicator(multiprocessing.Process):
def __init__(self, settings_json, stop_event):
"""
Create a Hasher Process, which will be bound to the stop_event, performing post-processing on downloaded Files.
"""
super().__init__()
self._settings = settings_json
self._stop_event = stop_event
self.progress = DownloaderProgress()
self.progress.clear(status="Starting up...")
self._session = None
self.daemon = True
def run(self):
""" Threaded loading of elements. """
settings.from_json(self._settings)
sql.init_from_settings()
self._session = sql.session()
self.progress.clear(status="Starting up...")
self.progress.set_running(True)
while not self._stop_event.is_set():
self._dedupe()
self.progress.set_status("Waiting for new files...")
self._stop_event.wait(2)
self._dedupe() # Run one final pass after downloading stops.
self.progress.set_running(False)
sql.close()
self.progress.clear("Finished.")
def _dedupe(self):
unfinished = self._session\
.query(File) \
.options(joinedload(File.urls))\
.filter(File.hash == None)\
.filter(File.downloaded == True)\
.all()
unfinished = list(
filter(lambda _f: not any(u.album_id for u in _f.urls),
unfinished)) # Filter out albums.
if not unfinished:
return
for idx, f in enumerate(unfinished):
self.progress.set_status("Deduplicating (%s) files..." %
(len(unfinished) - idx))
path = SanitizedRelFile(base=settings.get("output.base_dir"),
file_path=f.path)
is_album = any(u.album_id for u in f.urls)
if not path.is_file() or is_album:
continue
new_hash = FileHasher.get_best_hash(path.absolute())
# print('New hash for File:', f.id, '::', new_hash)
matches = [] if is_album else self._find_matching_files(
new_hash, ignore_id=f.id)
f.hash = new_hash
if len(matches):
print("Found duplicate files: ", new_hash, "::",
[(m.id, m.path) for m in matches])
best, others = self._choose_best_file(matches + [f])
print('Chose best File:', best.id)
for o in others:
self._upgrade_file(new_file=best, old_file=o)
self._session.commit()
self._prune()
def _find_matching_files(self, search_hash, ignore_id):
all_hashes = self._session \
.query(File) \
.options(joinedload(File.urls)) \
.filter(File.hash != None) \
.filter(File.downloaded == True) \
.filter(File.id != ignore_id) \
.all()
matches = []
for pm in all_hashes:
if any(u.album_id for u in pm.urls) or any(not u.processed
for u in pm.urls):
continue
if FileHasher.hamming_distance(search_hash, pm.hash) < 4:
matches.append(pm)
return matches
def _choose_best_file(self, files):
files = sorted(
files,
key=lambda f: SanitizedRelFile(
base=settings.get("output.base_dir"), file_path=f.path).size(),
reverse=True)
return files[0], files[1:]
def _upgrade_file(self, new_file, old_file):
print('Upgrading old file:', old_file.id, old_file.path, ' -> ',
new_file.id, new_file.path)
self._session.query(URL). \
filter(URL.file_id == old_file.id). \
update({URL.file_id: new_file.id})
file = SanitizedRelFile(base=settings.get("output.base_dir"),
file_path=old_file.path)
if file.is_file():
file.delete_file()
def _prune(self):
orphans = self._session.query(File).filter(~File.urls.any()).delete(
synchronize_session='fetch')
self._session.commit()
if orphans:
print("Deleted orphan Files:", orphans)
class Deduplicator(multiprocessing.Process):
def __init__(self, settings_json, stop_event, db_lock):
"""
Create a Hasher Process, which will be bound to the stop_event, performing post-processing on downloaded Files.
"""
super().__init__()
self._settings = settings_json
self._stop_event = stop_event
self._lock = db_lock
self.progress = DownloaderProgress()
self.progress.clear(status="Starting up...")
self._session = None
self.daemon = True
def run(self):
""" Threaded loading of elements. """
settings.from_json(self._settings)
sql.init_from_settings()
print("Starting up...", debug=True)
try:
self._session = sql.session()
self.progress.clear(status="Starting up...")
self.progress.set_running(True)
self.dedup_ignore_ids = set()
self.prune_counter = 0
self.special_hashes = self._session.query(Hash).filter(
Hash.id < 0).all()
while not self._stop_event.is_set():
#print("_stop_event is %s"%self._stop_event.is_set(), debug=True)
completed = self._dedupe()
if completed:
self.progress.set_status(
"Completed %s files. Ready for new files..." %
completed)
self._stop_event.wait(1)
else:
self._stop_event.wait(10)
print("_stop_event is %s" % self._stop_event.is_set(), debug=True)
self._dedupe() # Run one final pass after downloading stops.
self.progress.clear(status="Finished.", running=False)
except Exception as ex:
print('Deduplication Process Error:', ex)
self.progress.set_error(ex)
self.progress.set_running(False)
traceback.print_exc()
finally:
print("Finished process, _stop_event is %s" %
self._stop_event.is_set(),
debug=True)
sql.close()
def _dedupe(self):
# unfinished = self._session\
# .query(File) \
# .options(joinedload(File.urls))\
# .filter(File.hash == None)\
# .filter(File.downloaded == True)\
# .all()
start_time = datetime.now()
hashed = set(int(r.file_id) for r in self._session.query(Hash.file_id) \
.filter(Hash.full_hash != None, Hash.file_id != None))
downloaded = set(
r.id
for r in self._session.query(File).filter(File.downloaded == True))
# get downloaded files without a hash
search_ids = downloaded.difference(hashed).difference(
self.dedup_ignore_ids)
unfinished = self._session.query(File).filter(
File.id.in_(search_ids)).all()
unfinished = list(
filter(lambda _f: not any(u.album_id for u in _f.urls),
unfinished)) # Filter out albums.
#print("Working on %s files total"%len(unfinished), debug=True)
if not unfinished:
return 0
stats = {
'unique': 0,
'has_dup': 0,
'special_hash': 0,
'not_is_file': 0,
'is_album': 0
}
matches = []
last_printed = ''
for idx, f in enumerate(unfinished):
self.progress.set_status("Deduplicating %s of %s files..." %
(idx + 1, len(unfinished)))
#print("Working on %s/%s files"%(idx, len(unfinished)), debug=True)
path = SanitizedRelFile(base=settings.get("output.base_dir"),
file_path=f.path)
is_album = any(u.album_id for u in f.urls)
if not path.is_file():
stats['not_is_file'] += 1
self.dedup_ignore_ids.add(f.id)
continue
if is_album:
stats['is_album'] += 1
self.dedup_ignore_ids.add(f.id)
continue
if self._stop_event.is_set():
break
new_hash = FileHasher.get_best_hash(path.absolute())
# print('New hash for File:', f.id, '::', new_hash)
for h in self.special_hashes:
if new_hash == h.full_hash:
print("Found special hash:", h, "::\n", f, debug=True)
stats['special_hash'] += 1
with self._lock:
f.hash = Hash.make_hash(f, new_hash)
self._session.query(URL).filter(
URL.file_id == f.id).update(
{URL.file_id: h.file_id})
file = SanitizedRelFile(
base=settings.get("output.base_dir"),
file_path=f.path)
if file.is_file():
file.delete_file()
self._session.commit()
break
else: # not a special hash
matches = self._find_matching_files(new_hash, ignore_id=f.id)
if matches:
if new_hash == last_printed:
print("Found another duplicate:",
new_hash,
"::\n",
f,
debug=True)
elif len(matches) > 6:
printed = matches[:3] + [
"... %s total matches ..." % len(matches)
] + matches[-3:]
print("Found duplicate files: ",
new_hash,
"::\n",
'\n'.join(str(m) for m in [f] + printed),
debug=True)
else:
print("Found duplicate files: ",
new_hash,
"::\n",
'\n'.join(str(m) for m in [f] + matches),
debug=True)
stats['has_dup'] += 1
last_printed = new_hash
else:
stats['unique'] += 1
# print('\tActual matches:', matches)
with self._lock:
f.hash = Hash.make_hash(f, new_hash)
#print("Updating hash: ", f.id, f.hash.file_id, f.hash, debug=True)
if len(matches):
#print("Found duplicate files: ", new_hash, "::", [(m.id, m.path) for m in matches])
best, others = self._choose_best_file(matches + [f])
# print('Chose best File:', best.id)
for o in others:
self._upgrade_file(new_file=best, old_file=o)
self._session.commit()
if matches:
print("Completed %s of %s files..." %
(idx + 1, len(unfinished)),
debug=True)
dt = datetime.now() - start_time
print("Completed all %s files in %s sec. Counts = %s" %
(len(unfinished), str(dt), ', '.join(
'%s: %s' % (k, v) for k, v in stats.items() if v)),
debug=True)
# self.prune_counter += len(matches)
# if self.prune_counter >= 100:
# self.prune_counter = 0
#self.progress.set_status("Pruning orphaned files...")
#self._prune()
#print("Finished pruning.", debug=True)
return len(unfinished)
def _find_matching_files(self, search_hash, ignore_id):
sp = Hash.split_hash(search_hash)
all_hashes = self._session \
.query(File) \
.join(Hash, File.hash) \
.filter(
(Hash.full_hash == search_hash) |
(Hash.p1 == sp[0]) |
(Hash.p2 == sp[1]) |
(Hash.p3 == sp[2]) |
(Hash.p4 == sp[3])
).all()
# print(sp)
# print('Potential matches:', len(all_hashes), all_hashes)
return list(
filter(lambda f: self._check_hash_match(f, search_hash),
all_hashes))
def _check_hash_match(self, file, search_hash):
""" Compare the given hash against the given SQL File.
Returns invalid if the target File has albums, or is not fully processed.
"""
if not file.hash or any(u.album_id or not u.processed
for u in file.urls):
return False
#if FileHasher.hamming_distance(search_hash, file.hash.full_hash) >= 4:
if search_hash != file.hash.full_hash:
return False
return True
def _choose_best_file(self, files):
files = sorted(
files,
key=lambda f: SanitizedRelFile(
base=settings.get("output.base_dir"), file_path=f.path).size(),
reverse=True)
return files[0], files[1:]
def _upgrade_file(self, new_file, old_file):
# print('Upgrading old file:', old_file.id, old_file.path, ' -> ', new_file.id, new_file.path)
self._session.query(URL). \
filter(URL.file_id == old_file.id). \
update({URL.file_id: new_file.id})
file = SanitizedRelFile(base=settings.get("output.base_dir"),
file_path=old_file.path)
if file.is_file():
file.delete_file()
def _prune(self):
with self._lock:
files_id = set(r.id for r in self._session.query(File))
url_files_id = set(
int(r.file_id) for r in self._session.query(URL))
orphans = self._session.query(File).filter(
File.id.in_(files_id.difference(url_files_id))).delete(
synchronize_session='fetch')
#orphans = self._session.query(File).filter(~File.urls.any()).delete(synchronize_session='fetch')
self._session.commit()
if orphans:
print("Deleted orphan Files:", orphans, debug=True)