arsenetar
/
dupeguru
mirror of https://github.com/arsenetar/dupeguru.git
1
0
Fork 0
Code Releases Wiki Activity

match all orientations

This commit is contained in:
Bruno Cabral 2023-05-13 18:03:13 -07:00
parent 322d29a996
commit b00f0bf4f1
5 changed files with 43 additions and 32 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
core/engine.py
View File

@ -304,12 +304,12 @@ def getmatches_by_contents(files, bigsize=0, j=job.nulljob):
result.append(Match(first, second, 100))
continue
# if digests are the same (and not None) then files match
if first.digest_partial == second.digest_partial and first.digest_partial is not None:
if first.digest_partial is not None and first.digest_partial == second.digest_partial:
if bigsize > 0 and first.size > bigsize:
if first.digest_samples == second.digest_samples and first.digest_samples is not None:
if first.digest_samples is not None and first.digest_samples == second.digest_samples:
result.append(Match(first, second, 100))
else:
if first.digest == second.digest and first.digest is not None:
if first.digest is not None and first.digest == second.digest:
result.append(Match(first, second, 100))
group_count += 1
j.add_progress(desc=PROGRESS_MESSAGE % (len(result), group_count))

4
core/fs.py
View File

@ -199,7 +199,7 @@ class File:
# Slots for File make us save quite a bit of memory. In a memory test I've made with a lot of
# files, I saved 35% memory usage with "unread" files (no _read_info() call) and gains become
# even greater when we take into account read attributes (70%!). Yeah, it's worth it.
__slots__ = ("path", "is_ref", "words") + tuple(INITIAL_INFO.keys())
__slots__ = ("path", "unicode_path", "is_ref", "words") + tuple(INITIAL_INFO.keys())
def __init__(self, path):
for attrname in self.INITIAL_INFO:
@ -210,6 +210,8 @@ class File:
self.mtime = nonone(path.stat().st_mtime, 0)
else:
self.path = path
if self.path:
self.unicode_path = str(self.path)
def __repr__(self):
return f"<{self.__class__.__name__} {str(self.path)}>"

28
core/pe/cache_sqlite.py
View File

@ -15,10 +15,10 @@ from core.pe.cache import bytes_to_colors, colors_to_bytes
class SqliteCache:
"""A class to cache picture blocks in a sqlite backend."""
schema_version = 1
schema_version_description = "Changed from string to bytes for blocks."
schema_version = 2
schema_version_description = "Added blocks for all 8 orientations."
create_table_query = "CREATE TABLE IF NOT EXISTS pictures(path TEXT, mtime_ns INTEGER, blocks BLOB)"
create_table_query = "CREATE TABLE IF NOT EXISTS pictures(path TEXT, mtime_ns INTEGER, blocks BLOB, blocks2 BLOB, blocks3 BLOB, blocks4 BLOB, blocks5 BLOB, blocks6 BLOB, blocks7 BLOB, blocks8 BLOB)"
create_index_query = "CREATE INDEX IF NOT EXISTS idx_path on pictures (path)"
drop_table_query = "DROP TABLE IF EXISTS pictures"
drop_index_query = "DROP INDEX IF EXISTS idx_path"
@ -43,12 +43,12 @@ class SqliteCache:
# Optimized
def __getitem__(self, key):
if isinstance(key, int):
sql = "select blocks from pictures where rowid = ?"
sql = "select blocks, blocks2, blocks3, blocks4, blocks5, blocks6, blocks7, blocks8 from pictures where rowid = ?"
else:
sql = "select blocks from pictures where path = ?"
result = self.con.execute(sql, [key]).fetchone()
if result:
result = bytes_to_colors(result[0])
sql = "select blocks, blocks2, blocks3, blocks4, blocks5, blocks6, blocks7, blocks8 from pictures where path = ?"
blocks = self.con.execute(sql, [key]).fetchone()
if blocks:
result = [bytes_to_colors(block) for block in blocks]
return result
else:
raise KeyError(key)
@ -64,17 +64,17 @@ class SqliteCache:
return result[0][0]
def __setitem__(self, path_str, blocks):
blocks = colors_to_bytes(blocks)
blocks = [colors_to_bytes(block) for block in blocks]
if op.exists(path_str):
mtime = int(os.stat(path_str).st_mtime)
else:
mtime = 0
if path_str in self:
sql = "update pictures set blocks = ?, mtime_ns = ? where path = ?"
sql = "update pictures set blocks = ?, blocks2 = ?, blocks3 = ?, blocks4 = ?, blocks5 = ?, blocks6 = ?, blocks7 = ?, blocks8 = ?, mtime_ns = ? where path = ?"
else:
sql = "insert into pictures(blocks,mtime_ns,path) values(?,?,?)"
sql = "insert into pictures(blocks,blocks2,blocks3,blocks4,blocks5,blocks6,blocks7,blocks8,mtime_ns,path) values(?,?,?,?,?,?,?,?,?,?)"
try:
self.con.execute(sql, [blocks, mtime, path_str])
self.con.execute(sql, blocks + [mtime, path_str])
except sqlite.OperationalError:
logging.warning("Picture cache could not set value for key %r", path_str)
except sqlite.DatabaseError as e:
@ -136,9 +136,9 @@ class SqliteCache:
raise ValueError(path)
def get_multiple(self, rowids):
sql = "select rowid, blocks from pictures where rowid in (%s)" % ",".join(map(str, rowids))
sql = "select rowid, blocks, blocks2, blocks3, blocks4, blocks5, blocks6, blocks7, blocks8 from pictures where rowid in (%s)" % ",".join(map(str, rowids))
cur = self.con.execute(sql)
return ((rowid, bytes_to_colors(blocks)) for rowid, blocks in cur)
return ((rowid, [bytes_to_colors(blocks), bytes_to_colors(blocks2), bytes_to_colors(blocks3), bytes_to_colors(blocks4), bytes_to_colors(blocks5), bytes_to_colors(blocks6), bytes_to_colors(blocks7), bytes_to_colors(blocks8)]) for rowid, blocks, blocks2, blocks3, blocks4, blocks5, blocks6, blocks7, blocks8 in cur)
def purge_outdated(self):
"""Go through the cache and purge outdated records.

30
core/pe/matchblock.py
View File

@ -72,13 +72,12 @@ def prepare_pictures(pictures, cache_path, with_dimensions, j=job.nulljob):
# entry in iPhoto library.
logging.warning("We have a picture with a null path here")
continue
picture.unicode_path = str(picture.path)
logging.debug("Analyzing picture at %s", picture.unicode_path)
if with_dimensions:
picture.dimensions # pre-read dimensions
try:
if picture.unicode_path not in cache:
blocks = picture.get_blocks(BLOCK_COUNT_PER_SIDE)
blocks = [picture.get_blocks(BLOCK_COUNT_PER_SIDE, orientation) for orientation in range(1, 9)]
cache[picture.unicode_path] = blocks
prepared.append(picture)
except (OSError, ValueError) as e:
@ -125,7 +124,7 @@ def async_compare(ref_ids, other_ids, dbname, threshold, picinfo):
# picinfo is a dictionary {pic_id: (dimensions, is_ref)}
cache = get_cache(dbname, readonly=True)
limit = 100 - threshold
ref_pairs = list(cache.get_multiple(ref_ids))
ref_pairs = list(cache.get_multiple(ref_ids)) # (rowid, [b, b2, ..., b8])
if other_ids is not None:
other_pairs = list(cache.get_multiple(other_ids))
comparisons_to_do = [(r, o) for r in ref_pairs for o in other_pairs]
@ -137,15 +136,22 @@ def async_compare(ref_ids, other_ids, dbname, threshold, picinfo):
other_dimensions, other_is_ref = picinfo[other_id]
if ref_is_ref and other_is_ref:
continue
if ref_dimensions != other_dimensions:
rotated_ref_dimensions = (ref_dimensions[1], ref_dimensions[0])
if ref_dimensions != other_dimensions and rotated_ref_dimensions != other_dimensions:
continue
try:
diff = avgdiff(ref_blocks, other_blocks, limit, MIN_ITERATIONS)
percentage = 100 - diff
except (DifferentBlockCountError, NoBlocksError):
percentage = 0
if percentage >= threshold:
results.append((ref_id, other_id, percentage))
for orientation_ref in range(8):
for orientation_other in range(8):
try:
diff = avgdiff(ref_blocks[orientation_ref], other_blocks[orientation_other], limit, MIN_ITERATIONS)
percentage = 100 - diff
except (DifferentBlockCountError, NoBlocksError):
percentage = 0
if percentage >= threshold:
results.append((ref_id, other_id, percentage))
break
else:
continue
break
cache.close()
return results
@ -153,7 +159,7 @@ def async_compare(ref_ids, other_ids, dbname, threshold, picinfo):
def getmatches(pictures, cache_path, threshold, match_scaled=False, j=job.nulljob):
def get_picinfo(p):
if match_scaled:
return (None, p.is_ref)
return ((None, None), p.is_ref)
else:
return (p.dimensions, p.is_ref)

7
core/pe/photo.py
View File

@ -100,5 +100,8 @@ class Photo(fs.File):
elif field == "exif_timestamp":
self.exif_timestamp = self._get_exif_timestamp()
def get_blocks(self, block_count_per_side):
return self._plat_get_blocks(block_count_per_side, self._get_orientation())
def get_blocks(self, block_count_per_side, orientation: int = None):
if orientation is None:
return self._plat_get_blocks(block_count_per_side, self._get_orientation())
else:
return self._plat_get_blocks(block_count_per_side, orientation)