ComfyUI/app/assets/services/hashing.py
Luke Mino-Altherr 481a2fa263 refactor: rename functions to verb-based naming convention
Rename functions across app/assets/ to follow verb-based naming:
- is_scalar → check_is_scalar
- project_kv → expand_metadata_to_rows
- _visible_owner_clause → _build_visible_owner_clause
- _chunk_rows → _iter_row_chunks
- _at_least_one → _validate_at_least_one_field
- _tags_norm → _normalize_tags_field
- _ser_dt → _serialize_datetime
- _ser_updated → _serialize_updated_at
- _error_response → _build_error_response
- _validation_error_response → _build_validation_error_response
- file_sender → stream_file_chunks
- seed_assets_endpoint → seed_assets
- utcnow → get_utc_now
- _safe_sort_field → _validate_sort_field
- _safe_filename → _sanitize_filename
- fast_asset_file_check → check_asset_file_fast
- prefixes_for_root → get_prefixes_for_root
- blake3_hash → compute_blake3_hash
- blake3_hash_async → compute_blake3_hash_async
- _is_within → _check_is_within
- _rel → _compute_relative

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
2026-02-03 13:58:14 -08:00

76 lines
2.4 KiB
Python

from blake3 import blake3
from typing import IO
import os
import asyncio
DEFAULT_CHUNK = 8 * 1024 *1024 # 8MB
# NOTE: this allows hashing different representations of a file-like object
def compute_blake3_hash(
fp: str | IO[bytes],
chunk_size: int = DEFAULT_CHUNK,
) -> str:
"""
Returns a BLAKE3 hex digest for ``fp``, which may be:
- a filename (str/bytes) or PathLike
- an open binary file object
If ``fp`` is a file object, it must be opened in **binary** mode and support
``read``, ``seek``, and ``tell``. The function will seek to the start before
reading and will attempt to restore the original position afterward.
"""
# duck typing to check if input is a file-like object
if hasattr(fp, "read"):
return _hash_file_obj(fp, chunk_size)
with open(os.fspath(fp), "rb") as f:
return _hash_file_obj(f, chunk_size)
async def compute_compute_blake3_hash_async(
fp: str | IO[bytes],
chunk_size: int = DEFAULT_CHUNK,
) -> str:
"""Async wrapper for ``compute_blake3_hash_sync``.
Uses a worker thread so the event loop remains responsive.
"""
# If it is a path, open inside the worker thread to keep I/O off the loop.
if hasattr(fp, "read"):
return await asyncio.to_thread(compute_blake3_hash, fp, chunk_size)
def _worker() -> str:
with open(os.fspath(fp), "rb") as f:
return _hash_file_obj(f, chunk_size)
return await asyncio.to_thread(_worker)
def _hash_file_obj(file_obj: IO, chunk_size: int = DEFAULT_CHUNK) -> str:
"""
Hash an already-open binary file object by streaming in chunks.
- Seeks to the beginning before reading (if supported).
- Restores the original position afterward (if tell/seek are supported).
"""
if chunk_size <= 0:
chunk_size = DEFAULT_CHUNK
# in case file object is already open and not at the beginning, track so can be restored after hashing
orig_pos = file_obj.tell()
try:
# seek to the beginning before reading
if orig_pos != 0:
file_obj.seek(0)
h = blake3()
while True:
chunk = file_obj.read(chunk_size)
if not chunk:
break
h.update(chunk)
return h.hexdigest()
finally:
# restore original position in file object, if needed
if orig_pos != 0:
file_obj.seek(orig_pos)