mirror of
https://github.com/devine-dl/devine.git
synced 2025-11-04 03:44:49 +00:00
fix(requests): Fix multithreaded downloads
For some reason moving the download speed calculation code from the requests() function to the download() function makes it actually multi-threaded instead of sequential downloads.
This commit is contained in:
rlaphoenix
parent
5d1b54b8fa
commit
994ab152a4
@ -1,7 +1,7 @@
|
||||
import math
|
||||
import os
|
||||
import time
|
||||
from concurrent import futures
|
||||
from concurrent.futures import as_completed
|
||||
from concurrent.futures.thread import ThreadPoolExecutor
|
||||
from http.cookiejar import CookieJar
|
||||
from pathlib import Path
|
||||
@ -19,11 +19,14 @@ RETRY_WAIT = 2
|
||||
CHUNK_SIZE = 1024
|
||||
PROGRESS_WINDOW = 5
|
||||
|
||||
DOWNLOAD_SIZES = []
|
||||
LAST_SPEED_REFRESH = time.time()
|
||||
|
||||
def download(
|
||||
url: str,
|
||||
save_path: Path,
|
||||
session: Optional[Session] = None,
|
||||
segmented: bool = False,
|
||||
**kwargs: Any
|
||||
) -> Generator[dict[str, Any], None, None]:
|
||||
"""
|
||||
@ -48,10 +51,13 @@ def download(
|
||||
session: The Requests Session to make HTTP requests with. Useful to set Header,
|
||||
Cookie, and Proxy data. Connections are saved and re-used with the session
|
||||
so long as the server keeps the connection alive.
|
||||
segmented: If downloads are segments or parts of one bigger file.
|
||||
kwargs: Any extra keyword arguments to pass to the session.get() call. Use this
|
||||
for one-time request changes like a header, cookie, or proxy. For example,
|
||||
to request Byte-ranges use e.g., `headers={"Range": "bytes=0-128"}`.
|
||||
"""
|
||||
global LAST_SPEED_REFRESH
|
||||
|
||||
session = session or Session()
|
||||
|
||||
save_dir = save_path.parent
|
||||
@ -69,6 +75,7 @@ def download(
|
||||
file_downloaded=save_path,
|
||||
written=save_path.stat().st_size
|
||||
)
|
||||
# TODO: This should return, potential recovery bug
|
||||
|
||||
# TODO: Design a control file format so we know how much of the file is missing
|
||||
control_file.write_bytes(b"")
|
||||
@ -77,6 +84,8 @@ def download(
|
||||
try:
|
||||
while True:
|
||||
written = 0
|
||||
|
||||
# these are for single-url speed calcs only
|
||||
download_sizes = []
|
||||
last_speed_refresh = time.time()
|
||||
|
||||
@ -84,6 +93,7 @@ def download(
|
||||
stream = session.get(url, stream=True, **kwargs)
|
||||
stream.raise_for_status()
|
||||
|
||||
if not segmented:
|
||||
try:
|
||||
content_length = int(stream.headers.get("Content-Length", "0"))
|
||||
except ValueError:
|
||||
@ -101,11 +111,10 @@ def download(
|
||||
f.write(chunk)
|
||||
written += download_size
|
||||
|
||||
if not segmented:
|
||||
yield dict(advance=1)
|
||||
|
||||
now = time.time()
|
||||
time_since = now - last_speed_refresh
|
||||
|
||||
download_sizes.append(download_size)
|
||||
if time_since > PROGRESS_WINDOW or download_size < CHUNK_SIZE:
|
||||
data_size = sum(download_sizes)
|
||||
@ -114,10 +123,20 @@ def download(
|
||||
last_speed_refresh = now
|
||||
download_sizes.clear()
|
||||
|
||||
yield dict(
|
||||
file_downloaded=save_path,
|
||||
written=written
|
||||
)
|
||||
yield dict(file_downloaded=save_path, written=written)
|
||||
|
||||
if segmented:
|
||||
yield dict(advance=1)
|
||||
now = time.time()
|
||||
time_since = now - LAST_SPEED_REFRESH
|
||||
if written: # no size == skipped dl
|
||||
DOWNLOAD_SIZES.append(written)
|
||||
if DOWNLOAD_SIZES and time_since > PROGRESS_WINDOW:
|
||||
data_size = sum(DOWNLOAD_SIZES)
|
||||
download_speed = math.ceil(data_size / (time_since or 1))
|
||||
yield dict(downloaded=f"{filesize.decimal(download_speed)}/s")
|
||||
LAST_SPEED_REFRESH = now
|
||||
DOWNLOAD_SIZES.clear()
|
||||
break
|
||||
except Exception as e:
|
||||
save_path.unlink(missing_ok=True)
|
||||
@ -237,27 +256,19 @@ def requests(
|
||||
|
||||
yield dict(total=len(urls))
|
||||
|
||||
download_sizes = []
|
||||
last_speed_refresh = time.time()
|
||||
|
||||
try:
|
||||
with ThreadPoolExecutor(max_workers=max_workers) as pool:
|
||||
for i, future in enumerate(futures.as_completed((
|
||||
for future in as_completed(
|
||||
pool.submit(
|
||||
download,
|
||||
session=session,
|
||||
segmented=True,
|
||||
**url
|
||||
)
|
||||
for url in urls
|
||||
))):
|
||||
file_path, download_size = None, None
|
||||
):
|
||||
try:
|
||||
for status_update in future.result():
|
||||
if status_update.get("file_downloaded") and status_update.get("written"):
|
||||
file_path = status_update["file_downloaded"]
|
||||
download_size = status_update["written"]
|
||||
elif len(urls) == 1:
|
||||
# these are per-chunk updates, only useful if it's one big file
|
||||
yield status_update
|
||||
yield from future.result()
|
||||
except KeyboardInterrupt:
|
||||
DOWNLOAD_CANCELLED.set() # skip pending track downloads
|
||||
yield dict(downloaded="[yellow]CANCELLING")
|
||||
@ -274,22 +285,8 @@ def requests(
|
||||
# tell dl that it failed
|
||||
# the pool is already shut down, so exiting loop is fine
|
||||
raise
|
||||
else:
|
||||
yield dict(file_downloaded=file_path, written=download_size)
|
||||
yield dict(advance=1)
|
||||
|
||||
now = time.time()
|
||||
time_since = now - last_speed_refresh
|
||||
|
||||
if download_size: # no size == skipped dl
|
||||
download_sizes.append(download_size)
|
||||
|
||||
if download_sizes and (time_since > PROGRESS_WINDOW or i == len(urls)):
|
||||
data_size = sum(download_sizes)
|
||||
download_speed = math.ceil(data_size / (time_since or 1))
|
||||
yield dict(downloaded=f"{filesize.decimal(download_speed)}/s")
|
||||
last_speed_refresh = now
|
||||
download_sizes.clear()
|
||||
finally:
|
||||
DOWNLOAD_SIZES.clear()
|
||||
|
||||
|
||||
__all__ = ("requests",)
|
||||
|
||||
Loading…
x
Reference in New Issue
Block a user