mirror of
https://github.com/httpie/cli.git
synced 2024-11-24 08:22:22 +02:00
375 lines
11 KiB
Python
375 lines
11 KiB
Python
# coding=utf-8
|
|
"""
|
|
Download mode implementation.
|
|
|
|
"""
|
|
from __future__ import division
|
|
import os
|
|
import re
|
|
import sys
|
|
import errno
|
|
import mimetypes
|
|
import threading
|
|
from time import time
|
|
from datetime import timedelta
|
|
|
|
from .output import RawStream
|
|
from .models import HTTPResponse
|
|
from .utils import humanize_bytes
|
|
from .compat import urlsplit
|
|
|
|
|
|
PARTIAL_CONTENT = 206
|
|
|
|
|
|
CLEAR_LINE = '\r\033[K'
|
|
PROGRESS = ('{percentage: 6.2f}% ({downloaded})'
|
|
' of {total} ({speed}/s) ETA {eta}')
|
|
PROGRESS_NO_CONTENT_LENGTH = '{downloaded} ({speed}/s) ETA {eta}'
|
|
SUMMARY = 'Done. {downloaded} of {total} in {time:0.5f}s ({speed}/s)\n'
|
|
SPINNER = '|/-\\'
|
|
|
|
|
|
class ContentRangeError(ValueError):
|
|
pass
|
|
|
|
|
|
def _parse_content_range(content_range, resumed_from):
|
|
"""
|
|
Parse and validate Content-Range header.
|
|
|
|
<http://www.w3.org/Protocols/rfc2616/rfc2616-sec14.html>
|
|
|
|
:param content_range: the value of a Content-Range response header
|
|
eg. "bytes 21010-47021/47022"
|
|
:param resumed_from: first byte pos. from the Range request header
|
|
:return: total size of the response body when fully downloaded.
|
|
|
|
"""
|
|
pattern = (
|
|
'^bytes (?P<first_byte_pos>\d+)-(?P<last_byte_pos>\d+)'
|
|
'/(\*|(?P<instance_length>\d+))$'
|
|
)
|
|
match = re.match(pattern, content_range)
|
|
|
|
if not match:
|
|
raise ContentRangeError(
|
|
'Invalid Content-Range format %r' % content_range)
|
|
|
|
content_range_dict = match.groupdict()
|
|
first_byte_pos = int(content_range_dict['first_byte_pos'])
|
|
last_byte_pos = int(content_range_dict['last_byte_pos'])
|
|
instance_length = (
|
|
int(content_range_dict['instance_length'])
|
|
if content_range_dict['instance_length']
|
|
else None
|
|
)
|
|
|
|
# "A byte-content-range-spec with a byte-range-resp-spec whose
|
|
# last- byte-pos value is less than its first-byte-pos value,
|
|
# or whose instance-length value is less than or equal to its
|
|
# last-byte-pos value, is invalid. The recipient of an invalid
|
|
# byte-content-range- spec MUST ignore it and any content
|
|
# transferred along with it."
|
|
if (first_byte_pos >= last_byte_pos
|
|
or (instance_length is not None
|
|
and instance_length <= last_byte_pos)):
|
|
raise ContentRangeError(
|
|
'Invalid Content-Range returned: %r' % content_range)
|
|
|
|
if (first_byte_pos != resumed_from
|
|
or (instance_length is not None
|
|
and last_byte_pos + 1 != instance_length)):
|
|
# Not what we asked for.
|
|
raise ContentRangeError(
|
|
'Unexpected Content-Range returned (%r)'
|
|
' for the requested Range ("bytes=%d-")'
|
|
% (content_range, resumed_from)
|
|
)
|
|
|
|
return last_byte_pos + 1
|
|
|
|
|
|
class Download(object):
|
|
|
|
def __init__(self, output_file=None,
|
|
resume=False, progress_file=sys.stderr):
|
|
"""
|
|
:param resume: Should the download resume if partial download
|
|
already exists.
|
|
:type resume: bool
|
|
|
|
:param output_file: The file to store response body in. If not
|
|
provided, it will be guessed from the response.
|
|
:type output_file: file
|
|
|
|
:param progress_file: Where to report download progress.
|
|
:type progress_file: file
|
|
|
|
"""
|
|
self._output_file = output_file
|
|
self._resume = resume
|
|
self._resumed_from = 0
|
|
|
|
self._progress = Progress()
|
|
self._progress_reporter = ProgressReporter(
|
|
progress=self._progress,
|
|
output=progress_file
|
|
)
|
|
|
|
def pre_request(self, request_headers):
|
|
"""Called just before the HTTP request is sent.
|
|
|
|
Might alter `request_headers`.
|
|
|
|
:type request_headers: dict
|
|
|
|
"""
|
|
# Disable content encoding so that we can resume, etc.
|
|
request_headers['Accept-Encoding'] = None
|
|
if self._resume:
|
|
try:
|
|
bytes_have = os.path.getsize(self._output_file.name)
|
|
except OSError as e:
|
|
if e.errno != errno.ENOENT:
|
|
raise
|
|
else:
|
|
# Set ``Range`` header to resume the download
|
|
# TODO: Use "If-Range: mtime" to make sure it's fresh?
|
|
request_headers['Range'] = 'bytes=%d-' % bytes_have
|
|
self._resumed_from = bytes_have
|
|
|
|
def start(self, response):
|
|
"""
|
|
Initiate and return a stream for `response` body with progress
|
|
callback attached. Can be called only once.
|
|
|
|
:param response: Initiated response object with headers already fetched
|
|
:type response: requests.models.Response
|
|
|
|
:return: RawStream, output_file
|
|
|
|
"""
|
|
assert not self._progress.time_started
|
|
|
|
try:
|
|
total_size = int(response.headers['Content-Length'])
|
|
except (KeyError, ValueError):
|
|
total_size = None
|
|
|
|
if self._output_file:
|
|
if self._resume and response.status_code == PARTIAL_CONTENT:
|
|
content_range = response.headers.get('Content-Range')
|
|
if content_range:
|
|
total_size = _parse_content_range(
|
|
content_range, self._resumed_from)
|
|
|
|
else:
|
|
self._resumed_from = 0
|
|
self._output_file.seek(0)
|
|
self._output_file.truncate()
|
|
else:
|
|
# TODO: Should the filename be taken from response.history[0].url?
|
|
# Output file not specified. Pick a name that doesn't exist yet.
|
|
content_type = response.headers.get('Content-Type', '')
|
|
self._output_file = open(
|
|
self._get_unique_output_filename(
|
|
url=response.url,
|
|
content_type=content_type,
|
|
),
|
|
mode='a+b'
|
|
)
|
|
|
|
self._progress.started(
|
|
resumed_from=self._resumed_from,
|
|
total_size=total_size
|
|
)
|
|
|
|
stream = RawStream(
|
|
msg=HTTPResponse(response),
|
|
with_headers=False,
|
|
with_body=True,
|
|
on_body_chunk_downloaded=self._on_progress,
|
|
# TODO: Find the optimal chunk size.
|
|
# The smaller it is the slower it gets, but gives better feedback.
|
|
chunk_size=1024 * 8
|
|
)
|
|
|
|
self._progress_reporter.output.write(
|
|
'Saving to "%s"\n' % self._output_file.name)
|
|
self._progress_reporter.report()
|
|
|
|
return stream, self._output_file
|
|
|
|
def finish(self):
|
|
assert not self._output_file.closed
|
|
self._output_file.close()
|
|
self._progress.finished()
|
|
|
|
@property
|
|
def interrupted(self):
|
|
return (
|
|
self._output_file.closed
|
|
and self._progress.total_size
|
|
and self._progress.total_size != self._progress.downloaded
|
|
)
|
|
|
|
def _on_progress(self, chunk):
|
|
"""
|
|
A download progress callback.
|
|
|
|
:param chunk: A chunk of response body data that has just
|
|
been downloaded and written to the output.
|
|
:type chunk: bytes
|
|
|
|
"""
|
|
self._progress.chunk_downloaded(len(chunk))
|
|
|
|
def _get_unique_output_filename(self, url, content_type):
|
|
suffix = 0
|
|
while True:
|
|
fn = self._get_output_filename(url, content_type, suffix)
|
|
if not os.path.exists(fn):
|
|
return fn
|
|
suffix += 1
|
|
|
|
def _get_output_filename(self, url, content_type, suffix=None):
|
|
|
|
suffix = '' if not suffix else '-' + str(suffix)
|
|
|
|
fn = urlsplit(url).path.rstrip('/')
|
|
fn = os.path.basename(fn) if fn else 'index'
|
|
|
|
if '.' in fn:
|
|
base, ext = os.path.splitext(fn)
|
|
else:
|
|
base = fn
|
|
ext = mimetypes.guess_extension(content_type.split(';')[0]) or ''
|
|
|
|
return base + suffix + ext
|
|
|
|
|
|
class Progress(object):
|
|
|
|
def __init__(self):
|
|
self.downloaded = 0
|
|
self.total_size = None
|
|
self.resumed_from = 0
|
|
self.total_size_humanized = '?'
|
|
self.time_started = None
|
|
self.time_finished = None
|
|
|
|
def started(self, resumed_from=0, total_size=None):
|
|
assert self.time_started is None
|
|
if total_size is not None:
|
|
self.total_size_humanized = humanize_bytes(total_size)
|
|
self.total_size = total_size
|
|
self.downloaded = self.resumed_from = resumed_from
|
|
self.time_started = time()
|
|
|
|
def chunk_downloaded(self, size):
|
|
assert self.time_finished is None
|
|
self.downloaded += size
|
|
|
|
@property
|
|
def has_finished(self):
|
|
return self.time_finished is not None
|
|
|
|
def finished(self):
|
|
assert self.time_started is not None
|
|
assert self.time_finished is None
|
|
self.time_finished = time()
|
|
|
|
|
|
class ProgressReporter(object):
|
|
|
|
def __init__(self, progress, output, tick=.1, update_interval=1):
|
|
"""
|
|
|
|
:type progress: Progress
|
|
:type output: file
|
|
"""
|
|
self.progress = progress
|
|
self.output = output
|
|
self._prev_bytes = 0
|
|
self._prev_time = time()
|
|
self._speed = 0
|
|
self._spinner_pos = 0
|
|
self._tick = tick
|
|
self._update_interval = update_interval
|
|
self._status_line = ''
|
|
super(ProgressReporter, self).__init__()
|
|
|
|
def report(self):
|
|
if self.progress.has_finished:
|
|
self.sum_up()
|
|
else:
|
|
self.report_speed()
|
|
# TODO: quit on KeyboardInterrupt
|
|
threading.Timer(self._tick, self.report).start()
|
|
|
|
def report_speed(self):
|
|
|
|
now = time()
|
|
|
|
if now - self._prev_time >= self._update_interval:
|
|
|
|
downloaded = self.progress.downloaded
|
|
|
|
if self.progress.total_size:
|
|
template = PROGRESS
|
|
percentage = (
|
|
downloaded / self.progress.total_size * 100)
|
|
else:
|
|
template = PROGRESS_NO_CONTENT_LENGTH
|
|
percentage = None
|
|
|
|
try:
|
|
# TODO: Use a longer interval for the speed/eta calculation?
|
|
speed = ((downloaded - self._prev_bytes)
|
|
/ (now - self._prev_time))
|
|
eta = int((self.progress.total_size - downloaded) / speed)
|
|
eta = str(timedelta(seconds=eta))
|
|
except ZeroDivisionError:
|
|
speed = 0
|
|
eta = '?'
|
|
|
|
self._status_line = template.format(
|
|
percentage=percentage,
|
|
downloaded=humanize_bytes(downloaded),
|
|
total=self.progress.total_size_humanized,
|
|
speed=humanize_bytes(speed),
|
|
eta=eta,
|
|
)
|
|
|
|
self._prev_time = now
|
|
self._prev_bytes = downloaded
|
|
|
|
self.output.write(
|
|
CLEAR_LINE
|
|
+ SPINNER[self._spinner_pos]
|
|
+ ' '
|
|
+ self._status_line
|
|
)
|
|
self.output.flush()
|
|
|
|
self._spinner_pos = (
|
|
self._spinner_pos + 1
|
|
if self._spinner_pos + 1 != len(SPINNER)
|
|
else 0
|
|
)
|
|
|
|
def sum_up(self):
|
|
actually_downloaded = (
|
|
self.progress.downloaded - self.progress.resumed_from)
|
|
time_taken = self.progress.time_finished - self.progress.time_started
|
|
|
|
self.output.write(CLEAR_LINE)
|
|
self.output.write(SUMMARY.format(
|
|
downloaded=humanize_bytes(actually_downloaded),
|
|
total=humanize_bytes(self.progress.downloaded),
|
|
speed=humanize_bytes(actually_downloaded / time_taken),
|
|
time=time_taken,
|
|
))
|
|
self.output.flush()
|