mediafire_dl.py

import argparse
from collections import defaultdict
from datetime import datetime
from http import cookiejar
import json
import logging
import os
import re
import sys
import signal

import requests
from requests.adapters import HTTPAdapter
from requests.packages.urllib3.util.retry import Retry
from tqdm import tqdm, trange


logger = logging.getLogger(__name__)

ASCII_CONTROL = re.compile(r"[\x00-\x1f\x7f]")
FILENAME_FORBIDDEN = re.compile(r'[\\/:*?"<>|]')
LEADING_DASH_DOT = re.compile(r"_*[-.]")
TRAILING_DOTS = re.compile(r"\.+$")


def try_mkdir(path):
    try:
        os.mkdir(path)
    except FileExistsError:
        pass


def search_group(pattern, string, group=1):
    match = pattern.search(string)
    if match:
        if group is None:
            return match.groups()
        else:
            return match.group(group)


def sanitize_filename(filename, max_bytes=255):
    filename = ASCII_CONTROL.sub("", filename)
    # The forbidden characters are common enough that we replace them with an
    # underscore to show that a character was replaced.
    filename = FILENAME_FORBIDDEN.sub("_", filename)
    filename = filename.strip()
    # Prepend an underscore to files starting with a dot or dash--this prevents
    # the file from being hidden or being interpreted as a flag on Unix-likes.
    # We need _* to ensure that prepending an underscore doesn't conflict with
    # any existing files.
    if LEADING_DASH_DOT.match(filename):
        filename = "_" + filename
    # Limit the filename to max_bytes at most. We assume UTF-8.
    encoded = filename.encode()
    if len(encoded) > max_bytes:
        filename = encoded[:max_bytes].decode(errors="ignore")
        # Truncation might have exposed trailing spaces, so we strip again.
        filename = filename.strip()
    return filename


def get_clean_filename(qk, name):
    # Windows doesn't like files that end with a period
    name = TRAILING_DOTS.sub("_", name)
    root, ext = os.path.splitext(name)
    # We ignore the extension if it contains forbidden characters.
    if ASCII_CONTROL.search(ext) or FILENAME_FORBIDDEN.search(ext):
        root, ext = name, ""

    # Leave room for underscore, 15-character quickkey, extension, and ".json"
    sanitized = sanitize_filename(root, max_bytes=255 - 1 - 15 - len(ext) - 5)
    return f"{sanitized}_{qk}{ext}"


def get_clean_foldername(fk, name):
    # Leave room for underscore, 13-character folderkey, and ".json"
    sanitized = sanitize_filename(name, max_bytes=255 - 1 - 13 - 5)
    return f"{sanitized}_{fk}"


class NoCookiePolicy(cookiejar.DefaultCookiePolicy):
    return_ok = set_ok = domain_return_ok = path_return_ok = lambda self, *args: False


class MediafireError(Exception):
    def __init__(self, code, message):
        self.code = code
        self.message = message

    def __str__(self):
        # The default __str__ uses repr() on self.message, which is hard to
        # read if there are strings nested in the message (see the "Failed to
        # decode JSON" error)
        return f"({self.code}, {self.message})"


class MediafireDownloader:
    MEDIAFIRE_API_BASE = "https://www.mediafire.com/api/1.5/"

    DOWNLOAD_CHUNK_SIZE = 2 ** 20
    # Required for scraping custom folders
    USER_AGENT = "Mozilla/5.0 (Windows NT 10.0; rv:78.0) Gecko/20100101 Firefox/78.0"
    # Just below the max of 500 to be safe
    FILE_INFO_CHUNK_SIZE = 450
    # Between 100 to 1000
    FOLDER_CONTENTS_CHUNK_SIZE = 1000
    MAX_REDIRECTS = requests.models.DEFAULT_REDIRECT_LIMIT

    QUICKKEY_RE = re.compile(r"[a-z0-9]{11}|[a-z0-9]{15}")
    DL_URL_RE = re.compile(r"https://download\d+\.mediafire\.com/")
    DL_URL_SCRAPE_RE = re.compile(
        r'aria-label="Download file"\n +href="([^"]+)"\n +id="downloadButton"'
    )
    UPLOAD_COUNTRY_SCRAPE_RE = re.compile(
        r'<div class="lazyload DLExtraInfo-sectionGraphic flag" data-lazyclass="flag-(..)">'
    )
    CUSTOM_FOLDER_SCRAPE_RE = re.compile(r'gs= false,afI= "([a-z0-9]{13})",afQ= 0')
    ERROR_URL_RE = re.compile(r"mediafire\.com/error\.php\?errno=(\d+)")
    TAKEDOWN_SCRAPE_RE = re.compile(
        r"taken down on <b>([^<]+)</b>.*Company: <b>([^<]+)</b>.*Email: <b>([^>]+)</b>"
    )
    PASSWORD_PROTECTED_RE = re.compile(
        r'<label for="downloadp" style="display: none">Enter Password</label>'
    )

    def __init__(self, retry_total, retry_backoff):
        self.s = requests.Session()
        # We use a Session to reuse connections. But downloading lots of files
        # will accumulate cookies until they pass the 8k limit and cause "400
        # Request Header Or Cookie Too Large" errors
        self.s.cookies.set_policy(NoCookiePolicy())
        retry = Retry(total=retry_total, backoff_factor=retry_backoff)
        adapter = HTTPAdapter(max_retries=retry)
        self.s.mount("http://", adapter)
        self.s.mount("https://", adapter)

    def mf_api(self, method, data):
        data["response_format"] = "json"
        r = self.s.post(self.MEDIAFIRE_API_BASE + method + ".php", data=data)
        try:
            resp = r.json()["response"]
        except Exception as exc:
            raise MediafireError(
                None,
                f"Failed to decode JSON:\n{exc=}\n{method=}\n{data=}\n"
                f"{r.text=}\n{r.request.headers=}\n{r.headers=}",
            )

        if resp["result"] == "Success":
            return resp
        else:
            raise MediafireError(resp["error"], resp["message"])

    def get_file_info(self, quickkeys):
        if not quickkeys:
            return [], []

        def _validate_quickkey(k):
            valid = self.QUICKKEY_RE.fullmatch(k)
            if not valid:
                logger.warning("f{k} is not a valid quickkey")
            return valid

        quickkeys = list(filter(_validate_quickkey, quickkeys))
        file_info = []
        skipped = []
        for i in trange(
            0,
            len(quickkeys),
            self.FILE_INFO_CHUNK_SIZE,
            unit="req",
            desc="Get file info",
        ):
            chunk = quickkeys[i : i + self.FILE_INFO_CHUNK_SIZE]
            try:
                resp = self.mf_api("file/get_info", {"quick_key": ",".join(chunk)})
                if len(chunk) == 1:
                    file_info.append(resp["file_info"])
                else:
                    # There is an "s" when there is more than one quickkey
                    file_info.extend(resp["file_infos"])
                skipped.extend(filter(None, resp.get("skipped", "").split(",")))
            except MediafireError as err:
                if err.code != 110:
                    logger.error(f"Failed to get info for quickkeys {chunk}: {err}")
                skipped.extend(chunk)

        return file_info, skipped

    # The API call fails if even one folderkey is invalid, so we only request
    # info about one folder at a time. We don't validate folderkey because it
    # seems like custom folderkeys can be any string. We don't support the
    # `details` parameter because `get_content` is far more useful.
    def get_folder_info(self, folderkey):
        try:
            return self.mf_api("folder/get_info", {"folder_key": folderkey})[
                "folder_info"
            ]
        except MediafireError as err:
            if err.code != 112:
                logger.error(f"Failed to get info for folder {folderkey}: {err}")

    def get_folder_contents(self, folderkey, content_type="folders", owner_name=None):
        if content_type not in ["folders", "files"]:
            raise ValueError("content_type must be 'folders' or 'files'")

        data = {
            "folder_key": folderkey,
            "chunk_size": self.FOLDER_CONTENTS_CHUNK_SIZE,
            "content_type": content_type,
        }
        contents = []
        chunk = 1
        while True:
            data["chunk"] = chunk
            try:
                resp = self.mf_api("folder/get_content", data)
            except MediafireError as err:
                logger.error(
                    f"Failed to get chunk {chunk} of folder {folderkey}: {err}"
                )
                return contents
            chunk += 1

            content = resp["folder_content"]
            if owner_name:
                for c in content[content_type]:
                    # `get_content` does not return owner_name, so we hackily
                    # add it in
                    c["owner_name"] = owner_name

            contents.extend(content[content_type])
            if content["more_chunks"] == "no":
                break

        return contents

    # Download the file at the given URL (can be any URL, not just a Mediafire
    # file) to the given path
    def download_from_url(self, url, path, file_size=None):
        if file_size is None:
            # /conv/ links might redirect to ?size_id=[some number]
            head = self.s.head(url, allow_redirects=True)
            # XXX: If file_size is provided, this is probably a file that we
            # know exists (the API doesn't return anything for skipped files
            # and we don't try to download deleted files). If not, this is
            # probably a conv link, so it might not exist. This is tight
            # coupling, but I can't think of a better way right now.
            head.raise_for_status()
            if "Content-Length" in head.headers:
                file_size = int(head.headers.get("Content-Length"))
            else:
                logger.error(
                    f"No Content-Length found for {head.url} in {head.headers}"
                )
                return

        start_byte = os.path.getsize(path) if os.path.exists(path) else 0

        if file_size == 0:
            logger.warning(f"Skipping file with size of zero: {url}")
            return
        elif file_size < start_byte:
            logger.warning(
                f"Size on disk ({start_byte}) exceeds reported file size ({file_size}): {url}"
            )
            return
        elif file_size == start_byte:
            logger.debug(f"Already downloaded {url}")
            return

        desc = os.path.basename(path)
        if len(desc) > 60:
            desc = desc[:28] + "<..>" + desc[-28:]
        elif len(desc) < 60:
            desc = f"{desc:<60}"

        try:
            with open(path, "ab") as f, self.s.get(
                url, headers={"Range": f"bytes={start_byte}-"}, stream=True
            ) as r, tqdm(
                initial=start_byte,
                total=file_size,
                desc=desc,
                unit="B",
                unit_scale=True,
            ) as pbar:
                for chunk in r.iter_content(chunk_size=self.DOWNLOAD_CHUNK_SIZE):
                    f.write(chunk)
                    pbar.update(len(chunk))

                if pbar.n == 0:
                    logger.warning(
                        f"Downloaded nothing for {url=}, {start_byte=}, {file_size=}, "
                        f"{r.request.headers=}, {r.headers=}"
                    )
        except Exception as exc:
            logger.error(f"Failed to download {url}, skipping: {exc}")

    def scrape_download_page(self, url):
        extra_info = {}

        r = self.s.get(url, allow_redirects=False)
        redirects = 0
        while r.next:
            redirects += 1
            if redirects > self.MAX_REDIRECTS:
                raise Exception("Exceeded {self.MAX_REDIRECTS} redirects")

            location = r.headers["Location"]
            # Some quickkeys redirect to a direct download
            if self.DL_URL_RE.match(location):
                return location, extra_info

            r = self.s.send(r.next, allow_redirects=False)

        download_url = search_group(self.DL_URL_SCRAPE_RE, r.text)
        if download_url:
            extra_info["upload_country"] = search_group(
                self.UPLOAD_COUNTRY_SCRAPE_RE, r.text
            )
        elif self.PASSWORD_PROTECTED_RE.search(r.text):
            logger.warning(f"Password protected file: {url}")
        else:
            errno = search_group(self.ERROR_URL_RE, r.url)
            if errno == "324":
                logger.warning(f"File blocked by Google: {url}")
            elif errno == "378":
                logger.warning(f"File taken down for violating TOS: {url}")
                takedown_info = search_group(
                    self.TAKEDOWN_SCRAPE_RE, r.text, group=None
                )
                extra_info["takedown"] = dict(
                    zip(
                        ["date", "company", "email"],
                        takedown_info or [None, None, None],
                    )
                )
            elif errno == "380":
                logger.warning(f"File blocked via DCMA: {url}")
            elif errno == "386":
                logger.warning(f"File removed for violating TOS: {url}")
            elif errno == "388":
                logger.warning(f"File blocked by copyright: {url}")
            elif errno is not None:
                logger.warning(f"File not available, error {errno}: {url}")
            else:
                logger.warning(f"No download URL found: {url=}, {r.text=}")

        return download_url, extra_info

    def scrape_custom_folder(self, name):
        # Without a user agent, the <script> containing the folderkey will be
        # replaced by an "upgrade browser" popup
        r = self.s.get(
            "https://www.mediafire.com/" + name, headers={"User-Agent": self.USER_AGENT}
        )
        folderkey = self.CUSTOM_FOLDER_SCRAPE_RE.search(r.text)
        if folderkey:
            return folderkey.group(1)
        else:
            errno = search_group(self.ERROR_URL_RE, r.url)
            if errno == "370":
                logger.warning(f"Direct linking disabled for custom folder: {name}")
            elif errno is not None:
                logger.warning(f"Error {errno} for custom folder: {name}")
            else:
                logger.warning(f"Invalid custom folder: {name}")


if __name__ == "__main__":
    parser = argparse.ArgumentParser(
        description="Download files, folders, and metadata from Mediafire"
    )
    parser.add_argument(
        "input",
        help="Input JSON with file keys, folder keys, custom folder names, and conv links",
    )
    parser.add_argument("out_dir", help="Output directory")
    parser.add_argument(
        "--retry-total",
        help="Number of times to retry a request (default: %(default)s)",
        metavar="N",
        type=int,
        default=5,
    )
    parser.add_argument(
        "--retry-backoff",
        help="Retry after 0 sec, T sec, T^2 sec, etc. (default: %(default)s)",
        metavar="T",
        type=float,
        default=1.0,
    )
    parser.add_argument(
        "--log-file",
        help="Log warnings/errors to file (default: log messages are printed)",
        metavar="FILE",
    )
    parser.add_argument(
        "--metadata-only", help="Only get metadata", action="store_true"
    )
    parser.add_argument(
        "--indent", help="JSON indent (default: %(default)s)", type=int, default=None
    )
    args = parser.parse_args()

    if args.log_file:
        logging.basicConfig(filename=args.log_file)

    with open(args.input) as f:
        input_keys = json.load(f)

    os.makedirs(args.out_dir, exist_ok=True)
    mfdl = MediafireDownloader(args.retry_total, args.retry_backoff)

    # File, folders, and conv links which don't return an API response/don't exist
    skipped = []
    # Files which have a "delete_date"
    deleted = []

    def write_summary_json():
        # Not UTC or ISO 8601, but it's readable and filename-safe
        now = datetime.now().strftime("%Y-%m-%d_%H-%M-%S")
        summary_path = os.path.join(args.out_dir, f"mfdl_summary_{now}.json")
        with open(summary_path, "w") as f:
            json.dump(
                {
                    "skipped": skipped,
                    "deleted": deleted,
                },
                f,
                indent=args.indent,
            )
        print(f"Summary JSON written to: {summary_path}")

    def write_summary_handler(signal, frame):
        print("Exiting early...")
        write_summary_json()
        sys.exit(1)

    signal.signal(signal.SIGINT, write_summary_handler)

    for url in tqdm(input_keys.get("conv", []), desc="Get conv links", unit="link"):
        try:
            path = os.path.join(args.out_dir, url.split("/")[-1])
            mfdl.download_from_url(url, path)
        except requests.exceptions.HTTPError as err:
            logger.warning(f"Failed to download conv URL: {err}")
            skipped.append(url)

    folders = defaultdict(lambda: {"is_child": False, "children": []})
    folderkeys_seen = set()
    folderkeys_queue = set(input_keys.get("folder", []))

    for custom_name in tqdm(
        input_keys.get("custom_folder", []), desc="Get custom folder key"
    ):
        fk = mfdl.scrape_custom_folder(custom_name)
        if fk:
            folderkeys_queue.add(fk)
        else:
            skipped.append(custom_name)

    pbar = tqdm(desc="Get folder info", unit="dir")
    while folderkeys_queue:
        fk = folderkeys_queue.pop()
        folderkeys_seen.add(fk)

        if fk in folders:
            folder = folders[fk]
        else:
            info = mfdl.get_folder_info(fk)
            if info:
                folder = folders[fk]
                folder["info"] = info
            else:
                skipped.append(fk)
                continue

        if int(folder["info"]["folder_count"]) > 0:
            child_folders = mfdl.get_folder_contents(
                fk, content_type="folders", owner_name=folder["info"]["owner_name"]
            )
            folder["children"].extend(c["folderkey"] for c in child_folders)

            for child_info in child_folders:
                cfk = child_info["folderkey"]
                folders[cfk].update(info=child_info, is_child=True)
                if (cfk not in folderkeys_seen) and (cfk not in folderkeys_queue):
                    folderkeys_queue.add(cfk)

        pbar.update(1)

    pbar.close()

    quickkeys_seen = set()

    def process_file(info, path):
        try:
            qk = info["quickkey"]
            quickkeys_seen.add(qk)

            name = get_clean_filename(qk, info["filename"])

            if "delete_date" in info:
                deleted.append(qk)
            else:
                download_url, extra_info = mfdl.scrape_download_page(
                    info["links"]["normal_download"]
                )
                info.update(extra_info)

                if not args.metadata_only and download_url:
                    mfdl.download_from_url(
                        download_url,
                        os.path.join(path, name),
                        file_size=int(info["size"]),
                    )

            with open(os.path.join(path, name + ".json"), "w") as f:
                json.dump(info, f, indent=args.indent)
        except Exception as exc:
            # It's ugly to wrap everything in a try block, but it's more
            # important to not crash whenever possible
            logger.error(f"Failed to process file: {info=}, {path=}: {exc}")

    def process_folder(fk, path, pbar):
        folder = folders[fk]
        info = folder["info"]
        try:
            name = get_clean_foldername(fk, info["name"])
            folder_path = os.path.join(path, name)
            try_mkdir(folder_path)

            with open(os.path.join(path, name + ".json"), "w") as f:
                json.dump(info, f, indent=args.indent)

            if int(info["file_count"]) > 0:
                for child_info in mfdl.get_folder_contents(
                    fk, content_type="files", owner_name=info["owner_name"]
                ):
                    process_file(child_info, folder_path)

            for cfk in folder["children"]:
                process_folder(cfk, folder_path, pbar)

            pbar.update(1)
        except Exception as exc:
            # It's ugly to wrap everything in a try block, but it's more
            # important to not crash whenever possible
            logger.error(f"Failed to process folder: {info=}, {path=}: {exc}")

    with tqdm(desc="Process folders", total=len(folders), unit="dir") as pbar:
        for fk, folder in folders.items():
            if not folder["is_child"]:
                path = os.path.join(
                    args.out_dir, sanitize_filename(folder["info"]["owner_name"])
                )
                try_mkdir(path)
                process_folder(fk, path, pbar)

    quickkeys = list(set(input_keys.get("file", [])) - quickkeys_seen)
    file_infos, file_skipped = mfdl.get_file_info(quickkeys)
    skipped.extend(file_skipped)
    for file_info in file_infos:
        path = os.path.join(args.out_dir, sanitize_filename(file_info["owner_name"]))
        try_mkdir(path)
        process_file(file_info, path)

    write_summary_json()