|
| 1 | +from __future__ import annotations |
| 2 | + |
| 3 | +import dataclasses |
| 4 | +import urllib.parse |
| 5 | +import urllib.request |
| 6 | + |
| 7 | +from .datastructures import Headers |
| 8 | +from .exceptions import InvalidProxy |
| 9 | +from .headers import build_authorization_basic, build_host |
| 10 | +from .http11 import USER_AGENT |
| 11 | +from .uri import DELIMS, WebSocketURI |
| 12 | + |
| 13 | + |
| 14 | +__all__ = ["get_proxy", "parse_proxy", "Proxy"] |
| 15 | + |
| 16 | + |
| 17 | +@dataclasses.dataclass |
| 18 | +class Proxy: |
| 19 | + """ |
| 20 | + Proxy address. |
| 21 | +
|
| 22 | + Attributes: |
| 23 | + scheme: ``"socks5h"``, ``"socks5"``, ``"socks4a"``, ``"socks4"``, |
| 24 | + ``"https"``, or ``"http"``. |
| 25 | + host: Normalized to lower case. |
| 26 | + port: Always set even if it's the default. |
| 27 | + username: Available when the proxy address contains `User Information`_. |
| 28 | + password: Available when the proxy address contains `User Information`_. |
| 29 | +
|
| 30 | + .. _User Information: https://datatracker.ietf.org/doc/html/rfc3986#section-3.2.1 |
| 31 | +
|
| 32 | + """ |
| 33 | + |
| 34 | + scheme: str |
| 35 | + host: str |
| 36 | + port: int |
| 37 | + username: str | None = None |
| 38 | + password: str | None = None |
| 39 | + |
| 40 | + @property |
| 41 | + def user_info(self) -> tuple[str, str] | None: |
| 42 | + if self.username is None: |
| 43 | + return None |
| 44 | + assert self.password is not None |
| 45 | + return (self.username, self.password) |
| 46 | + |
| 47 | + |
| 48 | +def parse_proxy(proxy: str) -> Proxy: |
| 49 | + """ |
| 50 | + Parse and validate a proxy. |
| 51 | +
|
| 52 | + Args: |
| 53 | + proxy: proxy. |
| 54 | +
|
| 55 | + Returns: |
| 56 | + Parsed proxy. |
| 57 | +
|
| 58 | + Raises: |
| 59 | + InvalidProxy: If ``proxy`` isn't a valid proxy. |
| 60 | +
|
| 61 | + """ |
| 62 | + parsed = urllib.parse.urlparse(proxy) |
| 63 | + if parsed.scheme not in ["socks5h", "socks5", "socks4a", "socks4", "https", "http"]: |
| 64 | + raise InvalidProxy(proxy, f"scheme {parsed.scheme} isn't supported") |
| 65 | + if parsed.hostname is None: |
| 66 | + raise InvalidProxy(proxy, "hostname isn't provided") |
| 67 | + if parsed.path not in ["", "/"]: |
| 68 | + raise InvalidProxy(proxy, "path is meaningless") |
| 69 | + if parsed.query != "": |
| 70 | + raise InvalidProxy(proxy, "query is meaningless") |
| 71 | + if parsed.fragment != "": |
| 72 | + raise InvalidProxy(proxy, "fragment is meaningless") |
| 73 | + |
| 74 | + scheme = parsed.scheme |
| 75 | + host = parsed.hostname |
| 76 | + port = parsed.port or (443 if parsed.scheme == "https" else 80) |
| 77 | + username = parsed.username |
| 78 | + password = parsed.password |
| 79 | + # urllib.parse.urlparse accepts URLs with a username but without a |
| 80 | + # password. This doesn't make sense for HTTP Basic Auth credentials. |
| 81 | + if username is not None and password is None: |
| 82 | + raise InvalidProxy(proxy, "username provided without password") |
| 83 | + |
| 84 | + try: |
| 85 | + proxy.encode("ascii") |
| 86 | + except UnicodeEncodeError: |
| 87 | + # Input contains non-ASCII characters. |
| 88 | + # It must be an IRI. Convert it to a URI. |
| 89 | + host = host.encode("idna").decode() |
| 90 | + if username is not None: |
| 91 | + assert password is not None |
| 92 | + username = urllib.parse.quote(username, safe=DELIMS) |
| 93 | + password = urllib.parse.quote(password, safe=DELIMS) |
| 94 | + |
| 95 | + return Proxy(scheme, host, port, username, password) |
| 96 | + |
| 97 | + |
| 98 | +def get_proxy(uri: WebSocketURI) -> str | None: |
| 99 | + """ |
| 100 | + Return the proxy to use for connecting to the given WebSocket URI, if any. |
| 101 | +
|
| 102 | + """ |
| 103 | + if urllib.request.proxy_bypass(f"{uri.host}:{uri.port}"): |
| 104 | + return None |
| 105 | + |
| 106 | + # According to the _Proxy Usage_ section of RFC 6455, use a SOCKS5 proxy if |
| 107 | + # available, else favor the proxy for HTTPS connections over the proxy for |
| 108 | + # HTTP connections. |
| 109 | + |
| 110 | + # The priority of a proxy for WebSocket connections is unspecified. We give |
| 111 | + # it the highest priority. This makes it easy to configure a specific proxy |
| 112 | + # for websockets. |
| 113 | + |
| 114 | + # getproxies() may return SOCKS proxies as {"socks": "http://host:port"} or |
| 115 | + # as {"https": "socks5h://host:port"} depending on whether they're declared |
| 116 | + # in the operating system or in environment variables. |
| 117 | + |
| 118 | + proxies = urllib.request.getproxies() |
| 119 | + if uri.secure: |
| 120 | + schemes = ["wss", "socks", "https"] |
| 121 | + else: |
| 122 | + schemes = ["ws", "socks", "https", "http"] |
| 123 | + |
| 124 | + for scheme in schemes: |
| 125 | + proxy = proxies.get(scheme) |
| 126 | + if proxy is not None: |
| 127 | + if scheme == "socks" and proxy.startswith("http://"): |
| 128 | + proxy = "socks5h://" + proxy[7:] |
| 129 | + return proxy |
| 130 | + else: |
| 131 | + return None |
| 132 | + |
| 133 | + |
| 134 | +def prepare_connect_request( |
| 135 | + proxy: Proxy, |
| 136 | + ws_uri: WebSocketURI, |
| 137 | + user_agent_header: str | None = USER_AGENT, |
| 138 | +) -> bytes: |
| 139 | + host = build_host(ws_uri.host, ws_uri.port, ws_uri.secure, always_include_port=True) |
| 140 | + headers = Headers() |
| 141 | + headers["Host"] = build_host(ws_uri.host, ws_uri.port, ws_uri.secure) |
| 142 | + if user_agent_header is not None: |
| 143 | + headers["User-Agent"] = user_agent_header |
| 144 | + if proxy.username is not None: |
| 145 | + assert proxy.password is not None # enforced by parse_proxy() |
| 146 | + headers["Proxy-Authorization"] = build_authorization_basic( |
| 147 | + proxy.username, proxy.password |
| 148 | + ) |
| 149 | + # We cannot use the Request class because it supports only GET requests. |
| 150 | + return f"CONNECT {host} HTTP/1.1\r\n".encode() + headers.serialize() |
0 commit comments