hpx 2.4.9

#[cfg(all(target_os = "macos", feature = "system-proxy"))]
mod mac;
#[cfg(unix)]
mod uds;
#[cfg(all(windows, feature = "system-proxy"))]
mod win;

pub(crate) mod matcher;

use std::hash::{Hash, Hasher};
#[cfg(unix)]
use std::{path::Path, sync::Arc};

use http::{HeaderMap, Uri, header::HeaderValue};

use crate::{IntoUri, ext::UriExt};

// # Internals
//
// This module is a couple pieces:
//
// - The public builder API
// - The internal built types that our Connector knows how to use.
//
// The user creates a builder (`hpx::Proxy`), and configures any extras.
// Once that type is passed to the `ClientBuilder`, we convert it into the
// built matcher types, making use of `core`'s matchers.

/// Configuration of a proxy that a `Client` should pass requests to.
///
/// A `Proxy` has a couple pieces to it:
///
/// - a URI of how to talk to the proxy
/// - rules on what `Client` requests should be directed to the proxy
///
/// For instance, let's look at `Proxy::http`:
///
/// ```rust
/// # fn run() -> Result<(), Box<dyn std::error::Error>> {
/// let proxy = hpx::Proxy::http("https://secure.example")?;
/// # Ok(())
/// # }
/// ```
///
/// This proxy will intercept all HTTP requests, and make use of the proxy
/// at `https://secure.example`. A request to `http://hyper.rs` will talk
/// to your proxy. A request to `https://hyper.rs` will not.
///
/// Multiple `Proxy` rules can be configured for a `Client`. The `Client` will
/// check each `Proxy` in the order it was added. This could mean that a
/// `Proxy` added first with eager intercept rules, such as `Proxy::all`,
/// would prevent a `Proxy` later in the list from ever working, so take care.
///
/// By enabling the `"socks"` feature it is possible to use a socks proxy:
/// ```rust
/// # fn run() -> Result<(), Box<dyn std::error::Error>> {
/// let proxy = hpx::Proxy::http("socks5://192.168.1.1:9000")?;
/// # Ok(())
/// # }
/// ```
#[derive(Clone, Debug)]
pub struct Proxy {
    extra: Extra,
    scheme: ProxyScheme,
    no_proxy: Option<NoProxy>,
}

/// A configuration for filtering out requests that shouldn't be proxied
#[derive(Clone, Debug, Default)]
pub struct NoProxy {
    inner: String,
}

// ===== Internal =====

#[allow(clippy::large_enum_variant)]
#[derive(Clone, PartialEq, Eq)]
pub(crate) enum Intercepted {
    Proxy(matcher::Intercept),
    #[cfg(unix)]
    Unix(Arc<Path>),
}

#[derive(Debug, Clone, PartialEq, Eq, Hash)]
pub(crate) struct Matcher {
    inner: Box<matcher::Matcher>,
}

#[derive(Clone, Debug)]
enum ProxyScheme {
    All(Uri),
    Http(Uri),
    Https(Uri),
    #[cfg(unix)]
    Unix(Arc<Path>),
}

#[derive(Debug, Clone, Default, PartialEq, Eq)]
struct Extra {
    auth: Option<HeaderValue>,
    misc: Option<HeaderMap>,
}

// ===== impl Proxy =====

impl Proxy {
    /// Proxy all HTTP traffic to the passed URI.
    ///
    /// # Example
    ///
    /// ```
    /// # extern crate hpx;
    /// # fn run() -> Result<(), Box<dyn std::error::Error>> {
    /// let client = hpx::Client::builder()
    ///     .proxy(hpx::Proxy::http("https://my.prox")?)
    ///     .build()?;
    /// # Ok(())
    /// # }
    /// # fn main() {}
    /// ```
    pub fn http<U: IntoUri>(uri: U) -> crate::Result<Proxy> {
        uri.into_uri().map(ProxyScheme::Http).map(Proxy::new)
    }

    /// Proxy all HTTPS traffic to the passed URI.
    ///
    /// # Example
    ///
    /// ```
    /// # extern crate hpx;
    /// # fn run() -> Result<(), Box<dyn std::error::Error>> {
    /// let client = hpx::Client::builder()
    ///     .proxy(hpx::Proxy::https("https://example.prox:4545")?)
    ///     .build()?;
    /// # Ok(())
    /// # }
    /// # fn main() {}
    /// ```
    pub fn https<U: IntoUri>(uri: U) -> crate::Result<Proxy> {
        uri.into_uri().map(ProxyScheme::Https).map(Proxy::new)
    }

    /// Proxy **all** traffic to the passed URI.
    ///
    /// "All" refers to `https` and `http` URIs. Other schemes are not
    /// recognized by hpx.
    ///
    /// # Example
    ///
    /// ```
    /// # extern crate hpx;
    /// # fn run() -> Result<(), Box<dyn std::error::Error>> {
    /// let client = hpx::Client::builder()
    ///     .proxy(hpx::Proxy::all("http://pro.xy")?)
    ///     .build()?;
    /// # Ok(())
    /// # }
    /// # fn main() {}
    /// ```
    pub fn all<U: IntoUri>(uri: U) -> crate::Result<Proxy> {
        uri.into_uri().map(ProxyScheme::All).map(Proxy::new)
    }

    /// Proxy all traffic to the passed Unix Domain Socket path.
    ///
    /// # Example
    ///
    /// ```
    /// # extern crate hpx;
    /// # fn run() -> Result<(), Box<dyn std::error::Error>> {
    /// let client = hpx::Client::builder()
    ///     .proxy(hpx::Proxy::unix("/var/run/docker.sock")?)
    ///     .build()?;
    /// # Ok(())
    /// # }
    /// # fn main() {}
    /// ```
    #[cfg(unix)]
    pub fn unix<P: uds::IntoUnixSocket>(unix: P) -> crate::Result<Proxy> {
        Ok(Proxy::new(ProxyScheme::Unix(unix.unix_socket())))
    }

    fn new(scheme: ProxyScheme) -> Proxy {
        Proxy {
            extra: Extra {
                auth: None,
                misc: None,
            },
            scheme,
            no_proxy: None,
        }
    }

    /// Set the `Proxy-Authorization` header using Basic auth.
    ///
    /// # Example
    ///
    /// ```
    /// # extern crate hpx;
    /// # fn run() -> Result<(), Box<dyn std::error::Error>> {
    /// let proxy = hpx::Proxy::https("http://localhost:1234")?.basic_auth("Aladdin", "open sesame");
    /// # Ok(())
    /// # }
    /// # fn main() {}
    /// ```
    pub fn basic_auth(mut self, username: &str, password: &str) -> Proxy {
        match self.scheme {
            ProxyScheme::All(ref mut uri)
            | ProxyScheme::Http(ref mut uri)
            | ProxyScheme::Https(ref mut uri) => {
                let header = crate::util::basic_auth(username, Some(password));
                uri.set_userinfo(username, Some(password));
                self.extra.auth = Some(header);
            }
            #[cfg(unix)]
            ProxyScheme::Unix(_) => {
                // For Unix sockets, we don't set the auth header.
                // This is a no-op, but keeps the API consistent.
            }
        }

        self
    }

    /// Set the `Proxy-Authorization` header to a specified value.
    ///
    /// # Example
    ///
    /// ```
    /// # extern crate hpx;
    /// # use hpx::header::*;
    /// # fn run() -> Result<(), Box<dyn std::error::Error>> {
    /// let proxy = hpx::Proxy::https("http://localhost:1234")?
    ///     .custom_http_auth(HeaderValue::from_static("justletmeinalreadyplease"));
    /// # Ok(())
    /// # }
    /// # fn main() {}
    /// ```
    pub fn custom_http_auth(mut self, header_value: HeaderValue) -> Proxy {
        self.extra.auth = Some(header_value);
        self
    }

    /// Adds a Custom Headers to Proxy
    /// Adds custom headers to this Proxy
    ///
    /// # Example
    /// ```
    /// # extern crate hpx;
    /// # use hpx::header::*;
    /// # fn run() -> Result<(), Box<dyn std::error::Error>> {
    /// let mut headers = HeaderMap::new();
    /// headers.insert(USER_AGENT, "hpx".parse().unwrap());
    /// let proxy = hpx::Proxy::https("http://localhost:1234")?.custom_http_headers(headers);
    /// # Ok(())
    /// # }
    /// # fn main() {}
    /// ```
    pub fn custom_http_headers(mut self, headers: HeaderMap) -> Proxy {
        match self.scheme {
            ProxyScheme::All(_) | ProxyScheme::Http(_) | ProxyScheme::Https(_) => {
                self.extra.misc = Some(headers);
            }
            #[cfg(unix)]
            ProxyScheme::Unix(_) => {
                // For Unix sockets, we don't set custom headers.
                // This is a no-op, but keeps the API consistent.
            }
        }

        self
    }

    /// Adds a `No Proxy` exclusion list to this Proxy
    ///
    /// # Example
    ///
    /// ```
    /// # extern crate hpx;
    /// # fn run() -> Result<(), Box<dyn std::error::Error>> {
    /// let proxy = hpx::Proxy::https("http://localhost:1234")?
    ///     .no_proxy(hpx::NoProxy::from_string("direct.tld, sub.direct2.tld"));
    /// # Ok(())
    /// # }
    /// # fn main() {}
    /// ```
    pub fn no_proxy(mut self, no_proxy: Option<NoProxy>) -> Proxy {
        self.no_proxy = no_proxy;
        self
    }

    pub(crate) fn into_matcher(self) -> Matcher {
        let Proxy {
            scheme,
            extra,
            no_proxy,
        } = self;

        let no_proxy = no_proxy.as_ref().map_or("", |n| n.inner.as_ref());

        let inner = match scheme {
            ProxyScheme::All(uri) => matcher::Matcher::builder()
                .all(uri.to_string())
                .no(no_proxy)
                .build(extra),
            ProxyScheme::Http(uri) => matcher::Matcher::builder()
                .http(uri.to_string())
                .no(no_proxy)
                .build(extra),
            ProxyScheme::Https(uri) => matcher::Matcher::builder()
                .https(uri.to_string())
                .no(no_proxy)
                .build(extra),
            #[cfg(unix)]
            ProxyScheme::Unix(unix) => matcher::Matcher::builder()
                .unix(unix)
                .no(no_proxy)
                .build(extra),
        };

        Matcher {
            inner: Box::new(inner),
        }
    }
}

// ===== impl NoProxy =====

impl NoProxy {
    /// Returns a new no-proxy configuration based on environment variables (or `None` if no
    /// variables are set) see [self::NoProxy::from_string()] for the string format
    pub fn from_env() -> Option<NoProxy> {
        let raw = std::env::var("NO_PROXY")
            .or_else(|_| std::env::var("no_proxy"))
            .ok()?;

        // Per the docs, this returns `None` if no environment variable is set. We can only reach
        // here if an env var is set, so we return `Some(NoProxy::default)` if `from_string`
        // returns None, which occurs with an empty string.
        Some(Self::from_string(&raw).unwrap_or_default())
    }

    /// Returns a new no-proxy configuration based on a `no_proxy` string (or `None` if no variables
    /// are set)
    /// The rules are as follows:
    /// * The environment variable `NO_PROXY` is checked, if it is not set, `no_proxy` is checked
    /// * If neither environment variable is set, `None` is returned
    /// * Entries are expected to be comma-separated (whitespace between entries is ignored)
    /// * IP addresses (both IPv4 and IPv6) are allowed, as are optional subnet masks (by adding
    ///   /size, for example "`192.168.1.0/24`").
    /// * An entry "`*`" matches all hostnames (this is the only wildcard allowed)
    /// * Any other entry is considered a domain name (and may contain a leading dot, for example
    ///   `google.com` and `.google.com` are equivalent) and would match both that domain AND all
    ///   subdomains.
    ///
    /// For example, if `"NO_PROXY=google.com, 192.168.1.0/24"` was set, all the following would
    /// match (and therefore would bypass the proxy):
    /// * `http://google.com/`
    /// * `http://www.google.com/`
    /// * `http://192.168.1.42/`
    ///
    /// The URI `http://notgoogle.com/` would not match.
    pub fn from_string(no_proxy_list: &str) -> Option<Self> {
        Some(NoProxy {
            inner: no_proxy_list.into(),
        })
    }
}

// ===== impl Matcher =====

impl Matcher {
    pub(crate) fn system() -> Self {
        Self {
            inner: Box::new(matcher::Matcher::from_system()),
        }
    }

    /// Intercept the given destination URI, returning the intercepted
    /// proxy configuration if there is a match.
    #[inline]
    pub(crate) fn intercept(&self, dst: &Uri) -> Option<Intercepted> {
        self.inner.intercept(dst)
    }
}

// ===== impl Extra =====

impl Hash for Extra {
    fn hash<H: Hasher>(&self, state: &mut H) {
        self.auth.hash(state);
        if let Some(ref misc) = self.misc {
            for (k, v) in misc.iter() {
                k.as_str().hash(state);
                v.as_bytes().hash(state);
            }
        } else {
            1u8.hash(state);
        }
    }
}

#[cfg(test)]
mod tests {
    use super::*;

    fn uri(s: &str) -> Uri {
        s.parse().unwrap()
    }

    fn intercept(p: &Matcher, s: &Uri) -> matcher::Intercept {
        match p.intercept(s).unwrap() {
            Intercepted::Proxy(proxy) => proxy,
            _ => {
                unreachable!("intercepted_port should only be called with a Proxy matcher")
            }
        }
    }

    fn intercepted_uri(p: &Matcher, s: &str) -> Uri {
        match p.intercept(&s.parse().unwrap()).unwrap() {
            Intercepted::Proxy(proxy) => proxy.uri().clone(),
            _ => {
                unreachable!("intercepted_uri should only be called with a Proxy matcher")
            }
        }
    }

    #[test]
    fn test_http() {
        let target = "http://example.domain/";
        let p = Proxy::http(target).unwrap().into_matcher();

        let http = "http://hyper.rs";
        let other = "https://hyper.rs";

        assert_eq!(intercepted_uri(&p, http), target);
        assert!(p.intercept(&uri(other)).is_none());
    }

    #[test]
    fn test_https() {
        let target = "http://example.domain/";
        let p = Proxy::https(target).unwrap().into_matcher();

        let http = "http://hyper.rs";
        let other = "https://hyper.rs";

        assert!(p.intercept(&uri(http)).is_none());
        assert_eq!(intercepted_uri(&p, other), target);
    }

    #[test]
    fn test_all() {
        let target = "http://example.domain/";
        let p = Proxy::all(target).unwrap().into_matcher();

        let http = "http://hyper.rs";
        let https = "https://hyper.rs";
        // no longer supported
        // let other = "x-youve-never-heard-of-me-mr-proxy://hyper.rs";

        assert_eq!(intercepted_uri(&p, http), target);
        assert_eq!(intercepted_uri(&p, https), target);
        // assert_eq!(intercepted_uri(&p, other), target);
    }

    #[test]
    fn test_standard_with_custom_auth_header() {
        let target = "http://example.domain/";
        let p = Proxy::all(target)
            .unwrap()
            .custom_http_auth(http::HeaderValue::from_static("testme"))
            .into_matcher();

        let got = intercept(&p, &uri("http://anywhere.local"));
        let auth = got.basic_auth().unwrap();
        assert_eq!(auth, "testme");
    }

    #[test]
    fn test_maybe_has_http_auth() {
        let uri = uri("http://example.domain/");

        let m = Proxy::all("https://letme:in@yo.local")
            .unwrap()
            .into_matcher();

        let got = intercept(&m, &uri);
        assert!(got.basic_auth().is_some(), "https forwards");

        let m = Proxy::all("http://letme:in@yo.local")
            .unwrap()
            .into_matcher();

        let got = intercept(&m, &uri);
        assert!(got.basic_auth().is_some(), "http forwards");
    }

    #[test]
    fn test_maybe_has_http_custom_headers() {
        let uri = uri("http://example.domain/");

        let mut headers = HeaderMap::new();
        headers.insert("x-custom-header", HeaderValue::from_static("custom-value"));

        let m = Proxy::all("https://yo.local")
            .unwrap()
            .custom_http_headers(headers.clone())
            .into_matcher();

        match m.intercept(&uri).unwrap() {
            Intercepted::Proxy(proxy) => {
                let got_headers = proxy.custom_headers().unwrap();
                assert_eq!(got_headers, &headers, "https forwards");
            }
            _ => {
                unreachable!("Expected a Proxy Intercepted");
            }
        }

        let m = Proxy::all("http://yo.local")
            .unwrap()
            .custom_http_headers(headers.clone())
            .into_matcher();

        match m.intercept(&uri).unwrap() {
            Intercepted::Proxy(proxy) => {
                let got_headers = proxy.custom_headers().unwrap();
                assert_eq!(got_headers, &headers, "http forwards");
            }
            _ => {
                unreachable!("Expected a Proxy Intercepted");
            }
        }
    }

    fn test_socks_proxy_default_port(uri: &str, url2: &str, port: u16) {
        let m = Proxy::all(uri).unwrap().into_matcher();

        let http = "http://hyper.rs";
        let https = "https://hyper.rs";

        assert_eq!(intercepted_uri(&m, http).port_u16(), Some(1080));
        assert_eq!(intercepted_uri(&m, https).port_u16(), Some(1080));

        // custom port
        let m = Proxy::all(url2).unwrap().into_matcher();

        assert_eq!(intercepted_uri(&m, http).port_u16(), Some(port));
        assert_eq!(intercepted_uri(&m, https).port_u16(), Some(port));
    }

    #[test]
    fn test_socks4_proxy_default_port() {
        test_socks_proxy_default_port("socks4://example.com", "socks4://example.com:1234", 1234);
        test_socks_proxy_default_port("socks4a://example.com", "socks4a://example.com:1234", 1234);
    }

    #[test]
    fn test_socks5_proxy_default_port() {
        test_socks_proxy_default_port("socks5://example.com", "socks5://example.com:1234", 1234);
        test_socks_proxy_default_port("socks5h://example.com", "socks5h://example.com:1234", 1234);
    }
}