| @@ -0,0 +1,250 @@ |
| 1 | +/// A minimal URL parser that extracts only what we need for git URLs. |
| 2 | +/// This is a replacement for the `url` crate dependency. |
| 3 | +#[derive(Debug)] |
| 4 | +pub(crate) struct ParsedUrl<'a> { |
| 5 | + pub scheme: String, // Owned to allow normalization to lowercase |
| 6 | + pub username: &'a str, |
| 7 | + pub password: Option<&'a str>, |
| 8 | + pub host: Option<String>, // Owned to allow normalization to lowercase |
| 9 | + pub port: Option<u16>, |
| 10 | + pub path: &'a str, |
| 11 | +} |
| 12 | + |
| 13 | +/// Minimal parse error type to replace url::ParseError |
| 14 | +#[derive(Debug, Clone, PartialEq, Eq, thiserror::Error)] |
| 15 | +#[allow(missing_docs)] |
| 16 | +pub enum UrlParseError { |
| 17 | + #[error("relative URL without a base")] |
| 18 | + RelativeUrlWithoutBase, |
| 19 | + #[error("invalid port number - must be between 1-65535")] |
| 20 | + InvalidPort, |
| 21 | + #[error("invalid domain character")] |
| 22 | + InvalidDomainCharacter, |
| 23 | + #[error("Scheme requires host")] |
| 24 | + SchemeRequiresHost, |
| 25 | +} |
| 26 | + |
| 27 | +/// Check if a character is valid in a URL scheme. |
| 28 | +/// Valid scheme characters: alphanumeric, +, -, or . |
| 29 | +fn is_valid_scheme_char(c: char) -> bool { |
| 30 | + c.is_ascii_alphanumeric() || c == '+' || c == '-' || c == '.' |
| 31 | +} |
| 32 | + |
| 33 | +impl<'a> ParsedUrl<'a> { |
| 34 | + /// Parse a URL string into its components. |
| 35 | + /// Expected format: scheme://[user[:password]@]host[:port]/path |
| 36 | + pub(crate) fn parse(input: &'a str) -> Result<Self, UrlParseError> { |
| 37 | + // Find scheme by looking for first ':' |
| 38 | + let first_colon = input.find(':').ok_or(UrlParseError::RelativeUrlWithoutBase)?; |
| 39 | + let scheme_str = &input[..first_colon]; |
| 40 | + // Normalize scheme to lowercase for case-insensitive matching (matches url crate behavior) |
| 41 | + let scheme = scheme_str.to_ascii_lowercase(); |
| 42 | + let Some(after_scheme) = input[first_colon..].strip_prefix("://") else { |
| 43 | + return Err(UrlParseError::RelativeUrlWithoutBase); |
| 44 | + }; |
| 45 | + |
| 46 | + // Check for relative URL (scheme without proper authority) |
| 47 | + if scheme_str.is_empty() { |
| 48 | + return Err(UrlParseError::RelativeUrlWithoutBase); |
| 49 | + } |
| 50 | + |
| 51 | + // Validate scheme characters (check original before lowercase conversion) |
| 52 | + if !scheme_str.chars().all(is_valid_scheme_char) { |
| 53 | + return Err(UrlParseError::RelativeUrlWithoutBase); |
| 54 | + } |
| 55 | + |
| 56 | + // Find path start (first '/' after scheme) |
| 57 | + let path_start = after_scheme.find('/').unwrap_or(after_scheme.len()); |
| 58 | + let authority = &after_scheme[..path_start]; |
| 59 | + let path = if path_start < after_scheme.len() { |
| 60 | + &after_scheme[path_start..] |
| 61 | + } else { |
| 62 | + // No path specified - leave empty (caller can default to / if needed) |
| 63 | + "" |
| 64 | + }; |
| 65 | + |
| 66 | + // Parse authority: [user[:password]@]host[:port] |
| 67 | + let (username, password, host, port) = if let Some(at_pos) = authority.rfind('@') { |
| 68 | + // Has user info |
| 69 | + let user_info = &authority[..at_pos]; |
| 70 | + let host_port = &authority[at_pos + 1..]; |
| 71 | + |
| 72 | + let (user, pass) = if let Some(colon_pos) = user_info.find(':') { |
| 73 | + let pass_str = &user_info[colon_pos + 1..]; |
| 74 | + // Treat empty password as None |
| 75 | + let pass = if pass_str.is_empty() { None } else { Some(pass_str) }; |
| 76 | + (&user_info[..colon_pos], pass) |
| 77 | + } else { |
| 78 | + (user_info, None) |
| 79 | + }; |
| 80 | + |
| 81 | + let (h, p) = Self::parse_host_port(host_port)?; |
| 82 | + // If we have user info, we must have a host |
| 83 | + if h.is_none() { |
| 84 | + return Err(UrlParseError::InvalidDomainCharacter); |
| 85 | + } |
| 86 | + (user, pass, h, p) |
| 87 | + } else { |
| 88 | + // No user info |
| 89 | + let (h, p) = Self::parse_host_port(authority)?; |
| 90 | + ("", None, h, p) |
| 91 | + }; |
| 92 | + |
| 93 | + // Standard schemes (http, https, git, ssh) require a host |
| 94 | + // Scheme is already lowercase at this point |
| 95 | + let requires_host = matches!(scheme.as_str(), "http" | "https" | "git" | "ssh" | "ftp" | "ftps"); |
| 96 | + if requires_host && host.is_none() { |
| 97 | + return Err(UrlParseError::SchemeRequiresHost); |
| 98 | + } |
| 99 | + |
| 100 | + Ok(ParsedUrl { |
| 101 | + scheme, |
| 102 | + username, |
| 103 | + password, |
| 104 | + host, |
| 105 | + port, |
| 106 | + path, |
| 107 | + }) |
| 108 | + } |
| 109 | + |
| 110 | + fn parse_host_port(host_port: &str) -> Result<(Option<String>, Option<u16>), UrlParseError> { |
| 111 | + if host_port.is_empty() { |
| 112 | + return Ok((None, None)); |
| 113 | + } |
| 114 | + |
| 115 | + // Handle IPv6 addresses: [::1] or [::1]:port |
| 116 | + if host_port.starts_with('[') { |
| 117 | + if let Some(bracket_end) = host_port.find(']') { |
| 118 | + // IPv6 addresses are case-insensitive, normalize to lowercase |
| 119 | + let host = Some(host_port[..=bracket_end].to_ascii_lowercase()); |
| 120 | + let remaining = &host_port[bracket_end + 1..]; |
| 121 | + |
| 122 | + if remaining.is_empty() { |
| 123 | + return Ok((host, None)); |
| 124 | + } else if let Some(port_str) = remaining.strip_prefix(':') { |
| 125 | + let port = port_str.parse::<u16>().map_err(|_| UrlParseError::InvalidPort)?; |
| 126 | + // Validate port is in valid range (1-65535, port 0 is invalid) |
| 127 | + if port == 0 { |
| 128 | + return Err(UrlParseError::InvalidPort); |
| 129 | + } |
| 130 | + return Ok((host, Some(port))); |
| 131 | + } else { |
| 132 | + return Err(UrlParseError::InvalidDomainCharacter); |
| 133 | + } |
| 134 | + } else { |
| 135 | + return Err(UrlParseError::InvalidDomainCharacter); |
| 136 | + } |
| 137 | + } |
| 138 | + |
| 139 | + // Handle regular host:port |
| 140 | + // Use rfind to handle IPv6 addresses without brackets (edge case) |
| 141 | + if let Some(colon_pos) = host_port.rfind(':') { |
| 142 | + // Check if this looks like a port (all digits after colon) |
| 143 | + let potential_port = &host_port[colon_pos + 1..]; |
| 144 | + if potential_port.is_empty() { |
| 145 | + // Empty port like "host:" - strip the trailing colon |
| 146 | + let host_str = &host_port[..colon_pos]; |
| 147 | + return Ok((Some(Self::normalize_hostname(host_str)?), None)); |
| 148 | + } else if potential_port.chars().all(|c| c.is_ascii_digit()) { |
| 149 | + let host_str = &host_port[..colon_pos]; |
| 150 | + let host = Self::normalize_hostname(host_str)?; |
| 151 | + let port = potential_port.parse::<u16>().map_err(|_| UrlParseError::InvalidPort)?; |
| 152 | + // Validate port is in valid range (1-65535, port 0 is invalid) |
| 153 | + if port == 0 { |
| 154 | + return Err(UrlParseError::InvalidPort); |
| 155 | + } |
| 156 | + return Ok((Some(host), Some(port))); |
| 157 | + } |
| 158 | + } |
| 159 | + |
| 160 | + // No port, just host |
| 161 | + Ok((Some(Self::normalize_hostname(host_port)?), None)) |
| 162 | + } |
| 163 | + |
| 164 | + /// Check if a string looks like a valid DNS hostname (for normalization purposes) |
| 165 | + /// Valid DNS names contain only alphanumeric, hyphens, dots, underscores, and wildcards |
| 166 | + fn is_normalizable_hostname(host: &str) -> bool { |
| 167 | + // Allow alphanumeric, -, ., _, and * (for patterns) |
| 168 | + host.chars() |
| 169 | + .all(|c| c.is_ascii_alphanumeric() || matches!(c, '-' | '.' | '_' | '*')) |
| 170 | + } |
| 171 | + |
| 172 | + /// Validate and possibly normalize a hostname |
| 173 | + /// Valid DNS hostnames are normalized to lowercase |
| 174 | + /// Invalid strings (like injection attempts) are preserved as-is but ? is rejected |
| 175 | + fn normalize_hostname(host: &str) -> Result<String, UrlParseError> { |
| 176 | + // Reject ? character which git's url parser rejects |
| 177 | + if host.contains('?') { |
| 178 | + return Err(UrlParseError::InvalidDomainCharacter); |
| 179 | + } |
| 180 | + |
| 181 | + // Only normalize if it looks like a valid DNS hostname |
| 182 | + // Preserve case for security checks if it contains special characters |
| 183 | + if Self::is_normalizable_hostname(host) { |
| 184 | + Ok(host.to_ascii_lowercase()) |
| 185 | + } else { |
| 186 | + Ok(host.to_owned()) |
| 187 | + } |
| 188 | + } |
| 189 | +} |
| 190 | + |
| 191 | +#[cfg(test)] |
| 192 | +mod tests { |
| 193 | + use super::*; |
| 194 | + |
| 195 | + #[test] |
| 196 | + fn test_simple_url() { |
| 197 | + let url = ParsedUrl::parse("http://example.com/path").unwrap(); |
| 198 | + assert_eq!(url.scheme, "http"); |
| 199 | + assert_eq!(url.host.as_deref(), Some("example.com")); |
| 200 | + assert_eq!(url.path, "/path"); |
| 201 | + assert_eq!(url.username, ""); |
| 202 | + assert_eq!(url.password, None); |
| 203 | + assert_eq!(url.port, None); |
| 204 | + } |
| 205 | + |
| 206 | + #[test] |
| 207 | + fn test_url_with_port() { |
| 208 | + let url = ParsedUrl::parse("http://example.com:8080/path").unwrap(); |
| 209 | + assert_eq!(url.scheme, "http"); |
| 210 | + assert_eq!(url.host.as_deref(), Some("example.com")); |
| 211 | + assert_eq!(url.port, Some(8080)); |
| 212 | + assert_eq!(url.path, "/path"); |
| 213 | + } |
| 214 | + |
| 215 | + #[test] |
| 216 | + fn test_url_with_user() { |
| 217 | + let url = ParsedUrl::parse("http://user@example.com/path").unwrap(); |
| 218 | + assert_eq!(url.scheme, "http"); |
| 219 | + assert_eq!(url.username, "user"); |
| 220 | + assert_eq!(url.host.as_deref(), Some("example.com")); |
| 221 | + assert_eq!(url.path, "/path"); |
| 222 | + } |
| 223 | + |
| 224 | + #[test] |
| 225 | + fn test_url_with_user_and_password() { |
| 226 | + let url = ParsedUrl::parse("http://user:pass@example.com/path").unwrap(); |
| 227 | + assert_eq!(url.scheme, "http"); |
| 228 | + assert_eq!(url.username, "user"); |
| 229 | + assert_eq!(url.password, Some("pass")); |
| 230 | + assert_eq!(url.host.as_deref(), Some("example.com")); |
| 231 | + assert_eq!(url.path, "/path"); |
| 232 | + } |
| 233 | + |
| 234 | + #[test] |
| 235 | + fn test_url_with_ipv6() { |
| 236 | + let url = ParsedUrl::parse("http://[::1]/path").unwrap(); |
| 237 | + assert_eq!(url.scheme, "http"); |
| 238 | + assert_eq!(url.host.as_deref(), Some("[::1]")); |
| 239 | + assert_eq!(url.path, "/path"); |
| 240 | + } |
| 241 | + |
| 242 | + #[test] |
| 243 | + fn test_url_with_ipv6_and_port() { |
| 244 | + let url = ParsedUrl::parse("http://[::1]:8080/path").unwrap(); |
| 245 | + assert_eq!(url.scheme, "http"); |
| 246 | + assert_eq!(url.host.as_deref(), Some("[::1]")); |
| 247 | + assert_eq!(url.port, Some(8080)); |
| 248 | + assert_eq!(url.path, "/path"); |
| 249 | + } |
| 250 | +} |