diff options
author | Andreas Grois <andi@grois.info> | 2022-10-09 14:00:38 +0200 |
---|---|---|
committer | Andreas Grois <andi@grois.info> | 2022-10-09 14:00:38 +0200 |
commit | d6d345207530ec3232d937aeee3b0c9255b33129 (patch) | |
tree | 0645a54d0f695092ed6c0003777d2e212a99e5e5 /src/url_parsing/mod.rs | |
parent | 5e51b706d54a26470f33d1342f4666d5aab921fc (diff) |
Also add url_parsing to this crate.
Diffstat (limited to 'src/url_parsing/mod.rs')
-rw-r--r-- | src/url_parsing/mod.rs | 508 |
1 files changed, 508 insertions, 0 deletions
diff --git a/src/url_parsing/mod.rs b/src/url_parsing/mod.rs new file mode 100644 index 0000000..e0a95ad --- /dev/null +++ b/src/url_parsing/mod.rs @@ -0,0 +1,508 @@ +use crate::UrlParsing; +use std::ops::Deref; +use std::ops::Add; + +impl UrlParsing { + /// Computes a `used_text` from an input URL according to the passed in `UrlParsing` object. + /// Aims to be kinda compatible to Passwordmaker Pro. + pub(super) fn make_used_text_from_url(&self, input : &str, ) -> String { + parse_url(input).filter_by_settings(self).recombine() + } + + fn is_protocol_used(&self) -> bool{ + match self.use_protocol{ + crate::ProtocolUsageMode::Ignored => false, + crate::ProtocolUsageMode::Used + | crate::ProtocolUsageMode::UsedWithUndefinedIfEmpty + => true, + } + } + fn use_protocol_undefined_fallback(&self) -> bool{ + match self.use_protocol{ + crate::ProtocolUsageMode::Ignored + | crate::ProtocolUsageMode::Used + => false, + crate::ProtocolUsageMode::UsedWithUndefinedIfEmpty => true, + } + } +} + +#[cfg_attr(test, derive(PartialEq, Debug, Clone))] +struct UrlParts<'s> { + protocol : &'s str, + userinfo : &'s str, //Treating this field separate is an addition to the functionaliyt offered by PasswordMaker Pro + subdomain : &'s str, //this is not part of the official URI spec. But PasswordMaker Pro uses it. + domain: &'s str, + port: &'s str, //this would not need to be separated from path_query_fragment, but it's easier to parse if it's separate. + path_query_fragment: &'s str //we don't need to separate those. Passwordmaker doesn't either. +} + +impl<'s> UrlParts<'s> { + fn filter_by_settings(self, settings : &UrlParsing) -> UsedUrlParts<'s>{ + let has_protocol = settings.is_protocol_used() && !self.protocol.is_empty(); + UsedUrlParts{ + protocol: //PasswordMaker Pro compatibility: Protocol is handled _weird_... + if has_protocol { self.protocol } + else if settings.use_protocol_undefined_fallback() { "undefined" } + else { <&str>::default() }, + protocol_separator: if has_protocol { "://" } else { <&str>::default() }, //this is again some PasswordMaker Pro weirdness... + userinfo: if settings.use_userinfo { self.userinfo } else { <&str>::default() }, + subdomain: if settings.use_subdomains { self.subdomain } else { <&str>::default() }, + domain: if settings.use_domain { self.domain } else { <&str>::default() }, + port: if settings.use_port_path { self.port } else { <&str>::default() }, + path_query_fragment: if settings.use_port_path { self.path_query_fragment } else { <&str>::default() }, + } + } +} + +#[cfg_attr(test, derive(PartialEq, Debug))] +struct UsedUrlParts<'s> { + protocol : &'s str, + protocol_separator : &'s str, + userinfo : &'s str, //Treating this field separate is an addition to the functionaliyt offered by PasswordMaker Pro + subdomain : &'s str, //this is not part of the official URI spec. But PasswordMaker Pro uses it. + domain: &'s str, + port: &'s str, //this would not need to be separated from path_query_fragment, but it's easier to parse if it's separate. + path_query_fragment: &'s str //we don't need to separate those. Passwordmaker doesn't either. +} + +impl<'s> UsedUrlParts<'s> { + #[allow(clippy::doc_markdown)] + /// Tries to do assemble a string in a way that's at least somehow compatible to PasswordMaker Pro. + /// This prioritizes ease of use ("what the user expects") over correct URI parsing. + fn recombine(self) -> String { + //matching would need 64 arms... Too much work, soooo, a couple of ifs and less sanity instead. + let has_userinfo = !self.userinfo.is_empty(); + let has_subdomain = !self.subdomain.is_empty(); + let has_domain = !self.domain.is_empty(); + let has_port = !self.port.is_empty(); + let has_path_query_fragment = !self.path_query_fragment.is_empty(); + + //by doing all logic on &str, we save allocations to the very last moment. Also, the syntax is more readable. + let parts = [ + self.protocol, + self.protocol_separator, + self.userinfo, + if has_userinfo && (has_domain || has_subdomain || has_port|| has_path_query_fragment) { "@" } else { <&str>::default() }, + self.subdomain, + if has_subdomain && has_domain { "." } else { <&str>::default() }, + self.domain, + if has_port && (has_userinfo || has_domain || has_subdomain) { ":" } else { <&str>::default() }, + self.port, + self.path_query_fragment, + ]; + + let needed_size = parts.iter().map(Deref::deref).map(<str>::len).sum(); + parts.iter().map(Deref::deref).fold(String::with_capacity(needed_size), String::add) + } +} + +#[allow(clippy::doc_markdown)] +/// Parses the input URI in a way that resembles the behaviour of PasswordMaker Pro. This is intentionally not following the URI standard. +/// It priorizes ease-of-use over strictly following the URI standard. +/// The idea here is that users tend to input strings of the form "www.somedomain.com", what is not a valid URI (authority is not optional). +/// Input of this form should still work though, in order not to confuse users. +fn parse_url(input : &str) -> UrlParts{ + let maybe_protocol = input.split_once(':'); + let has_protocol = maybe_protocol.is_some(); + let (protocol, rest) = maybe_protocol.unwrap_or((<&str>::default(), input)); + let removed_authority_marker = rest.strip_prefix("//"); + let has_authority = removed_authority_marker.is_some(); + let rest = removed_authority_marker.unwrap_or(rest); + + //Authority stops at first / character. Or, if none encountered, at end of input. Slash is part of path. + //If there is a protocol, but no authority, we must treat everything after the intial ':' as path though. + let first_character_of_path = if has_protocol && !has_authority {Some(0)} else {rest.find('/')}; + let (authority, path_query_fragment) = first_character_of_path.map_or((rest, <&str>::default()),|mid| rest.split_at(mid)); + //must split authority at '@' characters. Otherwise ':' is ambigious. + let (userinfo, host_and_port) = authority.split_once('@').unwrap_or((<&str>::default(), authority)); + let (address, port) = host_and_port.split_once(':').unwrap_or((host_and_port, <&str>::default())); + let separator_between_subdom_and_domain = address.rmatch_indices('.').nth(1); + let (subdomain, domain_with_leading_dot) = separator_between_subdom_and_domain.map_or((<&str>::default(), address), |(i, _)| address.split_at(i)); + let domain = domain_with_leading_dot.strip_prefix('.').unwrap_or(domain_with_leading_dot); + UrlParts{protocol, userinfo, subdomain, domain, port, path_query_fragment} +} + +#[cfg(test)] +mod url_parsing_tests { + use crate::ProtocolUsageMode; + + use super::*; + + /// Just tries to split some example urls and checks if the result is as expected. This tests against PasswordMaker Pro behaviour, not proper URI format. + #[test] + fn uri_splitting_test_full_uri(){ + let input = "http://anon:12345@some.subdomain.of.some.domain.com:8080/some/path/with?query&and#fragment"; + let expected = UrlParts{ + protocol: "http", + userinfo: "anon:12345", + subdomain: "some.subdomain.of.some", + domain: "domain.com", + port: "8080", + path_query_fragment: "/some/path/with?query&and#fragment", + }; + let result = parse_url(input); + assert_eq!(result, expected); + } + #[test] + fn uri_splitting_test_no_userinfo(){ + let input = "http://some.subdomain.of.some.domain.com:8080/some/path/with?query&and#fragment"; + let expected = UrlParts{ + protocol: "http", + userinfo: <&str>::default(), + subdomain: "some.subdomain.of.some", + domain: "domain.com", + port: "8080", + path_query_fragment: "/some/path/with?query&and#fragment", + }; + let result = parse_url(input); + assert_eq!(result, expected); + } + #[test] + fn uri_splitting_test_no_port(){ + let input = "http://anon:12345@some.subdomain.of.some.domain.com/some/path/with?query&and#fragment"; + let expected = UrlParts{ + protocol: "http", + userinfo: "anon:12345", + subdomain: "some.subdomain.of.some", + domain: "domain.com", + port: <&str>::default(), + path_query_fragment: "/some/path/with?query&and#fragment", + }; + let result = parse_url(input); + assert_eq!(result, expected); + } + #[test] + fn uri_splitting_test_no_domain(){ + let input = "http://anon:12345@:8080/some/path/with?query&and#fragment"; + let expected = UrlParts{ + protocol: "http", + userinfo: "anon:12345", + subdomain: <&str>::default(), + domain: <&str>::default(), + port: "8080", + path_query_fragment: "/some/path/with?query&and#fragment", + }; + let result = parse_url(input); + assert_eq!(result, expected); + } + #[test] + fn uri_splitting_test_no_domain_no_port(){ + let input = "http://anon:12345@/some/path/with?query&and#fragment"; + let expected = UrlParts{ + protocol: "http", + userinfo: "anon:12345", + subdomain: <&str>::default(), + domain: <&str>::default(), + port: <&str>::default(), + path_query_fragment: "/some/path/with?query&and#fragment", + }; + let result = parse_url(input); + assert_eq!(result, expected); + } + #[test] + fn uri_splitting_test_empty_path(){ + let input = "http://anon:12345@some.subdomain.of.some.domain.com:8080"; + let expected = UrlParts{ + protocol: "http", + userinfo: "anon:12345", + subdomain: "some.subdomain.of.some", + domain: "domain.com", + port: "8080", + path_query_fragment: <&str>::default(), + }; + let result = parse_url(input); + assert_eq!(result, expected); + } + #[test] + fn uri_splitting_test_only_protocol_and_path(){ + let input = "http:some/path/"; + let expected = UrlParts{ + protocol: "http", + userinfo: <&str>::default(), + subdomain: <&str>::default(), + domain: <&str>::default(), + port: <&str>::default(), + path_query_fragment: "some/path/", + }; + let result = parse_url(input); + assert_eq!(result, expected); + } + + /// This triggers me. It should not work (scheme isn't optional), but users would miss it. + /// Password and Port are not included too, because those would be (correctly) identified as schemes. + #[test] + fn uri_splitting_missing_protocol(){ + let input = "anon@some.subdomain.of.some.domain.com/some/path/with?query&and#fragment"; + let expected = UrlParts{ + protocol: <&str>::default(), + userinfo: "anon", + subdomain: "some.subdomain.of.some", + domain: "domain.com", + port: <&str>::default(), + path_query_fragment: "/some/path/with?query&and#fragment", + }; + let result = parse_url(input); + assert_eq!(result, expected); + } + #[test] + fn uri_splitting_just_domain_and_path(){ + let input = "some.subdomain.of.some.domain.com/some/path/with?query&and#fragment"; + let expected = UrlParts{ + protocol: <&str>::default(), + userinfo: <&str>::default(), + subdomain: "some.subdomain.of.some", + domain: "domain.com", + port: <&str>::default(), + path_query_fragment: "/some/path/with?query&and#fragment", + }; + let result = parse_url(input); + assert_eq!(result, expected); + } + #[test] + fn uri_splitting_just_domain_and_subdomain(){ + let input = "some.subdomain.of.some.domain.com"; + let expected = UrlParts{ + protocol: <&str>::default(), + userinfo: <&str>::default(), + subdomain: "some.subdomain.of.some", + domain: "domain.com", + port: <&str>::default(), + path_query_fragment: <&str>::default(), + }; + let result = parse_url(input); + assert_eq!(result, expected); + } + #[test] + fn uri_splitting_just_domain(){ + let input = "domain.com"; + let expected = UrlParts{ + protocol: <&str>::default(), + userinfo: <&str>::default(), + subdomain: <&str>::default(), + domain: "domain.com", + port: <&str>::default(), + path_query_fragment: <&str>::default(), + }; + let result = parse_url(input); + assert_eq!(result, expected); + } + #[test] + fn uri_splitting_only_protocol(){ + let input = "ftp:"; + let expected = UrlParts{ + protocol: "ftp", + userinfo: <&str>::default(), + subdomain: <&str>::default(), + domain: <&str>::default(), + port: <&str>::default(), + path_query_fragment: <&str>::default(), + }; + let result = parse_url(input); + assert_eq!(result, expected); + } + + // Above tests are incomplete. I mean, there are 64 combinations... And then there could be errors... + // Soo, let's just pretend it's fine, and if there are bugs, add the specific buggy input. + //----------------------------------------------------------------------------- + // Reassembly tests + // Again our valid input range is 64 values... And again we just test some samples that are known to be weird. + // For everything else, let's wait for bug reports. + + /// However, for settings application, every combination can be tested. + #[test] + fn apply_settings_to_url_parts_no_undefined_protocol(){ + for i in 0..64 { + let settings = UrlParsing { + use_protocol: if i%2 == 0 { ProtocolUsageMode::Used } else { ProtocolUsageMode::Ignored }, + use_userinfo: (i/2)%2 == 0, + use_subdomains: (i/4)%2 == 0, + use_domain: (i/8)%2 == 0, + use_port_path: (i/16)%2 == 0, + }; + + let inputs = UrlParts { + protocol: if (i/32)%2 == 0 {"proto"} else {""}, + userinfo: "plasmic", + subdomain: "pirate", + domain: "hordes", + port: "420", + path_query_fragment: "under/blackened#banners", + }; + + let output = inputs.clone().filter_by_settings(&settings); + if settings.is_protocol_used() { assert_eq!(output.protocol, inputs.protocol) } else { assert_eq!(output.protocol, "") }; + if settings.is_protocol_used() && !inputs.protocol.is_empty() { assert_eq!(output.protocol_separator, "://") } else { assert_eq!(output.protocol_separator, "") }; + if settings.use_userinfo { assert_eq!(output.userinfo, inputs.userinfo) } else { assert_eq!(output.userinfo, "")}; + if settings.use_subdomains { assert_eq!(output.subdomain, inputs.subdomain) } else { assert_eq!(output.subdomain, "")}; + if settings.use_domain { assert_eq!(output.domain, inputs.domain) } else { assert_eq!(output.domain, "")}; + if settings.use_port_path { assert_eq!(output.port, inputs.port) } else { assert_eq!(output.port, "")}; + if settings.use_port_path { assert_eq!(output.path_query_fragment, inputs.path_query_fragment) } else { assert_eq!(output.path_query_fragment, "")}; + } + } + #[test] + fn apply_settings_to_url_parts_undefined_protocol(){ + for i in 0..64 { + let settings = UrlParsing { + use_protocol: if i%2 == 0 { ProtocolUsageMode::UsedWithUndefinedIfEmpty } else { ProtocolUsageMode::Ignored }, + use_userinfo: (i/2)%2 == 0, + use_subdomains: (i/4)%2 == 0, + use_domain: (i/8)%2 == 0, + use_port_path: (i/16)%2 == 0, + }; + + let inputs = UrlParts { + protocol: if (i/32)%2 == 0 {"proto"} else {""}, + userinfo: "plasmic", + subdomain: "pirate", + domain: "hordes", + port: "420", + path_query_fragment: "under/blackened#banners", + }; + + let output = inputs.clone().filter_by_settings(&settings); + if settings.is_protocol_used() { + if !inputs.protocol.is_empty() { + assert_eq!(output.protocol, inputs.protocol) + } else { + assert_eq!(output.protocol, "undefined") + } + } else { + assert_eq!(output.protocol, "") + }; + if settings.is_protocol_used() && !inputs.protocol.is_empty() { assert_eq!(output.protocol_separator, "://") } else { assert_eq!(output.protocol_separator, "") }; + if settings.use_userinfo { assert_eq!(output.userinfo, inputs.userinfo) } else { assert_eq!(output.userinfo, "")}; + if settings.use_subdomains { assert_eq!(output.subdomain, inputs.subdomain) } else { assert_eq!(output.subdomain, "")}; + if settings.use_domain { assert_eq!(output.domain, inputs.domain) } else { assert_eq!(output.domain, "")}; + if settings.use_port_path { assert_eq!(output.port, inputs.port) } else { assert_eq!(output.port, "")}; + if settings.use_port_path { assert_eq!(output.path_query_fragment, inputs.path_query_fragment) } else { assert_eq!(output.path_query_fragment, "")}; + } + } + + #[test] + fn recombine_full_url_test() { + let input = UsedUrlParts{ + protocol: "xmpp", + protocol_separator: "://", + userinfo: "horst:12345", + subdomain: "www", + domain: "example.com", + port: "8080", + path_query_fragment: "/some/path", + }; + let result = input.recombine(); + assert_eq!(result, "xmpp://horst:12345@www.example.com:8080/some/path"); + } + #[test] + fn recombine_user_but_no_subdomain() { + let input = UsedUrlParts{ + protocol: "xmpp", + protocol_separator: "://", + userinfo: "horst:12345", + subdomain: <&str>::default(), + domain: "example.com", + port: "8080", + path_query_fragment: "/some/path", + }; + let result = input.recombine(); + assert_eq!(result, "xmpp://horst:12345@example.com:8080/some/path"); + } + #[test] + fn recombine_no_user_but_subdomain() { + let input = UsedUrlParts{ + protocol: "xmpp", + protocol_separator: "://", + userinfo: <&str>::default(), + subdomain: "w3", + domain: "example.com", + port: "8080", + path_query_fragment: "/some/path", + }; + let result = input.recombine(); + assert_eq!(result, "xmpp://w3.example.com:8080/some/path"); + } + #[test] + fn recombine_no_user_no_subdomain() { + let input = UsedUrlParts{ + protocol: "xmpp", + protocol_separator: "://", + userinfo: <&str>::default(), + subdomain: <&str>::default(), + domain: "example.com", + port: "8080", + path_query_fragment: "/some/path", + }; + let result = input.recombine(); + assert_eq!(result, "xmpp://example.com:8080/some/path"); + } + #[test] + fn recombine_no_user_no_subdomain_no_port() { + let input = UsedUrlParts{ + protocol: "xmpp", + protocol_separator: "://", + userinfo: <&str>::default(), + subdomain: <&str>::default(), + domain: "example.com", + port: <&str>::default(), + path_query_fragment: "/some/path", + }; + let result = input.recombine(); + assert_eq!(result, "xmpp://example.com/some/path"); + } + #[test] + fn recombine_undefined_protocol() { + let input = UsedUrlParts{ + protocol: "undefined", + protocol_separator: <&str>::default(), + userinfo: "horst:12345", + subdomain: "www", + domain: "example.com", + port: "8080", + path_query_fragment: "/some/path", + }; + let result = input.recombine(); + assert_eq!(result, "undefinedhorst:12345@www.example.com:8080/some/path"); + } + #[test] + fn recombine_undefined_protocol_no_user_no_subdomain() { + let input = UsedUrlParts{ + protocol: "undefined", + protocol_separator: <&str>::default(), + userinfo: <&str>::default(), + subdomain: <&str>::default(), + domain: "example.com", + port: <&str>::default(), + path_query_fragment: "/some/path", + }; + let result = input.recombine(); + assert_eq!(result, "undefinedexample.com/some/path"); + } + #[test] + fn recombine_no_protocol() { + let input = UsedUrlParts{ + protocol: <&str>::default(), + protocol_separator: <&str>::default(), + userinfo: <&str>::default(), + subdomain: "www", + domain: "example.com", + port: <&str>::default(), + path_query_fragment: "/some/path", + }; + let result = input.recombine(); + assert_eq!(result, "www.example.com/some/path"); + } + #[test] + fn recombine_empty_path() { + let input = UsedUrlParts{ + protocol: "xmpp", + protocol_separator: "://", + userinfo: "horst:12345", + subdomain: "www", + domain: "example.com", + port: "8080", + path_query_fragment: <&str>::default(), + }; + let result = input.recombine(); + assert_eq!(result, "xmpp://horst:12345@www.example.com:8080"); + } +}
\ No newline at end of file |