| 1 | //! Decompose a URL typed or pasted into the URL bar. This is the inverse of |
| 2 | //! [`super::client`]'s `build_url`: that composes `base + path + query` into a |
| 3 | //! request, this splits a pasted URL back into the pieces a |
| 4 | //! [`crate::model::SavedRequest`] stores. |
| 5 | |
| 6 | use url::Url; |
| 7 | |
| 8 | use crate::model::KeyValueRow; |
| 9 | |
| 10 | /// The pieces a URL-bar entry decomposes into. |
| 11 | #[derive(Debug, Clone, Default, PartialEq, Eq)] |
| 12 | pub struct UrlParts { |
| 13 | /// Origin (`scheme://host[:port]`) when an absolute http(s) URL was pasted; |
| 14 | /// `None` for a bare path, which is what the field normally holds. |
| 15 | pub origin: Option<String>, |
| 16 | /// Always leading-slash (or a `{{var}}` template), relative to `origin`. |
| 17 | pub path: String, |
| 18 | /// `None` when the input had no `?` at all, meaning "leave the existing |
| 19 | /// query rows alone"; `Some(rows)` when it did, even if empty — a bare `?` |
| 20 | /// clears them. |
| 21 | pub query: Option<Vec<KeyValueRow>>, |
| 22 | } |
| 23 | |
| 24 | pub fn split_url_input(input: &str) -> UrlParts { |
| 25 | let input = input.trim(); |
| 26 | if input.is_empty() { |
| 27 | return UrlParts { |
| 28 | origin: None, |
| 29 | path: "/".into(), |
| 30 | query: None, |
| 31 | }; |
| 32 | } |
| 33 | |
| 34 | // `Url::parse` succeeds for anything with a scheme, including nonsense like |
| 35 | // `localhost:8080/pets` (scheme `localhost`, path `8080/pets`), so the |
| 36 | // scheme and host checks are load-bearing rather than cosmetic. |
| 37 | if let Ok(u) = Url::parse(input) |
| 38 | && matches!(u.scheme(), "http" | "https") |
| 39 | && u.host().is_some() |
| 40 | { |
| 41 | return UrlParts { |
| 42 | origin: Some(u.origin().ascii_serialization()), |
| 43 | path: restore_braces(u.path()), |
| 44 | query: u.query().map(parse_query), |
| 45 | // The fragment is deliberately dropped: it is never sent to a server. |
| 46 | }; |
| 47 | } |
| 48 | |
| 49 | let no_fragment = input.split_once('#').map_or(input, |(head, _)| head); |
| 50 | let (path, query) = match no_fragment.split_once('?') { |
| 51 | Some((p, q)) => (p, Some(parse_query(q))), |
| 52 | None => (no_fragment, None), |
| 53 | }; |
| 54 | // `{{baseUrl}}/pets` is a template, not a relative path — leave it verbatim. |
| 55 | let path = if path.is_empty() { |
| 56 | "/".to_string() |
| 57 | } else if path.starts_with('/') || path.starts_with("{{") { |
| 58 | path.to_string() |
| 59 | } else { |
| 60 | format!("/{path}") |
| 61 | }; |
| 62 | UrlParts { |
| 63 | origin: None, |
| 64 | path, |
| 65 | query, |
| 66 | } |
| 67 | } |
| 68 | |
| 69 | /// Undo the `url` crate's percent-encoding of `{` and `}`, which are in its |
| 70 | /// path encode set — so `/pets/{id}` comes back as `/pets/%7Bid%7D` and would |
| 71 | /// break both `{pathParam}` replacement and `{{variable}}` substitution. Only |
| 72 | /// the braces are restored; a blanket decode would corrupt segments the user |
| 73 | /// percent-encoded on purpose. |
| 74 | fn restore_braces(s: &str) -> String { |
| 75 | s.replace("%7B", "{") |
| 76 | .replace("%7b", "{") |
| 77 | .replace("%7D", "}") |
| 78 | .replace("%7d", "}") |
| 79 | } |
| 80 | |
| 81 | /// Query string to enabled rows. Values are decoded here and re-encoded by |
| 82 | /// `build_url`'s `.query(&pairs)`, so they round-trip. |
| 83 | fn parse_query(raw: &str) -> Vec<KeyValueRow> { |
| 84 | url::form_urlencoded::parse(raw.as_bytes()) |
| 85 | .map(|(k, v)| KeyValueRow::new(k.as_ref(), v.as_ref(), true)) |
| 86 | .filter(|r| !r.key.is_empty()) |
| 87 | .collect() |
| 88 | } |
| 89 | |
| 90 | #[cfg(test)] |
| 91 | mod tests { |
| 92 | use super::*; |
| 93 | |
| 94 | fn rows(parts: &UrlParts) -> Vec<(String, String)> { |
| 95 | parts |
| 96 | .query |
| 97 | .as_ref() |
| 98 | .map(|q| q.iter().map(|r| (r.key.clone(), r.value.clone())).collect()) |
| 99 | .unwrap_or_default() |
| 100 | } |
| 101 | |
| 102 | #[test] |
| 103 | fn splits_a_full_url() { |
| 104 | let p = split_url_input("https://api.example.com/v1/pets?limit=10"); |
| 105 | assert_eq!(p.origin.as_deref(), Some("https://api.example.com")); |
| 106 | assert_eq!(p.path, "/v1/pets"); |
| 107 | assert_eq!(rows(&p), vec![("limit".to_string(), "10".to_string())]); |
| 108 | } |
| 109 | |
| 110 | #[test] |
| 111 | fn bare_path_has_no_origin() { |
| 112 | let p = split_url_input("/v1/pets"); |
| 113 | assert_eq!(p.origin, None); |
| 114 | assert_eq!(p.path, "/v1/pets"); |
| 115 | assert_eq!(p.query, None); |
| 116 | } |
| 117 | |
| 118 | #[test] |
| 119 | fn relative_path_gains_a_leading_slash() { |
| 120 | assert_eq!(split_url_input("pets/42").path, "/pets/42"); |
| 121 | } |
| 122 | |
| 123 | #[test] |
| 124 | fn host_with_port_and_no_scheme_is_relative() { |
| 125 | // `Url::parse` accepts this with scheme "localhost"; we must not. |
| 126 | let p = split_url_input("localhost:8080/pets"); |
| 127 | assert_eq!(p.origin, None); |
| 128 | assert_eq!(p.path, "/localhost:8080/pets"); |
| 129 | } |
| 130 | |
| 131 | #[test] |
| 132 | fn keeps_brace_placeholders_unencoded() { |
| 133 | let p = split_url_input("https://api.example.com/pets/{petId}/photos"); |
| 134 | assert_eq!(p.path, "/pets/{petId}/photos"); |
| 135 | } |
| 136 | |
| 137 | #[test] |
| 138 | fn keeps_double_brace_variables_in_a_relative_path() { |
| 139 | let p = split_url_input("{{prefix}}/pets"); |
| 140 | assert_eq!(p.origin, None); |
| 141 | assert_eq!(p.path, "{{prefix}}/pets"); |
| 142 | } |
| 143 | |
| 144 | #[test] |
| 145 | fn drops_the_fragment() { |
| 146 | assert_eq!( |
| 147 | split_url_input("https://api.example.com/pets#section").path, |
| 148 | "/pets" |
| 149 | ); |
| 150 | assert_eq!(split_url_input("/pets#section").path, "/pets"); |
| 151 | assert_eq!(split_url_input("/pets?a=1#section").path, "/pets"); |
| 152 | assert_eq!( |
| 153 | rows(&split_url_input("/pets?a=1#section")), |
| 154 | vec![("a".to_string(), "1".to_string())] |
| 155 | ); |
| 156 | } |
| 157 | |
| 158 | #[test] |
| 159 | fn empty_input_becomes_root_path() { |
| 160 | let p = split_url_input(" "); |
| 161 | assert_eq!(p.path, "/"); |
| 162 | assert_eq!(p.origin, None); |
| 163 | assert_eq!(p.query, None); |
| 164 | } |
| 165 | |
| 166 | #[test] |
| 167 | fn absent_question_mark_leaves_query_none() { |
| 168 | assert_eq!(split_url_input("https://api.example.com/pets").query, None); |
| 169 | assert_eq!(split_url_input("/pets").query, None); |
| 170 | } |
| 171 | |
| 172 | #[test] |
| 173 | fn bare_question_mark_clears_the_query() { |
| 174 | assert_eq!(split_url_input("/pets?").query, Some(Vec::new())); |
| 175 | assert_eq!( |
| 176 | split_url_input("https://api.example.com/pets?").query, |
| 177 | Some(Vec::new()) |
| 178 | ); |
| 179 | } |
| 180 | |
| 181 | #[test] |
| 182 | fn decodes_query_values() { |
| 183 | let p = split_url_input("/search?q=hello%20world&tag=a%2Bb"); |
| 184 | assert_eq!( |
| 185 | rows(&p), |
| 186 | vec![ |
| 187 | ("q".to_string(), "hello world".to_string()), |
| 188 | ("tag".to_string(), "a+b".to_string()), |
| 189 | ] |
| 190 | ); |
| 191 | } |
| 192 | |
| 193 | #[test] |
| 194 | fn default_ports_are_dropped_from_the_origin() { |
| 195 | assert_eq!( |
| 196 | split_url_input("https://api.example.com:443/pets") |
| 197 | .origin |
| 198 | .as_deref(), |
| 199 | Some("https://api.example.com") |
| 200 | ); |
| 201 | assert_eq!( |
| 202 | split_url_input("http://localhost:8080/pets") |
| 203 | .origin |
| 204 | .as_deref(), |
| 205 | Some("http://localhost:8080") |
| 206 | ); |
| 207 | } |
| 208 | |
| 209 | #[test] |
| 210 | fn query_rows_are_enabled_and_skip_empty_keys() { |
| 211 | let p = split_url_input("/pets?a=1&=2&b"); |
| 212 | let q = p.query.unwrap(); |
| 213 | assert_eq!(q.len(), 2); |
| 214 | assert!(q.iter().all(|r| r.enabled)); |
| 215 | assert_eq!(q[1].key, "b"); |
| 216 | assert_eq!(q[1].value, ""); |
| 217 | } |
| 218 | } |