microcosm.blue
Constellation, Spacedust, Slingshot, UFOs: atproto crates and services for microcosm
add a horrible at-uri parsing/normalization routine
ughhooooooofff
bad-example.com
+243
src/at_uri.rs
+243
src/at_uri.rs
···
1
+
use fluent_uri::{Uri, UriRef};
2
+
use std::sync::LazyLock;
3
+
4
+
static BASE: LazyLock<Uri<&str>> = LazyLock::new(|| Uri::parse("https://example.com").unwrap());
5
+
6
+
// normalizing is a bit opinionated but eh
7
+
/// see "Full AT URI Syntax" at https://atproto.com/specs/at-uri-scheme
8
+
/// this parser is intentinonally lax: it should accept all valid at-uris, and
9
+
/// may accept some invalid at-uris.
10
+
///
11
+
/// at the moment this implementation is quite bad and incomplete
12
+
pub fn parse_at_uri(s: &str) -> Option<String> {
13
+
// for now, just working through the rules laid out in the docs in order,
14
+
// without much regard for efficiency for now.
15
+
16
+
// The overall URI is restricted to a subset of ASCII characters
17
+
if !s.is_ascii() {
18
+
return None;
19
+
}
20
+
// // A-Za-z0-9 . - _ ~
21
+
// if !s.chars().all(|c| matches!(c, 'A'..='Z' | 'a'..='z' | '0'..='9' | '.' | '-' | '_' | '~')) {
22
+
// return None
23
+
// }
24
+
25
+
// Maximum overall length is 8 kilobytes (which may be shortened in the future)
26
+
if s.len() > (8 * 2_usize.pow(10)) {
27
+
return None;
28
+
}
29
+
30
+
// Hex-encoding of characters is permitted (but in practice not necessary)
31
+
// -> decode any unreserved characters. from rfc 3986:
32
+
// -> For consistency, percent-encoded octets in the ranges of ALPHA
33
+
// -> (%41-%5A and %61-%7A), DIGIT (%30-%39), hyphen (%2D), period (%2E),
34
+
// -> underscore (%5F), or tilde (%7E) should not be created by URI
35
+
// -> producers and, when found in a URI, should be decoded to their
36
+
// -> corresponding unreserved characters by URI normalizers.
37
+
let s = if let Some((unencoded_prefix, rest)) = s.split_once('%') {
38
+
let mut out = String::with_capacity(s.len());
39
+
out.push_str(unencoded_prefix);
40
+
for segment in rest.split('%') {
41
+
let Some((hex2, unencoded_suffix)) = segment.split_at_checked(2) else {
42
+
return None; // bail: % must always be followed by 2 hex digits
43
+
};
44
+
let Ok(decoded) = u8::from_str_radix(hex2, 16).map(char::from) else {
45
+
return None; // bail: % must be followed by decodable hex
46
+
};
47
+
if matches!(decoded, 'A'..='Z' | 'a'..='z' | '0'..='9' | '.' | '-' | '_' | '~') {
48
+
out.push(decoded);
49
+
} else {
50
+
out.push('%');
51
+
out.push_str(&hex2.to_ascii_uppercase()); // norm
52
+
}
53
+
out.push_str(unencoded_suffix);
54
+
}
55
+
out
56
+
} else {
57
+
s.to_string()
58
+
};
59
+
60
+
// The URI scheme is `at`, and an authority part preceded with double slashes is always
61
+
// required, so the URI always starts at://
62
+
// -> the spec doesn't explicitly say, but it seems like uri schemes are case-insensitive
63
+
let (proto, rest) = s.split_at_checked(5)?;
64
+
if !proto.eq_ignore_ascii_case("at://") {
65
+
return None;
66
+
}
67
+
68
+
// An authority section is required and must be non-empty. the authority can be either an
69
+
// atproto Handle, or a DID meeting the restrictions for use with atproto. note that the
70
+
// authority part can not be interpreted as a host:port pair, because of the use of colon
71
+
// characters (:) in DIDs. Colons and unreserved characters should not be escaped in DIDs,
72
+
// but other reserved characters (including #, /, $, &, @) must be escaped.
73
+
// Note that none of the current "blessed" DID methods for atproto allow these
74
+
// characters in DID identifiers
75
+
76
+
// An optional path section may follow the authority. The path may contain multiple segments
77
+
// separated by a single slash (/). Generic URI path normalization rules may be used.
78
+
79
+
// An optional query part is allowed, following generic URI syntax restrictions
80
+
81
+
// An optional fragment part is allowed, using JSON Path syntax
82
+
83
+
// -> work backwards from fragment, query, path -> authority
84
+
let mut base = rest;
85
+
let (mut fragment, mut query, mut path) = (None, None, None);
86
+
if let Some((pre, f)) = base.split_once('#') {
87
+
base = pre;
88
+
fragment = Some(f);
89
+
}
90
+
if let Some((pre, q)) = base.split_once('?') {
91
+
base = pre;
92
+
query = Some(q);
93
+
}
94
+
if let Some((pre, p)) = base.split_once('/') {
95
+
base = pre;
96
+
path = Some(p);
97
+
}
98
+
let mut authority = base.to_string();
99
+
100
+
if authority.is_empty() {
101
+
return None;
102
+
}
103
+
104
+
// Normalization: Authority as handle: lowercased
105
+
if !authority.starts_with("did:") {
106
+
// lowercase handles
107
+
authority.make_ascii_lowercase();
108
+
}
109
+
110
+
// Normalization: No trailing slashes in path part
111
+
// Normalization: No duplicate slashes or "dot" sections in path part (/./ or /abc/../ for example)
112
+
// -> be so lazy
113
+
let path = match path {
114
+
Some(p) => {
115
+
let p = p.trim_end_matches('/');
116
+
let uri_ref = UriRef::parse(p).ok()?; // fully bail if we can't parse path
117
+
let resolved = uri_ref.resolve_against(&*BASE).unwrap(); // both fail conditions are specific to BASE
118
+
let normalized = resolved.normalize().path().to_string();
119
+
let without_trailing_slashes = normalized.trim_end_matches('/');
120
+
Some(without_trailing_slashes.to_string())
121
+
}
122
+
None => None,
123
+
};
124
+
125
+
let mut out = format!("at://{authority}");
126
+
if let Some(p) = path {
127
+
// no need for `/` -- it's added by fluent_uri normalization
128
+
out.push_str(&p);
129
+
}
130
+
if let Some(q) = query {
131
+
out.push('?');
132
+
out.push_str(q);
133
+
}
134
+
if let Some(f) = fragment {
135
+
out.push('#');
136
+
out.push_str(f);
137
+
}
138
+
139
+
Some(out)
140
+
141
+
// there's a more normalization to do still. ugh.
142
+
}
143
+
144
+
#[cfg(test)]
145
+
mod tests {
146
+
use super::*;
147
+
148
+
#[test]
149
+
fn test_at_uri_parse() {
150
+
for (case, expected, detail) in vec![
151
+
("", None, "empty"),
152
+
(" ", None, "whitespace"),
153
+
("https://bad-example.com", None, "not at scheme"),
154
+
("at://µcosm.bad-example.com", None, "not ascii"),
155
+
(
156
+
"at://bad-example.com",
157
+
Some("at://bad-example.com"),
158
+
"handle, authority-only",
159
+
),
160
+
(
161
+
"at://did:plc:hdhoaan3xa3jiuq4fg4mefid",
162
+
Some("at://did:plc:hdhoaan3xa3jiuq4fg4mefid"),
163
+
"DID, authority-only",
164
+
),
165
+
(
166
+
"at://bad-example.com/app.bsky.feed.post/3jwdwj2ctlk26",
167
+
Some("at://bad-example.com/app.bsky.feed.post/3jwdwj2ctlk26"),
168
+
"bsky post (handle)",
169
+
),
170
+
(
171
+
"at://did:plc:hdhoaan3xa3jiuq4fg4mefid/app.bsky.feed.post/3ldqksainxc27",
172
+
Some("at://did:plc:hdhoaan3xa3jiuq4fg4mefid/app.bsky.feed.post/3ldqksainxc27"),
173
+
"bsky post (DID)",
174
+
),
175
+
(
176
+
"AT://bad-example.com",
177
+
Some("at://bad-example.com"),
178
+
"scheme case is normalized",
179
+
),
180
+
(
181
+
"at://bad-example.com",
182
+
Some("at://bad-example.com"),
183
+
"scheme case is normalized",
184
+
),
185
+
(
186
+
"at://bad-example.com?q=z",
187
+
Some("at://bad-example.com?q=z"),
188
+
"query is allowed",
189
+
),
190
+
(
191
+
"at://bad-example.com#a",
192
+
Some("at://bad-example.com#a"),
193
+
"fragment is allowed",
194
+
),
195
+
(
196
+
"at://bad-example.com/%",
197
+
None,
198
+
"invalid percent-encoding: ends with %",
199
+
),
200
+
(
201
+
"at://bad-example.com/%2",
202
+
None,
203
+
"invalid percent-encoding: ends with only one digit after %",
204
+
),
205
+
(
206
+
"at://bad-example.com/%ZZ",
207
+
None,
208
+
"invalid percent-encoding: non-hex after %",
209
+
),
210
+
(
211
+
"at://bad-example.com/%3A",
212
+
Some("at://bad-example.com/%3A"),
213
+
"valid percent-encoding is left",
214
+
),
215
+
(
216
+
"at://bad-example.com/%3a",
217
+
Some("at://bad-example.com/%3A"),
218
+
"valid percent-encoding is hex-uppercased",
219
+
),
220
+
(
221
+
"at://bad-example.com/%61/%62",
222
+
Some("at://bad-example.com/a/b"),
223
+
"unreserved characters are percent-decoded",
224
+
),
225
+
(
226
+
"at://bad-example.com/a/../b",
227
+
Some("at://bad-example.com/b"),
228
+
"paths have traversals resolved (oof)",
229
+
),
230
+
(
231
+
"at://bad-example.com/../",
232
+
Some("at://bad-example.com"),
233
+
"paths always have trailing slashes removed",
234
+
),
235
+
] {
236
+
assert_eq!(
237
+
parse_at_uri(case),
238
+
expected.map(|s| s.to_string()),
239
+
"{detail}"
240
+
);
241
+
}
242
+
}
243
+
}
+6
-5
src/lib.rs
+6
-5
src/lib.rs
···
1
1
use fluent_uri::Uri;
2
2
3
+
pub mod at_uri;
4
+
3
5
#[derive(Debug, PartialEq)]
4
6
pub enum Link {
5
7
AtUri(String),
6
8
Uri(String),
7
9
}
8
10
9
-
// normalizing is a bit opinionated
10
-
pub fn parse_at_uri(_s: &str) -> Option<String> {
11
-
// TODO
12
-
None
11
+
// normalizing is a bit opinionated but ehhh
12
+
pub fn parse_at_uri(s: &str) -> Option<String> {
13
+
at_uri::parse_at_uri(s)
13
14
}
14
15
15
-
// normalizing is a bit opinionated
16
+
// normalizing is a bit opinionated but eh
16
17
pub fn parse_uri(s: &str) -> Option<String> {
17
18
Uri::parse(s).map(|u| u.normalize().into_string()).ok()
18
19
}