Skip to main content

str_utils/
remove_all_invisible_characters.rs

1use alloc::{borrow::Cow, str::from_utf8_unchecked, string::String};
2
3use crate::to_substring_in_place;
4
5/// To extend `str` and `Cow<str>` to have `remove_all_invisible_characters` method.
6pub trait RemoveInvisibleCharacters<'a> {
7    /// Removes all invisible or non-printable characters from a given string.
8    ///
9    /// This function filters out a comprehensive set of Unicode characters that are typically
10    /// invisible or used for control or formatting purposes. This includes:
11    ///
12    /// - ASCII control characters (U+0000 to U+001F and U+007F), exclude `\t` and `\n`
13    /// - Zero-width characters and format controls:
14    ///   - U+200B (Zero Width Space)
15    ///   - U+200C (Zero Width Non-Joiner)
16    ///   - U+200D (Zero Width Joiner)
17    ///   - U+200E to U+200F (Directional marks)
18    ///   - U+202A to U+202E (Directional formatting)
19    ///   - U+2060 to U+2064 (Word Joiner and Invisible Math Symbols)
20    ///   - U+2066 to U+2069 (Bidi Isolates)
21    ///   - U+FEFF (Byte Order Mark / Zero Width No-Break Space)
22    ///
23    /// These characters can interfere with text rendering, parsing, and display,
24    /// and are often used in text-based attacks (e.g., for spoofing).
25    fn remove_all_invisible_characters(self) -> Cow<'a, str>;
26}
27
28impl<'a> RemoveInvisibleCharacters<'a> for &'a str {
29    fn remove_all_invisible_characters(self) -> Cow<'a, str> {
30        let s = self;
31        let bytes = s.as_bytes();
32
33        let length = bytes.len();
34
35        let mut p = 0;
36
37        let check_character_whether_to_remove = |p: usize, e: u8, width: usize| -> bool {
38            match width {
39                1 => {
40                    match e {
41                        // ascii controls, only remain \t, \n
42                        0..=8 | 11..=13 | 14..=31 | 127 => return true,
43                        _ => (),
44                    }
45                },
46                3 => match e {
47                    0xE2 => match bytes[p + 1] {
48                        // zero width characters and bidirectional controls
49                        0x80 => match bytes[p + 2] {
50                            0x8B..=0x8F | 0xAA..=0xAE => return true,
51                            _ => (),
52                        },
53                        // word joiner, invisible times/separator/plus and isolate characters
54                        0x81 => match bytes[p + 2] {
55                            0xA0 | 0xA2..=0xA4 | 0xA6..=0xA9 => return true,
56                            _ => (),
57                        },
58                        _ => (),
59                    },
60                    // zero width character
61                    0xEF if bytes[p + 1] == 0xBB && bytes[p + 2] == 0xBF => {
62                        return true;
63                    },
64                    _ => (),
65                },
66                _ => (),
67            }
68
69            false
70        };
71
72        let width = loop {
73            if p == length {
74                return Cow::Borrowed(s);
75            }
76
77            let e = bytes[p];
78
79            let width = unsafe { utf8_width::get_width_assume_valid(e) };
80
81            if check_character_whether_to_remove(p, e, width) {
82                break width;
83            } else {
84                p += width;
85            }
86        };
87
88        let heading_normal_characters_end_index = p;
89
90        p += width;
91
92        // there are four situations which can use a string slice:
93        // 1. <invisible_characters>
94        // 2. <normal_characters><invisible_characters>
95        // 3. <invisible_characters><normal_characters>
96        // 4. <invisible_characters><normal_characters><invisible_characters>
97
98        // continue to find more invisible characters
99        let width = loop {
100            if p == length {
101                // situation 1 or situation 2
102
103                return Cow::Borrowed(unsafe {
104                    from_utf8_unchecked(&bytes[..heading_normal_characters_end_index])
105                });
106            }
107
108            let e = bytes[p];
109
110            let width = unsafe { utf8_width::get_width_assume_valid(e) };
111
112            if check_character_whether_to_remove(p, e, width) {
113                p += width;
114            } else {
115                break width;
116            }
117        };
118
119        let following_invisible_characters_end_index = p;
120
121        p += width;
122
123        // continue to find more normal characters
124        let width = loop {
125            if p == length {
126                // situation 3
127
128                return Cow::Borrowed(unsafe {
129                    from_utf8_unchecked(&bytes[following_invisible_characters_end_index..])
130                });
131            }
132
133            let e = bytes[p];
134
135            let width = unsafe { utf8_width::get_width_assume_valid(e) };
136
137            if check_character_whether_to_remove(p, e, width) {
138                break width;
139            } else {
140                p += width;
141            }
142        };
143
144        let following_normal_characters_end_index = p;
145
146        p += width;
147
148        // continue to find more invisible characters
149        let width = loop {
150            if p == length {
151                // situation 4
152
153                return Cow::Borrowed(unsafe {
154                    from_utf8_unchecked(
155                        &bytes[following_invisible_characters_end_index
156                            ..following_normal_characters_end_index],
157                    )
158                });
159            }
160
161            let e = bytes[p];
162
163            let width = unsafe { utf8_width::get_width_assume_valid(e) };
164
165            if check_character_whether_to_remove(p, e, width) {
166                p += width;
167            } else {
168                break width;
169            }
170        };
171
172        let mut new_v = bytes
173            [following_invisible_characters_end_index..following_normal_characters_end_index]
174            .to_vec();
175
176        let mut start = p;
177
178        p += width;
179
180        loop {
181            if p == length {
182                break;
183            }
184
185            let e = bytes[p];
186
187            let width = unsafe { utf8_width::get_width_assume_valid(e) };
188
189            if check_character_whether_to_remove(p, e, width) {
190                new_v.extend_from_slice(&bytes[start..p]);
191
192                start = p + width;
193            }
194
195            p += width;
196        }
197
198        new_v.extend_from_slice(&bytes[start..p]);
199
200        Cow::Owned(unsafe { String::from_utf8_unchecked(new_v) })
201    }
202}
203
204impl<'a> RemoveInvisibleCharacters<'a> for Cow<'a, str> {
205    #[inline]
206    fn remove_all_invisible_characters(self) -> Cow<'a, str> {
207        match self {
208            Cow::Borrowed(s) => s.remove_all_invisible_characters(),
209            Cow::Owned(s) => match s.remove_all_invisible_characters() {
210                Cow::Borrowed(ss) => Cow::Owned(unsafe { to_substring_in_place!(s, ss) }),
211                Cow::Owned(s) => Cow::Owned(s),
212            },
213        }
214    }
215}