9207498870

Committed 23 May 2024 07:11AM UTC coverage: 76.113% (-0.3%) from 76.402%

Build # 9207498870

Build Type

push

github

Committed by

web-flow

Commit Message

Add to `IsoDurationParser` documentation in `ixdtf` (#4916)

Run Details

53397 of 70155 relevant lines covered (76.11%)

514353.71 hits per line

Source File
Press 'n' to go to next uncovered line, 'b' for previous

71.74

/components/segmenter/src/rule_segmenter.rs

// This file is part of ICU4X. For terms of use, please see the file
// called LICENSE at the top level of the ICU4X source tree
// (online at: https://github.com/unicode-org/icu4x/blob/main/LICENSE ).

use crate::complex::ComplexPayloads;
use crate::indices::{Latin1Indices, Utf16Indices};
use crate::provider::*;
use crate::WordType;
use core::str::CharIndices;
use utf8_iter::Utf8CharIndices;

/// A trait allowing for RuleBreakIterator to be generalized to multiple string
/// encoding methods and granularity such as grapheme cluster, word, etc.
pub trait RuleBreakType<'l, 's> {
    /// The iterator over characters.
    type IterAttr: Iterator<Item = (usize, Self::CharType)> + Clone + core::fmt::Debug;

    /// The character type.
    type CharType: Copy + Into<u32> + core::fmt::Debug;

    fn get_current_position_character_len(iter: &RuleBreakIterator<'l, 's, Self>) -> usize;

    fn handle_complex_language(
        iter: &mut RuleBreakIterator<'l, 's, Self>,
        left_codepoint: Self::CharType,
    ) -> Option<usize>;
}

/// Implements the [`Iterator`] trait over the segmenter boundaries of the given string.
///
/// Lifetimes:
///
/// - `'l` = lifetime of the segmenter object from which this iterator was created
/// - `'s` = lifetime of the string being segmented
///
/// The [`Iterator::Item`] is an [`usize`] representing index of a code unit
/// _after_ the boundary (for a boundary at the end of text, this index is the length
/// of the [`str`] or array of code units).
#[derive(Debug)]
pub struct RuleBreakIterator<'l, 's, Y: RuleBreakType<'l, 's> + ?Sized> {
    pub(crate) iter: Y::IterAttr,
    pub(crate) len: usize,
    pub(crate) current_pos_data: Option<(usize, Y::CharType)>,
    pub(crate) result_cache: alloc::vec::Vec<usize>,
    pub(crate) data: &'l RuleBreakDataV1<'l>,
    pub(crate) complex: Option<&'l ComplexPayloads>,
    pub(crate) boundary_property: u8,
}

impl<'l, 's, Y: RuleBreakType<'l, 's> + ?Sized> Iterator for RuleBreakIterator<'l, 's, Y> {
    type Item = usize;

    fn next(&mut self) -> Option<Self::Item> {
        // If we have break point cache by previous run, return this result
        if let Some(&first_result) = self.result_cache.first() {
            let mut i = 0;
            loop {
                if i == first_result {
                    self.result_cache = self.result_cache.iter().skip(1).map(|r| r - i).collect();
                    return self.get_current_position();
                }
                i += Y::get_current_position_character_len(self);
                self.advance_iter();
                if self.is_eof() {
                    self.result_cache.clear();
                    self.boundary_property = self.data.complex_property;
                    return Some(self.len);
                }
            }
        }

        if self.is_eof() {
            self.advance_iter();
            if self.is_eof() && self.len == 0 {
                // Empty string. Since `self.current_pos_data` is always going to be empty,
                // we never read `self.len` except for here, so we can use it to mark that
                // we have already returned the single empty-string breakpoint.
                self.len = 1;
                return Some(0);
            }
            let Some(right_prop) = self.get_current_break_property() else {
                // iterator already reaches to EOT. Reset boundary property for word-like.
                self.boundary_property = 0;
                return None;
            };
            // SOT x anything
            if matches!(
                self.get_break_state_from_table(self.data.sot_property, right_prop),
                BreakState::Break | BreakState::NoMatch
            ) {
                self.boundary_property = 0; // SOT is special type
                return self.get_current_position();
            }
        }

        'a: loop {
            debug_assert!(!self.is_eof());
            let left_codepoint = self.get_current_codepoint()?;
            let left_prop = self.get_break_property(left_codepoint);
            self.advance_iter();

            let Some(right_prop) = self.get_current_break_property() else {
                self.boundary_property = left_prop;
                return Some(self.len);
            };

            // Some segmenter rules doesn't have language-specific rules, we have to use LSTM (or dictionary) segmenter.
            // If property is marked as SA, use it
            if right_prop == self.data.complex_property {
                if left_prop != self.data.complex_property {
                    // break before SA
                    self.boundary_property = left_prop;
                    return self.get_current_position();
                }
                let break_offset = Y::handle_complex_language(self, left_codepoint);
                if break_offset.is_some() {
                    return break_offset;
                }
            }

            match self.get_break_state_from_table(left_prop, right_prop) {
                BreakState::Keep => continue,
                BreakState::Break | BreakState::NoMatch => {
                    self.boundary_property = left_prop;
                    return self.get_current_position();
                }
                BreakState::Index(mut index) | BreakState::Intermediate(mut index) => {
                    // This isn't simple rule set. We need marker to restore iterator to previous position.
                    let mut previous_iter = self.iter.clone();
                    let mut previous_pos_data = self.current_pos_data;
                    let mut previous_left_prop = left_prop;

                    loop {
                        self.advance_iter();

                        let Some(prop) = self.get_current_break_property() else {
                            // Reached EOF. But we are analyzing multiple characters now, so next break may be previous point.
                            self.boundary_property = index;
                            if self.get_break_state_from_table(index, self.data.eot_property)
                                == BreakState::NoMatch
                            {
                                self.boundary_property = previous_left_prop;
                                self.iter = previous_iter;
                                self.current_pos_data = previous_pos_data;
                                return self.get_current_position();
                            }
                            // EOF
                            return Some(self.len);
                        };

                        let previous_break_state_is_cp_prop =
                            index <= self.data.last_codepoint_property;

                        match self.get_break_state_from_table(index, prop) {
                            BreakState::Keep => continue 'a,
                            BreakState::NoMatch => {
                                self.boundary_property = previous_left_prop;
                                self.iter = previous_iter;
                                self.current_pos_data = previous_pos_data;
                                return self.get_current_position();
                            }
                            BreakState::Break => return self.get_current_position(),
                            BreakState::Intermediate(i) => {
                                index = i;
                                if previous_break_state_is_cp_prop {
                                    // Move marker
                                    previous_left_prop = index;
                                }
                                previous_iter = self.iter.clone();
                                previous_pos_data = self.current_pos_data;
                            }
                            BreakState::Index(i) => {
                                index = i;
                                if previous_break_state_is_cp_prop {
                                    // Move marker
                                    previous_iter = self.iter.clone();
                                    previous_pos_data = self.current_pos_data;
                                    previous_left_prop = index;
                                }
                            }
                        }
                    }
                }
            }
        }
    }
}

impl<'l, 's, Y: RuleBreakType<'l, 's> + ?Sized> RuleBreakIterator<'l, 's, Y> {
    pub(crate) fn advance_iter(&mut self) {
        self.current_pos_data = self.iter.next();
    }

    pub(crate) fn is_eof(&self) -> bool {
        self.current_pos_data.is_none()
    }

    pub(crate) fn get_current_break_property(&self) -> Option<u8> {
        self.get_current_codepoint()
            .map(|c| self.get_break_property(c))
    }

    pub(crate) fn get_current_position(&self) -> Option<usize> {
        self.current_pos_data.map(|(pos, _)| pos)
    }

    pub(crate) fn get_current_codepoint(&self) -> Option<Y::CharType> {
        self.current_pos_data.map(|(_, codepoint)| codepoint)
    }

    fn get_break_property(&self, codepoint: Y::CharType) -> u8 {
        // Note: Default value is 0 == UNKNOWN
        self.data.property_table.get32(codepoint.into())
    }

    fn get_break_state_from_table(&self, left: u8, right: u8) -> BreakState {
        let idx = left as usize * self.data.property_count as usize + right as usize;
        // We use unwrap_or to fall back to the base case and prevent panics on bad data.
        self.data
            .break_state_table
            .get(idx)
            .unwrap_or(BreakState::Keep)
    }

    /// Return the status value of break boundary.
    /// If segmenter isn't word, always return WordType::None
    pub fn word_type(&self) -> WordType {
        if self.result_cache.first().is_some() {
            // Dictionary type (CJ and East Asian) is letter.
            return WordType::Letter;
        }
        if self.boundary_property == 0 {
            // break position is SOT / Any
            return WordType::None;
        }
        self.data
            .word_type_table
            .get((self.boundary_property - 1) as usize)
            .unwrap_or(WordType::None)
    }

    /// Return true when break boundary is word-like such as letter/number/CJK
    /// If segmenter isn't word, return false
    pub fn is_word_like(&self) -> bool {
        self.word_type().is_word_like()
    }
}

#[derive(Debug)]
pub struct RuleBreakTypeUtf8;

impl<'l, 's> RuleBreakType<'l, 's> for RuleBreakTypeUtf8 {
    type IterAttr = CharIndices<'s>;
    type CharType = char;

    fn get_current_position_character_len(iter: &RuleBreakIterator<Self>) -> usize {
        iter.get_current_codepoint().map_or(0, |c| c.len_utf8())
    }

    fn handle_complex_language(
        _: &mut RuleBreakIterator<Self>,
        _: Self::CharType,
    ) -> Option<usize> {
        unreachable!()
    }
}

#[derive(Debug)]
pub struct RuleBreakTypePotentiallyIllFormedUtf8;

impl<'l, 's> RuleBreakType<'l, 's> for RuleBreakTypePotentiallyIllFormedUtf8 {
    type IterAttr = Utf8CharIndices<'s>;
    type CharType = char;

    fn get_current_position_character_len(iter: &RuleBreakIterator<Self>) -> usize {
        iter.get_current_codepoint().map_or(0, |c| c.len_utf8())
    }

    fn handle_complex_language(
        _: &mut RuleBreakIterator<Self>,
        _: Self::CharType,
    ) -> Option<usize> {
        unreachable!()
    }
}

#[derive(Debug)]
pub struct RuleBreakTypeLatin1;

impl<'l, 's> RuleBreakType<'l, 's> for RuleBreakTypeLatin1 {
    type IterAttr = Latin1Indices<'s>;
    type CharType = u8;

    fn get_current_position_character_len(_: &RuleBreakIterator<Self>) -> usize {
        unreachable!()
    }

    fn handle_complex_language(
        _: &mut RuleBreakIterator<Self>,
        _: Self::CharType,
    ) -> Option<usize> {
        unreachable!()
    }
}

#[derive(Debug)]
pub struct RuleBreakTypeUtf16;

impl<'l, 's> RuleBreakType<'l, 's> for RuleBreakTypeUtf16 {
    type IterAttr = Utf16Indices<'s>;
    type CharType = u32;

    fn get_current_position_character_len(iter: &RuleBreakIterator<Self>) -> usize {
        match iter.get_current_codepoint() {
            None => 0,
            Some(ch) if ch >= 0x10000 => 2,
            _ => 1,
        }
    }

    fn handle_complex_language(
        _: &mut RuleBreakIterator<Self>,
        _: Self::CharType,
    ) -> Option<usize> {
        unreachable!()
    }
}

1	// This file is part of ICU4X. For terms of use, please see the file	15,273✔
2	// called LICENSE at the top level of the ICU4X source tree
3	// (online at: https://github.com/unicode-org/icu4x/blob/main/LICENSE ).
4
5	use crate::complex::ComplexPayloads;
6	use crate::indices::{Latin1Indices, Utf16Indices};
7	use crate::provider::*;
8	use crate::WordType;
9	use core::str::CharIndices;
10	use utf8_iter::Utf8CharIndices;
11
12	/// A trait allowing for RuleBreakIterator to be generalized to multiple string
13	/// encoding methods and granularity such as grapheme cluster, word, etc.
14	pub trait RuleBreakType<'l, 's> {
15	/// The iterator over characters.
16	type IterAttr: Iterator<Item = (usize, Self::CharType)> + Clone + core::fmt::Debug;
17
18	/// The character type.
19	type CharType: Copy + Into<u32> + core::fmt::Debug;
20
21	fn get_current_position_character_len(iter: &RuleBreakIterator<'l, 's, Self>) -> usize;
22
23	fn handle_complex_language(
24	iter: &mut RuleBreakIterator<'l, 's, Self>,
25	left_codepoint: Self::CharType,
26	) -> Option<usize>;
27	}
28
29	/// Implements the [`Iterator`] trait over the segmenter boundaries of the given string.
30	///
31	/// Lifetimes:
32	///
33	/// - `'l` = lifetime of the segmenter object from which this iterator was created
34	/// - `'s` = lifetime of the string being segmented
35	///
36	/// The [`Iterator::Item`] is an [`usize`] representing index of a code unit
37	/// _after_ the boundary (for a boundary at the end of text, this index is the length
38	/// of the [`str`] or array of code units).
39	#[derive(Debug)]
40	pub struct RuleBreakIterator<'l, 's, Y: RuleBreakType<'l, 's> + ?Sized> {
41	pub(crate) iter: Y::IterAttr,
42	pub(crate) len: usize,
43	pub(crate) current_pos_data: Option<(usize, Y::CharType)>,
44	pub(crate) result_cache: alloc::vec::Vec<usize>,
45	pub(crate) data: &'l RuleBreakDataV1<'l>,
46	pub(crate) complex: Option<&'l ComplexPayloads>,
47	pub(crate) boundary_property: u8,
48	}
49
50	impl<'l, 's, Y: RuleBreakType<'l, 's> + ?Sized> Iterator for RuleBreakIterator<'l, 's, Y> {
51	type Item = usize;
52
53	fn next(&mut self) -> Option<Self::Item> {	91,116✔
54	// If we have break point cache by previous run, return this result
55	if let Some(&first_result) = self.result_cache.first() {	91,116✔
56	let mut i = 0;	2✔
57	loop {	2✔
58	if i == first_result {	6✔
59	self.result_cache = self.result_cache.iter().skip(1).map(\|r\| r - i).collect();	×
60	return self.get_current_position();	×
61	}
62	i += Y::get_current_position_character_len(self);	6✔
63	self.advance_iter();	6✔
64	if self.is_eof() {	6✔
65	self.result_cache.clear();	2✔
66	self.boundary_property = self.data.complex_property;	2✔
67	return Some(self.len);	2✔
68	}
69	}
70	}
71
72	if self.is_eof() {	91,114✔
73	self.advance_iter();	7,756✔
74	if self.is_eof() && self.len == 0 {	7,756✔
75	// Empty string. Since `self.current_pos_data` is always going to be empty,
76	// we never read `self.len` except for here, so we can use it to mark that
77	// we have already returned the single empty-string breakpoint.
78	self.len = 1;	3✔
79	return Some(0);	3✔
80	}
81	let Some(right_prop) = self.get_current_break_property() else {	7,753✔
82	// iterator already reaches to EOT. Reset boundary property for word-like.
83	self.boundary_property = 0;	3,861✔
84	return None;	3,861✔
85	};
86	// SOT x anything
87	if matches!(	3,892✔
88	self.get_break_state_from_table(self.data.sot_property, right_prop),	3,892✔
89	BreakState::Break \| BreakState::NoMatch
90	) {
91	self.boundary_property = 0; // SOT is special type	3,892✔
92	return self.get_current_position();	3,892✔
93	}
94	}
95
96	'a: loop {
97	debug_assert!(!self.is_eof());	189,053✔
98	let left_codepoint = self.get_current_codepoint()?;	189,053✔
99	let left_prop = self.get_break_property(left_codepoint);	189,053✔
100	self.advance_iter();	189,053✔
101
102	let Some(right_prop) = self.get_current_break_property() else {	189,053✔
103	self.boundary_property = left_prop;	3,663✔
104	return Some(self.len);	3,663✔
105	};
106
107	// Some segmenter rules doesn't have language-specific rules, we have to use LSTM (or dictionary) segmenter.
108	// If property is marked as SA, use it
109	if right_prop == self.data.complex_property {	185,390✔
110	if left_prop != self.data.complex_property {	2✔
111	// break before SA
112	self.boundary_property = left_prop;	×
113	return self.get_current_position();	×
114	}
115	let break_offset = Y::handle_complex_language(self, left_codepoint);	2✔
116	if break_offset.is_some() {	2✔
117	return break_offset;	2✔
118	}
119	}
120
121	match self.get_break_state_from_table(left_prop, right_prop) {	185,388✔
122	BreakState::Keep => continue,
123	BreakState::Break \| BreakState::NoMatch => {
124	self.boundary_property = left_prop;	72,026✔
125	return self.get_current_position();	72,026✔
126	}
127	BreakState::Index(mut index) \| BreakState::Intermediate(mut index) => {	15,276✔
128	// This isn't simple rule set. We need marker to restore iterator to previous position.
129	let mut previous_iter = self.iter.clone();	15,276✔
130	let mut previous_pos_data = self.current_pos_data;	15,276✔
131	let mut previous_left_prop = left_prop;	15,276✔
132
133	loop {	15,276✔
134	self.advance_iter();	22,823✔
135
136	let Some(prop) = self.get_current_break_property() else {	22,814✔
137	// Reached EOF. But we are analyzing multiple characters now, so next break may be previous point.
138	self.boundary_property = index;	226✔
139	if self.get_break_state_from_table(index, self.data.eot_property)	226✔
140	== BreakState::NoMatch
141	{
142	self.boundary_property = previous_left_prop;	16✔
143	self.iter = previous_iter;	16✔
144	self.current_pos_data = previous_pos_data;	16✔
145	return self.get_current_position();	16✔
146	}
147	// EOF
148	return Some(self.len);	210✔
149	};
150
151	let previous_break_state_is_cp_prop =
152	index <= self.data.last_codepoint_property;	22,591✔
153
154	match self.get_break_state_from_table(index, prop) {	22,583✔
155	BreakState::Keep => continue 'a,
156	BreakState::NoMatch => {
157	self.boundary_property = previous_left_prop;	962✔
158	self.iter = previous_iter;	962✔
159	self.current_pos_data = previous_pos_data;	962✔
160	return self.get_current_position();	962✔
161	}
162	BreakState::Break => return self.get_current_position(),	6,469✔
163	BreakState::Intermediate(i) => {	638✔
164	index = i;	638✔
165	if previous_break_state_is_cp_prop {	638✔
166	// Move marker
167	previous_left_prop = index;	94✔
168	}
169	previous_iter = self.iter.clone();	638✔
170	previous_pos_data = self.current_pos_data;	638✔
171	}	638✔
172	BreakState::Index(i) => {	6,909✔
173	index = i;	6,909✔
174	if previous_break_state_is_cp_prop {	8,603✔
175	// Move marker
176	previous_iter = self.iter.clone();	1,694✔
177	previous_pos_data = self.current_pos_data;	1,694✔
178	previous_left_prop = index;	1,694✔
179	}
180	}
181	}
182	}
183	}	15,266✔
184	}
185	}
186	}	91,106✔
187	}
188
189	impl<'l, 's, Y: RuleBreakType<'l, 's> + ?Sized> RuleBreakIterator<'l, 's, Y> {
190	pub(crate) fn advance_iter(&mut self) {	242,959✔
191	self.current_pos_data = self.iter.next();	242,959✔
192	}	242,959✔
193
194	pub(crate) fn is_eof(&self) -> bool {	322,721✔
195	self.current_pos_data.is_none()	322,721✔
196	}	322,721✔
197
198	pub(crate) fn get_current_break_property(&self) -> Option<u8> {	242,598✔
199	self.get_current_codepoint()	485,196✔
200	.map(\|c\| self.get_break_property(c))	469,060✔
201	}	242,598✔
202
203	pub(crate) fn get_current_position(&self) -> Option<usize> {	92,515✔
204	self.current_pos_data.map(\|(pos, _)\| pos)	185,022✔
205	}	92,515✔
206
207	pub(crate) fn get_current_codepoint(&self) -> Option<Y::CharType> {	442,031✔
208	self.current_pos_data.map(\|(_, codepoint)\| codepoint)	867,932✔
209	}	442,031✔
210
211	fn get_break_property(&self, codepoint: Y::CharType) -> u8 {	424,757✔
212	// Note: Default value is 0 == UNKNOWN
213	self.data.property_table.get32(codepoint.into())	424,757✔
214	}	424,757✔
215
216	fn get_break_state_from_table(&self, left: u8, right: u8) -> BreakState {	212,008✔
217	let idx = left as usize * self.data.property_count as usize + right as usize;	212,008✔
218	// We use unwrap_or to fall back to the base case and prevent panics on bad data.
219	self.data	424,016✔
220	.break_state_table
221	.get(idx)
222	.unwrap_or(BreakState::Keep)	212,008✔
223	}	212,008✔
224
225	/// Return the status value of break boundary.
226	/// If segmenter isn't word, always return WordType::None
227	pub fn word_type(&self) -> WordType {	×
228	if self.result_cache.first().is_some() {	×
229	// Dictionary type (CJ and East Asian) is letter.
230	return WordType::Letter;	×
231	}
232	if self.boundary_property == 0 {	×
233	// break position is SOT / Any
234	return WordType::None;	×
235	}
236	self.data	×
237	.word_type_table
238	.get((self.boundary_property - 1) as usize)	×
239	.unwrap_or(WordType::None)	×
240	}	×
241
242	/// Return true when break boundary is word-like such as letter/number/CJK
243	/// If segmenter isn't word, return false
244	pub fn is_word_like(&self) -> bool {
245	self.word_type().is_word_like()
246	}
247	}
248
249	#[derive(Debug)]	×
250	pub struct RuleBreakTypeUtf8;
251
252	impl<'l, 's> RuleBreakType<'l, 's> for RuleBreakTypeUtf8 {
253	type IterAttr = CharIndices<'s>;
254	type CharType = char;
255
256	fn get_current_position_character_len(iter: &RuleBreakIterator<Self>) -> usize {	×
257	iter.get_current_codepoint().map_or(0, \|c\| c.len_utf8())	×
258	}	×
259
260	fn handle_complex_language(	×
261	_: &mut RuleBreakIterator<Self>,
262	_: Self::CharType,
263	) -> Option<usize> {
264	unreachable!()	×
265	}
266	}
267
268	#[derive(Debug)]	×
269	pub struct RuleBreakTypePotentiallyIllFormedUtf8;
270
271	impl<'l, 's> RuleBreakType<'l, 's> for RuleBreakTypePotentiallyIllFormedUtf8 {
272	type IterAttr = Utf8CharIndices<'s>;
273	type CharType = char;
274
275	fn get_current_position_character_len(iter: &RuleBreakIterator<Self>) -> usize {	×
276	iter.get_current_codepoint().map_or(0, \|c\| c.len_utf8())	×
277	}	×
278
279	fn handle_complex_language(	×
280	_: &mut RuleBreakIterator<Self>,
281	_: Self::CharType,
282	) -> Option<usize> {
283	unreachable!()	×
284	}
285	}
286
287	#[derive(Debug)]	×
288	pub struct RuleBreakTypeLatin1;
289
290	impl<'l, 's> RuleBreakType<'l, 's> for RuleBreakTypeLatin1 {
291	type IterAttr = Latin1Indices<'s>;
292	type CharType = u8;
293
294	fn get_current_position_character_len(_: &RuleBreakIterator<Self>) -> usize {	×
295	unreachable!()	×
296	}
297
298	fn handle_complex_language(	×
299	_: &mut RuleBreakIterator<Self>,
300	_: Self::CharType,
301	) -> Option<usize> {
302	unreachable!()	×
303	}
304	}
305
306	#[derive(Debug)]	×
307	pub struct RuleBreakTypeUtf16;
308
309	impl<'l, 's> RuleBreakType<'l, 's> for RuleBreakTypeUtf16 {
310	type IterAttr = Utf16Indices<'s>;
311	type CharType = u32;
312
313	fn get_current_position_character_len(iter: &RuleBreakIterator<Self>) -> usize {	×
314	match iter.get_current_codepoint() {	×
315	None => 0,	×
316	Some(ch) if ch >= 0x10000 => 2,	×
317	_ => 1,	×
318	}
319	}	×
320
321	fn handle_complex_language(	×
322	_: &mut RuleBreakIterator<Self>,
323	_: Self::CharType,
324	) -> Option<usize> {
325	unreachable!()	×
326	}
327	}

zbraniecki / icu4x / 9207498870

Source File Press 'n' to go to next uncovered line, 'b' for previous

Source File
Press 'n' to go to next uncovered line, 'b' for previous