1
  2
  3
  4
  5
  6
  7
  8
  9
 10
 11
 12
 13
 14
 15
 16
 17
 18
 19
 20
 21
 22
 23
 24
 25
 26
 27
 28
 29
 30
 31
 32
 33
 34
 35
 36
 37
 38
 39
 40
 41
 42
 43
 44
 45
 46
 47
 48
 49
 50
 51
 52
 53
 54
 55
 56
 57
 58
 59
 60
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
//! Parallel iterator types for [ranges][std::range],
//! the type for values created by `a..b` expressions
//!
//! You will rarely need to interact with this module directly unless you have
//! need to name one of the iterator types.
//!
//! ```
//! use rayon::prelude::*;
//!
//! let r = (0..100u64).into_par_iter()
//!                    .sum();
//!
//! // compare result with sequential calculation
//! assert_eq!((0..100).sum::<u64>(), r);
//! ```
//!
//! [std::range]: https://doc.rust-lang.org/core/ops/struct.Range.html

use iter::plumbing::*;
use iter::*;
use std::ops::Range;
use std::usize;

/// Parallel iterator over a range, implemented for all integer types.
///
/// **Note:** The `zip` operation requires `IndexedParallelIterator`
/// which is not implemented for `u64`, `i64`, `u128`, or `i128`.
///
/// ```
/// use rayon::prelude::*;
///
/// let p = (0..25usize).into_par_iter()
///                   .zip(0..25usize)
///                   .filter(|&(x, y)| x % 5 == 0 || y % 5 == 0)
///                   .map(|(x, y)| x * y)
///                   .sum::<usize>();
///
/// let s = (0..25usize).zip(0..25)
///                   .filter(|&(x, y)| x % 5 == 0 || y % 5 == 0)
///                   .map(|(x, y)| x * y)
///                   .sum();
///
/// assert_eq!(p, s);
/// ```
#[derive(Debug, Clone)]
pub struct Iter<T> {
    range: Range<T>,
}

impl<T> IntoParallelIterator for Range<T>
where
    Iter<T>: ParallelIterator,
{
    type Item = <Iter<T> as ParallelIterator>::Item;
    type Iter = Iter<T>;

    fn into_par_iter(self) -> Self::Iter {
        Iter { range: self }
    }
}

struct IterProducer<T> {
    range: Range<T>,
}

impl<T> IntoIterator for IterProducer<T>
where
    Range<T>: Iterator,
{
    type Item = <Range<T> as Iterator>::Item;
    type IntoIter = Range<T>;

    fn into_iter(self) -> Self::IntoIter {
        self.range
    }
}

macro_rules! indexed_range_impl {
    ( $t:ty ) => {
        impl ParallelIterator for Iter<$t> {
            type Item = $t;

            fn drive_unindexed<C>(self, consumer: C) -> C::Result
            where
                C: UnindexedConsumer<Self::Item>,
            {
                bridge(self, consumer)
            }

            fn opt_len(&self) -> Option<usize> {
                Some(self.len())
            }
        }

        impl IndexedParallelIterator for Iter<$t> {
            fn drive<C>(self, consumer: C) -> C::Result
            where
                C: Consumer<Self::Item>,
            {
                bridge(self, consumer)
            }

            fn len(&self) -> usize {
                self.range.len()
            }

            fn with_producer<CB>(self, callback: CB) -> CB::Output
            where
                CB: ProducerCallback<Self::Item>,
            {
                callback.callback(IterProducer { range: self.range })
            }
        }

        impl Producer for IterProducer<$t> {
            type Item = <Range<$t> as Iterator>::Item;
            type IntoIter = Range<$t>;
            fn into_iter(self) -> Self::IntoIter {
                self.range
            }

            fn split_at(self, index: usize) -> (Self, Self) {
                assert!(index <= self.range.len());
                // For signed $t, the length and requested index could be greater than $t::MAX, and
                // then `index as $t` could wrap to negative, so wrapping_add is necessary.
                let mid = self.range.start.wrapping_add(index as $t);
                let left = self.range.start..mid;
                let right = mid..self.range.end;
                (IterProducer { range: left }, IterProducer { range: right })
            }
        }
    };
}

trait UnindexedRangeLen<L> {
    fn len(&self) -> L;
}

macro_rules! unindexed_range_impl {
    ( $t:ty, $len_t:ty ) => {
        impl UnindexedRangeLen<$len_t> for Range<$t> {
            fn len(&self) -> $len_t {
                let &Range { start, end } = self;
                if end > start {
                    end.wrapping_sub(start) as $len_t
                } else {
                    0
                }
            }
        }

        impl ParallelIterator for Iter<$t> {
            type Item = $t;

            fn drive_unindexed<C>(self, consumer: C) -> C::Result
            where
                C: UnindexedConsumer<Self::Item>,
            {
                #[inline]
                fn offset(start: $t) -> impl Fn(usize) -> $t {
                    move |i| start.wrapping_add(i as $t)
                }

                if let Some(len) = self.opt_len() {
                    // Drive this in indexed mode for better `collect`.
                    (0..len)
                        .into_par_iter()
                        .map(offset(self.range.start))
                        .drive(consumer)
                } else {
                    bridge_unindexed(IterProducer { range: self.range }, consumer)
                }
            }

            fn opt_len(&self) -> Option<usize> {
                let len = self.range.len();
                if len <= usize::MAX as $len_t {
                    Some(len as usize)
                } else {
                    None
                }
            }
        }

        impl UnindexedProducer for IterProducer<$t> {
            type Item = $t;

            fn split(mut self) -> (Self, Option<Self>) {
                let index = self.range.len() / 2;
                if index > 0 {
                    let mid = self.range.start.wrapping_add(index as $t);
                    let right = mid..self.range.end;
                    self.range.end = mid;
                    (self, Some(IterProducer { range: right }))
                } else {
                    (self, None)
                }
            }

            fn fold_with<F>(self, folder: F) -> F
            where
                F: Folder<Self::Item>,
            {
                folder.consume_iter(self)
            }
        }
    };
}

// all Range<T> with ExactSizeIterator
indexed_range_impl! {u8}
indexed_range_impl! {u16}
indexed_range_impl! {u32}
indexed_range_impl! {usize}
indexed_range_impl! {i8}
indexed_range_impl! {i16}
indexed_range_impl! {i32}
indexed_range_impl! {isize}

// other Range<T> with just Iterator
unindexed_range_impl! {u64, u64}
unindexed_range_impl! {i64, u64}
unindexed_range_impl! {u128, u128}
unindexed_range_impl! {i128, u128}

#[test]
fn check_range_split_at_overflow() {
    // Note, this split index overflows i8!
    let producer = IterProducer { range: -100i8..100 };
    let (left, right) = producer.split_at(150);
    let r1: i32 = left.range.map(i32::from).sum();
    let r2: i32 = right.range.map(i32::from).sum();
    assert_eq!(r1 + r2, -100);
}

#[test]
fn test_i128_len_doesnt_overflow() {
    use std::{i128, u128};

    // Using parse because some versions of rust don't allow long literals
    let octillion: i128 = "1000000000000000000000000000".parse().unwrap();
    let producer = IterProducer {
        range: 0..octillion,
    };

    assert_eq!(octillion as u128, producer.range.len());
    assert_eq!(octillion as u128, (0..octillion).len());
    assert_eq!(2 * octillion as u128, (-octillion..octillion).len());

    assert_eq!(u128::MAX, (i128::MIN..i128::MAX).len());
}

#[test]
fn test_u64_opt_len() {
    use std::{u64, usize};
    assert_eq!(Some(100), (0..100u64).into_par_iter().opt_len());
    assert_eq!(
        Some(usize::MAX),
        (0..usize::MAX as u64).into_par_iter().opt_len()
    );
    if (usize::MAX as u64) < u64::MAX {
        assert_eq!(
            None,
            (0..(usize::MAX as u64).wrapping_add(1))
                .into_par_iter()
                .opt_len()
        );
        assert_eq!(None, (0..u64::MAX).into_par_iter().opt_len());
    }
}

#[test]
fn test_u128_opt_len() {
    use std::{u128, usize};
    assert_eq!(Some(100), (0..100u128).into_par_iter().opt_len());
    assert_eq!(
        Some(usize::MAX),
        (0..usize::MAX as u128).into_par_iter().opt_len()
    );
    assert_eq!(None, (0..1 + usize::MAX as u128).into_par_iter().opt_len());
    assert_eq!(None, (0..u128::MAX).into_par_iter().opt_len());
}

// `usize as i64` can overflow, so make sure to wrap it appropriately
// when using the `opt_len` "indexed" mode.
#[test]
#[cfg(target_pointer_width = "64")]
fn test_usize_i64_overflow() {
    use std::i64;
    use ThreadPoolBuilder;

    let iter = (-2..i64::MAX).into_par_iter();
    assert_eq!(iter.opt_len(), Some(i64::MAX as usize + 2));

    // always run with multiple threads to split into, or this will take forever...
    let pool = ThreadPoolBuilder::new().num_threads(8).build().unwrap();
    pool.install(|| assert_eq!(iter.find_last(|_| true), Some(i64::MAX - 1)));
}