numeric 0.1.4

N-dimensional matrix class for Rust
Documentation 
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
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391

use tensor::Tensor;
use traits::TensorTrait;
use std::ops::{Add, Sub, Mul, Div, Rem, Neg, BitAnd, BitOr, BitXor};
use std::cmp::max;

fn compatible_shapes_for_elementwise_op(sh1: &[usize], sh2: &[usize]) -> bool {
    let mut ok = true;
    if sh1.len() == sh2.len() {
        for i in 0..sh1.len() {
            ok = ok && (sh1[i] == sh2[i] || sh1[i] == 1 || sh2[i] == 1);
        }
    } else if sh1.len() > sh2.len() {
        let diff = sh1.len() - sh2.len();
        for i in 0..sh2.len() {
            ok = ok && (sh1[i+diff] == sh2[i] || sh1[i+diff] == 1 || sh2[i] == 1);
        }
    } else if sh2.len() > sh1.len() {
        let diff = sh2.len() - sh1.len();
        for i in 0..sh1.len() {
            ok = ok && (sh1[i] == sh2[i+diff] || sh1[i] == 1 || sh2[i+diff] == 1);
        }
    }
    ok
}

fn shape_for_elementwise_op(sh1: &[usize], sh2: &[usize]) -> Vec<usize> {
    let mut out_sh = vec![0usize; max(sh2.len(), sh1.len())];
    if sh1.len() == sh2.len() {
        for i in 0..sh1.len() {
            out_sh[i] = max(sh1[i], sh2[i]);
        }
    } else if sh1.len() > sh2.len() {
        assert!(false);
    } else if sh2.len() > sh1.len() {
        assert!(false);
    }
    out_sh
}

/// Help function for test_compatible_shapes_for_elementwise_op, since it is a
/// symmetric function.
#[allow(dead_code)]
fn test_compatible_assert_symmetric(sh1: &[usize], sh2: &[usize], expected: bool) -> () {
    assert!(compatible_shapes_for_elementwise_op(sh1, sh2) == expected);
    assert!(compatible_shapes_for_elementwise_op(sh2, sh1) == expected);
}

#[test]
fn test_compatible_shapes_for_elementwise_op() {
    // Compatible test cases
    test_compatible_assert_symmetric(&[2, 30], &[2, 30], true);
    test_compatible_assert_symmetric(&[2, 3, 4], &[2, 3, 4], true);
    test_compatible_assert_symmetric(&[1, 1, 1], &[1, 1, 1], true);
    test_compatible_assert_symmetric(&[2, 3, 4, 5], &[2, 3, 4, 5], true);

    // One broadcasted axis
    test_compatible_assert_symmetric(&[1, 3, 4], &[2, 3, 4], true);
    test_compatible_assert_symmetric(&[10, 1, 5], &[10, 2, 5], true);
    test_compatible_assert_symmetric(&[10, 20, 1], &[10, 20, 100], true);
    test_compatible_assert_symmetric(&[10, 20, 1, 300, 2], &[10, 20, 100, 300, 2], true);

    // Two broadcasted axes
    test_compatible_assert_symmetric(&[1, 1, 4], &[2, 3, 4], true);
    test_compatible_assert_symmetric(&[10, 1, 1], &[10, 2, 5], true);
    test_compatible_assert_symmetric(&[1, 20, 1], &[10, 20, 100], true);
    test_compatible_assert_symmetric(&[1, 20, 1, 300, 2], &[10, 20, 100, 300, 2], true);

    // Mixed broadcasting
    test_compatible_assert_symmetric(&[10, 1, 100, 300, 2], &[10, 20, 1, 300, 2], true);
    test_compatible_assert_symmetric(&[1, 5, 100, 1, 1], &[10, 1, 1, 300, 2], true);
    test_compatible_assert_symmetric(&[1, 1, 100, 1, 1], &[10, 1, 1, 300, 2], true);

    // One is lower dimensional
    test_compatible_assert_symmetric(&[1, 100, 300, 2], &[10, 20, 1, 300, 2], true);
    test_compatible_assert_symmetric(&[100, 300, 2], &[10, 20, 1, 300, 2], true);
    test_compatible_assert_symmetric(&[300, 2], &[10, 20, 1, 300, 2], true);
    test_compatible_assert_symmetric(&[2], &[10, 20, 1, 300, 2], true);
    test_compatible_assert_symmetric(&[1], &[10, 20, 1, 300, 2], true);

    // Scalars
    test_compatible_assert_symmetric(&[], &[10, 20, 1, 300, 2], true);
    test_compatible_assert_symmetric(&[], &[], true);

    // Incompatible test cases
    test_compatible_assert_symmetric(&[2, 3, 4], &[3, 2, 4], false);
    test_compatible_assert_symmetric(&[2, 3], &[3, 3], false);
    test_compatible_assert_symmetric(&[5], &[3], false);
    test_compatible_assert_symmetric(&[1, 5], &[3], false);
    test_compatible_assert_symmetric(&[1, 5], &[1, 3], false);
    test_compatible_assert_symmetric(&[1, 2, 3, 4], &[5, 4], false);
}

macro_rules! add_impl {
    ($trait_name:ident, $func_name:ident, $func_name_with_mul:ident) => (
        // T <op> T
        impl<T: TensorTrait + $trait_name<Output=T>> $trait_name for Tensor<T> {
            type Output = Tensor<T>;
            fn $func_name(mut self, rhs: Self::Output) -> Self::Output {
                if rhs.is_scalar() {
                    self.canonize_inplace();
                    {
                        let n = self.size();
                        let mut data = self.slice_mut();
                        let v: T = rhs.scalar_value() as T;
                        for i in 0..n {
                            data[i] = data[i].$func_name(v);
                        }
                    }
                    self
                } else if self.is_scalar() {
                    let mut t = Tensor::empty(&rhs.shape);
                    {
                        let n = t.size();
                        let mut data = t.slice_mut();
                        let v = self.scalar_value();
                        for i in 0..n {
                            data[i] = v.$func_name(rhs.data[i]);
                        }
                    }
                    t
                } else {
                    self.canonize_inplace();
                    {
                        let n = self.size();
                        assert_eq!(self.shape, rhs.shape);
                        let mut data = self.slice_mut();
                        for i in 0..n {
                            data[i] = data[i].$func_name(rhs.data[i]);
                        }
                    }
                    self
                }
            }
        }

        // T <op> &T
        impl<'a, T: TensorTrait + $trait_name<Output=T>> $trait_name<&'a Tensor<T>> for Tensor<T> {
            type Output = Tensor<T>;
            fn $func_name(mut self, rhs: &Self::Output) -> Self::Output {
                if rhs.is_scalar() {
                    self.canonize_inplace();
                    {
                        let n = self.size();
                        let mut data = self.slice_mut();
                        let v = rhs.scalar_value();
                        for i in 0..n {
                            data[i] = data[i].$func_name(v);
                        }
                    }
                    self
                } else if self.is_scalar() {
                    let mut t: Tensor<T> = Tensor::empty(&rhs.shape);
                    {
                        let n = t.size();
                        let mut data = t.slice_mut();
                        let v = self.scalar_value();
                        for i in 0..n {
                            data[i] = v.$func_name(rhs.data[i]);
                        }
                    }
                    t
                } else {
                    let rhs0 = rhs.canonize();
                    assert_eq!(self.shape, rhs0.shape);
                    self.canonize_inplace();
                    let rhs0 = rhs.canonize();
                    {
                        let n = self.size();
                        let mut data = self.slice_mut();
                        for i in 0..n {
                            data[i] = data[i].$func_name(rhs0.data[i]);
                        }
                    }
                    self
                }
            }
        }

        // T <op> &T  (with out)
        impl<T: TensorTrait + $trait_name<Output=T>> Tensor<T> {
            pub fn $func_name_with_mul(&self, rhs: &Tensor<T>, out: &mut Tensor<T>) -> () {
                out.canonize_inplace();
                if rhs.is_scalar() {
                    assert!(out.shape() == self.shape());
                    let n = out.size();
                    let mut data = out.slice_mut();
                    let v = rhs.scalar_value();
                    for i in 0..n {
                        data[i] = data[i].$func_name(v);
                    }
                } else if self.is_scalar() {
                    assert!(out.shape() == rhs.shape());
                    let mut data = out.slice_mut();
                    let v = self.scalar_value();
                    for i in 0..rhs.size() {
                        data[i] = v.$func_name(rhs.data[i]);
                    }
                } else {
                    assert_eq!(self.shape, rhs.shape);
                    let n = out.size();
                    let mut data = out.slice_mut();
                    for i in 0..n {
                        data[i] = self.data[i].$func_name(rhs.data[i]);
                    }
                }
            }
        }

        // &T <op> &T
        impl<'a, T: TensorTrait + $trait_name<Output=T>> $trait_name<&'a Tensor<T>> for &'a Tensor<T> {
            type Output = Tensor<T>;
            fn $func_name(self, rhs: &Self::Output) -> Self::Output {
                if rhs.is_scalar() {
                    let mut t = self.canonize();
                    {
                        let n = t.size();
                        let mut data = t.mem_slice_mut();
                        let v = rhs.scalar_value();
                        for i in 0..n {
                            data[i] = data[i].$func_name(v);
                        }
                    }
                    t
                } else if self.is_scalar() {
                    let mut t = Tensor::empty(&rhs.shape);
                    {
                        let n = t.size();
                        let mut data = t.mem_slice_mut();
                        let v = self.scalar_value();
                        for i in 0..n {
                            data[i] = v.$func_name(rhs.data[i]);
                        }
                    }
                    t
                } else {
                    assert!(compatible_shapes_for_elementwise_op(&self.shape(), &rhs.shape()));

                    let ndim = max(self.ndim(), rhs.ndim());
                    let t1 = self.with_ndim(ndim);
                    let t2 = rhs.with_ndim(ndim);

                    let shape = shape_for_elementwise_op(&t1.shape(), &t2.shape());
                    let mut t: Tensor<T> = Tensor::empty(&shape);

                    let n = t.size();

                    let mut flat_i = t1.mem_offset as isize;
                    let mut flat_j = t2.mem_offset as isize;
                    let mut kk = vec![0isize; ndim];
                    let mut i = 0;
                    {
                        let mut data = t.slice_mut();
                        let t1_data = t1.mem_slice();
                        let t2_data = t2.mem_slice();
                        let mut cur_dim = ndim - 1;
                        while i < n {
                            data[i] = t1_data[flat_i as usize].$func_name(t2_data[flat_j as usize]);
                            if t1.shape[ndim - 1] > 1 {
                                flat_i += t1.strides[ndim - 1];
                            }
                            if t2.shape[ndim - 1] > 1 {
                                flat_j += t2.strides[ndim - 1];
                            }

                            kk[cur_dim] += 1;
                            while kk[cur_dim] == shape[cur_dim] as isize && cur_dim > 0 {
                                kk[cur_dim] = 0;
                                if t1.shape[cur_dim] > 1 {
                                    flat_i -= t1.strides[cur_dim] * (t1.shape[cur_dim] as isize);
                                }
                                if t2.shape[cur_dim] > 1 {
                                    flat_j -= t2.strides[cur_dim] * (t2.shape[cur_dim] as isize);
                                }
                                cur_dim -= 1;
                                kk[cur_dim] += 1;
                                if t1.shape[cur_dim] > 1 {
                                    flat_i += t1.strides[cur_dim];
                                }
                                if t2.shape[cur_dim] > 1 {
                                    flat_j += t2.strides[cur_dim];
                                }
                            }

                            i += 1;
                            cur_dim = ndim - 1;
                        }
                    }
                    t
                }
            }
        }

        // T <op> S
        impl<T: TensorTrait + $trait_name<Output=T>> $trait_name<T> for Tensor<T> {
            type Output = Tensor<T>;
            fn $func_name(mut self, rhs: T) -> Self::Output {
                self.canonize_inplace();
                {
                    let n = self.size();
                    let mut data = self.slice_mut();
                    for i in 0..n {
                        data[i] = data[i].$func_name(rhs);
                    }
                }
                self
            }
        }

        // &T <op> S
        impl<'a, T: TensorTrait + $trait_name<Output=T>> $trait_name<T> for &'a Tensor<T> {
            type Output = Tensor<T>;
            fn $func_name(self, rhs: T) -> Self::Output {
                let mut t = self.canonize();
                {
                    let n = t.size();
                    let mut data = t.mem_slice_mut();
                    for i in 0..n {
                        data[i] = data[i].$func_name(rhs);
                    }
                }
                t
            }
        }
    )
}

// Any operation supported on T should be supported on Tensor<T>, as long as T supports TensorTrait
add_impl!(Add, add, add_with_out);
add_impl!(Sub, sub, sub_with_out);
add_impl!(Mul, mul, mul_with_out);
add_impl!(Div, div, div_with_out);
add_impl!(Rem, rem, rem_with_out);

add_impl!(BitAnd, bitand, bitand_with_out);
add_impl!(BitOr, bitor, bitor_with_out);
add_impl!(BitXor, bitxor, bitxor_with_out);

// -T
impl<T: TensorTrait + Neg<Output=T>> Neg for Tensor<T> {
    type Output = Tensor<T>;
    fn neg(mut self) -> Self::Output {
        self.canonize_inplace();
        {
            let n = self.size();
            let mut data = self.slice_mut();
            for i in 0..n {
                data[i] = -data[i];
            }
        }
        self
    }
}

// -&T
impl<'a, T: TensorTrait + Neg<Output=T>> Neg for &'a Tensor<T> {
    type Output = Tensor<T>;
    fn neg(self) -> Self::Output {
        let mut t = Tensor::empty(&self.shape);
        {
            let mut data = t.slice_mut();
            for (i, v) in self.iter().enumerate() {
                data[i] = -v;
            }
        }
        t
    }
}

/*
#[test]
fn test_negate1() {
    let t0 = tensor![1.0, 10.0; 23.0, -4.0];
    let t1 = -&t0;

    let ta = tensor![-1.0, -10.0; -23.0, 4.0];
    assert!(t1 == ta);
}

#[test]
fn test_negate2() {
    let t0 = Tensor::range(120).reshape(&[5, 4, 2, 3]);
    let t1 = t0.index(&[AxisIndex::Index(3),
                        AxisIndex::StridedSlice(Some(3), None, -2),
                        AxisIndex::Index(1),
                        AxisIndex::StridedSlice(None, None, 2)]);
    let t2 = -&t1;

    let ta = tensor![-93, -95; -81, -83];
    assert!(t2 == ta);
}
*/