dataprof 0.7.1

High-performance data profiler with ISO 8000/25012 quality metrics for CSV, JSON/JSONL, and Parquet files
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

/// SIMD-accelerated numerical computations for data profiling
/// Uses the `wide` crate for portable SIMD operations
use wide::*;

/// SIMD-accelerated statistical computations
pub struct SimdStats {
    pub count: usize,
    pub sum: f64,
    pub sum_squares: f64,
    pub min: f64,
    pub max: f64,
}

impl Default for SimdStats {
    fn default() -> Self {
        Self::new()
    }
}

impl SimdStats {
    pub fn new() -> Self {
        Self {
            count: 0,
            sum: 0.0,
            sum_squares: 0.0,
            min: f64::INFINITY,
            max: f64::NEG_INFINITY,
        }
    }

    pub fn mean(&self) -> f64 {
        if self.count > 0 {
            self.sum / self.count as f64
        } else {
            0.0
        }
    }

    pub fn variance(&self) -> f64 {
        if self.count <= 1 {
            return 0.0;
        }

        let n = self.count as f64;
        let mean = self.mean();
        (self.sum_squares - n * mean * mean) / n
    }

    pub fn std_dev(&self) -> f64 {
        self.variance().sqrt()
    }
}

/// Compute basic statistics using SIMD operations
pub fn compute_stats_simd(values: &[f64]) -> SimdStats {
    if values.is_empty() {
        return SimdStats::new();
    }

    let mut stats = SimdStats::new();
    stats.count = values.len();

    // Process in SIMD chunks of 4 f64 values
    let chunk_size = 4;
    let chunks = values.chunks_exact(chunk_size);
    let remainder = chunks.remainder();

    // SIMD variables for accumulation
    let mut sum_vec = f64x4::splat(0.0);
    let mut sum_squares_vec = f64x4::splat(0.0);
    let mut min_vec = f64x4::splat(f64::INFINITY);
    let mut max_vec = f64x4::splat(f64::NEG_INFINITY);

    // Process SIMD chunks
    for chunk in chunks {
        let vec = f64x4::new([chunk[0], chunk[1], chunk[2], chunk[3]]);

        // Accumulate sum
        sum_vec += vec;

        // Accumulate sum of squares
        sum_squares_vec += vec * vec;

        // Update min/max
        min_vec = min_vec.min(vec);
        max_vec = max_vec.max(vec);
    }

    // Reduce SIMD vectors to scalar values
    let sum_array: [f64; 4] = sum_vec.to_array();
    let sum_squares_array: [f64; 4] = sum_squares_vec.to_array();
    let min_array: [f64; 4] = min_vec.to_array();
    let max_array: [f64; 4] = max_vec.to_array();

    stats.sum = sum_array[0] + sum_array[1] + sum_array[2] + sum_array[3];
    stats.sum_squares =
        sum_squares_array[0] + sum_squares_array[1] + sum_squares_array[2] + sum_squares_array[3];
    stats.min = min_array[0]
        .min(min_array[1])
        .min(min_array[2])
        .min(min_array[3]);
    stats.max = max_array[0]
        .max(max_array[1])
        .max(max_array[2])
        .max(max_array[3]);

    // Process remainder values (non-SIMD)
    for &value in remainder {
        stats.sum += value;
        stats.sum_squares += value * value;
        stats.min = stats.min.min(value);
        stats.max = stats.max.max(value);
    }

    stats
}

/// Fast parallel sum using SIMD
pub fn sum_simd(values: &[f64]) -> f64 {
    if values.is_empty() {
        return 0.0;
    }

    let chunk_size = 4;
    let chunks = values.chunks_exact(chunk_size);
    let remainder = chunks.remainder();

    let mut sum_vec = f64x4::splat(0.0);

    for chunk in chunks {
        let vec = f64x4::new([chunk[0], chunk[1], chunk[2], chunk[3]]);
        sum_vec += vec;
    }

    let sum_array: [f64; 4] = sum_vec.to_array();
    let mut total = sum_array[0] + sum_array[1] + sum_array[2] + sum_array[3];

    // Add remainder
    for &value in remainder {
        total += value;
    }

    total
}

/// Fast min/max finding using SIMD
pub fn min_max_simd(values: &[f64]) -> (f64, f64) {
    if values.is_empty() {
        return (0.0, 0.0);
    }

    if values.len() == 1 {
        return (values[0], values[0]);
    }

    let chunk_size = 4;
    let chunks = values.chunks_exact(chunk_size);
    let remainder = chunks.remainder();

    let mut min_vec = f64x4::splat(f64::INFINITY);
    let mut max_vec = f64x4::splat(f64::NEG_INFINITY);

    for chunk in chunks {
        let vec = f64x4::new([chunk[0], chunk[1], chunk[2], chunk[3]]);
        min_vec = min_vec.min(vec);
        max_vec = max_vec.max(vec);
    }

    let min_array: [f64; 4] = min_vec.to_array();
    let max_array: [f64; 4] = max_vec.to_array();

    let mut min_val = min_array[0]
        .min(min_array[1])
        .min(min_array[2])
        .min(min_array[3]);
    let mut max_val = max_array[0]
        .max(max_array[1])
        .max(max_array[2])
        .max(max_array[3]);

    // Process remainder
    for &value in remainder {
        min_val = min_val.min(value);
        max_val = max_val.max(value);
    }

    (min_val, max_val)
}

/// SIMD-accelerated dot product (useful for correlation computations)
pub fn dot_product_simd(a: &[f64], b: &[f64]) -> f64 {
    assert_eq!(a.len(), b.len());

    if a.is_empty() {
        return 0.0;
    }

    let chunk_size = 4;
    let chunks_a = a.chunks_exact(chunk_size);
    let chunks_b = b.chunks_exact(chunk_size);
    let remainder_a = chunks_a.remainder();
    let remainder_b = chunks_b.remainder();

    let mut dot_vec = f64x4::splat(0.0);

    for (chunk_a, chunk_b) in chunks_a.zip(chunks_b) {
        let vec_a = f64x4::new([chunk_a[0], chunk_a[1], chunk_a[2], chunk_a[3]]);
        let vec_b = f64x4::new([chunk_b[0], chunk_b[1], chunk_b[2], chunk_b[3]]);

        dot_vec += vec_a * vec_b;
    }

    let dot_array: [f64; 4] = dot_vec.to_array();
    let mut result = dot_array[0] + dot_array[1] + dot_array[2] + dot_array[3];

    // Process remainder
    for (&val_a, &val_b) in remainder_a.iter().zip(remainder_b.iter()) {
        result += val_a * val_b;
    }

    result
}

/// Check if SIMD is beneficial for the given data size
pub fn should_use_simd(data_size: usize) -> bool {
    // SIMD is beneficial for larger datasets due to setup overhead
    data_size >= 64
}

/// Auto-choose between SIMD and regular computation
pub fn compute_stats_auto(values: &[f64]) -> SimdStats {
    if should_use_simd(values.len()) && is_simd_available() {
        compute_stats_simd(values)
    } else {
        compute_stats_fallback(values)
    }
}

/// Check if SIMD is available on current platform
pub fn is_simd_available() -> bool {
    // The wide crate handles platform detection internally
    // For now, we assume SIMD is available on most modern platforms
    true
}

/// Fallback non-SIMD implementation
fn compute_stats_fallback(values: &[f64]) -> SimdStats {
    let mut stats = SimdStats::new();
    stats.count = values.len();

    if values.is_empty() {
        return stats;
    }

    stats.min = values[0];
    stats.max = values[0];

    for &value in values {
        stats.sum += value;
        stats.sum_squares += value * value;
        stats.min = stats.min.min(value);
        stats.max = stats.max.max(value);
    }

    stats
}

#[cfg(test)]
mod tests {
    use super::*;

    #[test]
    fn test_simd_stats() {
        let values = vec![1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0];
        let stats = compute_stats_simd(&values);

        assert_eq!(stats.count, 8);
        assert!((stats.sum - 36.0).abs() < 1e-10);
        assert_eq!(stats.min, 1.0);
        assert_eq!(stats.max, 8.0);
        assert!((stats.mean() - 4.5).abs() < 1e-10);
    }

    #[test]
    fn test_simd_vs_fallback() {
        let values: Vec<f64> = (1..=100).map(|x| x as f64).collect();

        let simd_stats = compute_stats_simd(&values);
        let fallback_stats = compute_stats_fallback(&values);

        assert!((simd_stats.sum - fallback_stats.sum).abs() < 1e-10);
        assert!((simd_stats.min - fallback_stats.min).abs() < 1e-10);
        assert!((simd_stats.max - fallback_stats.max).abs() < 1e-10);
        assert!((simd_stats.mean() - fallback_stats.mean()).abs() < 1e-10);
    }

    #[test]
    fn test_min_max_simd() {
        let values = vec![3.0, 1.0, 4.0, 1.0, 5.0, 9.0, 2.0, 6.0];
        let (min, max) = min_max_simd(&values);

        assert_eq!(min, 1.0);
        assert_eq!(max, 9.0);
    }

    #[test]
    fn test_dot_product_simd() {
        let a = vec![1.0, 2.0, 3.0, 4.0];
        let b = vec![5.0, 6.0, 7.0, 8.0];
        let dot = dot_product_simd(&a, &b);

        // 1*5 + 2*6 + 3*7 + 4*8 = 5 + 12 + 21 + 32 = 70
        assert!((dot - 70.0).abs() < 1e-10);
    }

    #[test]
    fn test_auto_selection() {
        let small_values = vec![1.0, 2.0, 3.0];
        let large_values: Vec<f64> = (1..=1000).map(|x| x as f64).collect();

        // Both should work correctly regardless of implementation chosen
        let small_stats = compute_stats_auto(&small_values);
        let large_stats = compute_stats_auto(&large_values);

        assert!((small_stats.mean() - 2.0).abs() < 1e-10);
        assert!((large_stats.mean() - 500.5).abs() < 1e-10);
    }
}