vortex_array/expr/stats/
mod.rs

1// SPDX-License-Identifier: Apache-2.0
2// SPDX-FileCopyrightText: Copyright the Vortex contributors
3
4use std::fmt::Debug;
5use std::fmt::Display;
6use std::fmt::Formatter;
7
8use enum_iterator::Sequence;
9use enum_iterator::all;
10use num_enum::IntoPrimitive;
11use num_enum::TryFromPrimitive;
12use vortex_dtype::DType;
13use vortex_dtype::DecimalDType;
14use vortex_dtype::MAX_PRECISION;
15use vortex_dtype::Nullability::NonNullable;
16use vortex_dtype::Nullability::Nullable;
17use vortex_dtype::PType;
18
19mod bound;
20mod precision;
21mod provider;
22mod stat_bound;
23
24pub use bound::*;
25pub use precision::*;
26pub use provider::*;
27pub use stat_bound::*;
28
29#[derive(
30    Debug,
31    Clone,
32    Copy,
33    PartialEq,
34    Eq,
35    PartialOrd,
36    Ord,
37    Hash,
38    Sequence,
39    IntoPrimitive,
40    TryFromPrimitive,
41)]
42#[repr(u8)]
43pub enum Stat {
44    /// Whether all values are the same (nulls are not equal to other non-null values,
45    /// so this is true iff all values are null or all values are the same non-null value)
46    IsConstant = 0,
47    /// Whether the non-null values in the array are sorted (i.e., we skip nulls)
48    IsSorted = 1,
49    /// Whether the non-null values in the array are strictly sorted (i.e., sorted with no duplicates)
50    IsStrictSorted = 2,
51    /// The maximum value in the array (ignoring nulls, unless all values are null)
52    Max = 3,
53    /// The minimum value in the array (ignoring nulls, unless all values are null)
54    Min = 4,
55    /// The sum of the non-null values of the array.
56    Sum = 5,
57    /// The number of null values in the array
58    NullCount = 6,
59    /// The uncompressed size of the array in bytes
60    UncompressedSizeInBytes = 7,
61    /// The number of NaN values in the array
62    NaNCount = 8,
63}
64
65/// These structs allow the extraction of the bound from the `Precision` value.
66/// They tie together the Stat and the StatBound, which allows the bound to be extracted.
67pub struct Max;
68
69pub struct Min;
70
71pub struct Sum;
72
73pub struct IsConstant;
74
75pub struct IsSorted;
76
77pub struct IsStrictSorted;
78
79pub struct NullCount;
80
81pub struct UncompressedSizeInBytes;
82
83pub struct NaNCount;
84
85impl StatType<bool> for IsConstant {
86    type Bound = Precision<bool>;
87
88    const STAT: Stat = Stat::IsConstant;
89}
90
91impl StatType<bool> for IsSorted {
92    type Bound = Precision<bool>;
93
94    const STAT: Stat = Stat::IsSorted;
95}
96
97impl StatType<bool> for IsStrictSorted {
98    type Bound = Precision<bool>;
99
100    const STAT: Stat = Stat::IsStrictSorted;
101}
102
103impl<T: PartialOrd + Clone> StatType<T> for NullCount {
104    type Bound = UpperBound<T>;
105
106    const STAT: Stat = Stat::NullCount;
107}
108
109impl<T: PartialOrd + Clone> StatType<T> for UncompressedSizeInBytes {
110    type Bound = UpperBound<T>;
111
112    const STAT: Stat = Stat::UncompressedSizeInBytes;
113}
114
115impl<T: PartialOrd + Clone + Debug> StatType<T> for Max {
116    type Bound = UpperBound<T>;
117
118    const STAT: Stat = Stat::Max;
119}
120
121impl<T: PartialOrd + Clone + Debug> StatType<T> for Min {
122    type Bound = LowerBound<T>;
123
124    const STAT: Stat = Stat::Min;
125}
126
127impl<T: PartialOrd + Clone + Debug> StatType<T> for Sum {
128    type Bound = Precision<T>;
129
130    const STAT: Stat = Stat::Sum;
131}
132
133impl<T: PartialOrd + Clone> StatType<T> for NaNCount {
134    type Bound = UpperBound<T>;
135
136    const STAT: Stat = Stat::NaNCount;
137}
138
139impl Stat {
140    /// Whether the statistic is commutative (i.e., whether merging can be done independently of ordering)
141    /// e.g., min/max are commutative, but is_sorted is not
142    pub fn is_commutative(&self) -> bool {
143        // NOTE: we prefer this syntax to force a compile error if we add a new stat
144        match self {
145            Self::IsConstant
146            | Self::Max
147            | Self::Min
148            | Self::NullCount
149            | Self::Sum
150            | Self::NaNCount
151            | Self::UncompressedSizeInBytes => true,
152            Self::IsSorted | Self::IsStrictSorted => false,
153        }
154    }
155
156    /// Whether the statistic has the same dtype as the array it's computed on
157    pub fn has_same_dtype_as_array(&self) -> bool {
158        matches!(self, Stat::Min | Stat::Max)
159    }
160
161    /// Return the [`DType`] of the statistic scalar assuming the array is of the given [`DType`].
162    pub fn dtype(&self, data_type: &DType) -> Option<DType> {
163        Some(match self {
164            Self::IsConstant => DType::Bool(NonNullable),
165            Self::IsSorted => DType::Bool(NonNullable),
166            Self::IsStrictSorted => DType::Bool(NonNullable),
167            Self::Max if matches!(data_type, DType::Null) => return None,
168            Self::Max => data_type.clone(),
169            Self::Min if matches!(data_type, DType::Null) => return None,
170            Self::Min => data_type.clone(),
171            Self::NullCount => DType::Primitive(PType::U64, NonNullable),
172            Self::UncompressedSizeInBytes => DType::Primitive(PType::U64, NonNullable),
173            Self::NaNCount => {
174                // Only floating points support NaN counts.
175                if let DType::Primitive(ptype, ..) = data_type
176                    && ptype.is_float()
177                {
178                    DType::Primitive(PType::U64, NonNullable)
179                } else {
180                    return None;
181                }
182            }
183            Self::Sum => {
184                // Any array that cannot be summed has a sum DType of null.
185                // Any array that can be summed, but overflows, has a sum _value_ of null.
186                // Therefore, we make integer sum stats nullable.
187                match data_type {
188                    DType::Bool(_) => DType::Primitive(PType::U64, Nullable),
189                    DType::Primitive(ptype, _) => match ptype {
190                        PType::U8 | PType::U16 | PType::U32 | PType::U64 => {
191                            DType::Primitive(PType::U64, Nullable)
192                        }
193                        PType::I8 | PType::I16 | PType::I32 | PType::I64 => {
194                            DType::Primitive(PType::I64, Nullable)
195                        }
196                        PType::F16 | PType::F32 | PType::F64 => {
197                            // Float sums cannot overflow, but all null floats still end up as null
198                            DType::Primitive(PType::F64, Nullable)
199                        }
200                    },
201                    DType::Extension(ext_dtype) => self.dtype(ext_dtype.storage_dtype())?,
202                    DType::Decimal(decimal_dtype, _) => {
203                        // Both Spark and DataFusion use this heuristic.
204                        // - https://github.com/apache/spark/blob/fcf636d9eb8d645c24be3db2d599aba2d7e2955a/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/expressions/aggregate/Sum.scala#L66
205                        // - https://github.com/apache/datafusion/blob/4153adf2c0f6e317ef476febfdc834208bd46622/datafusion/functions-aggregate/src/sum.rs#L188
206                        let precision = u8::min(MAX_PRECISION, decimal_dtype.precision() + 10);
207                        DType::Decimal(
208                            DecimalDType::new(precision, decimal_dtype.scale()),
209                            Nullable,
210                        )
211                    }
212                    // Unsupported types
213                    _ => return None,
214                }
215            }
216        })
217    }
218
219    pub fn name(&self) -> &str {
220        match self {
221            Self::IsConstant => "is_constant",
222            Self::IsSorted => "is_sorted",
223            Self::IsStrictSorted => "is_strict_sorted",
224            Self::Max => "max",
225            Self::Min => "min",
226            Self::NullCount => "null_count",
227            Self::UncompressedSizeInBytes => "uncompressed_size_in_bytes",
228            Self::Sum => "sum",
229            Self::NaNCount => "nan_count",
230        }
231    }
232
233    pub fn all() -> impl Iterator<Item = Stat> {
234        all::<Self>()
235    }
236}
237
238impl Display for Stat {
239    fn fmt(&self, f: &mut Formatter<'_>) -> std::fmt::Result {
240        write!(f, "{}", self.name())
241    }
242}
243
244#[cfg(test)]
245mod test {
246    use enum_iterator::all;
247
248    use crate::arrays::PrimitiveArray;
249    use crate::expr::stats::Stat;
250
251    #[test]
252    fn min_of_nulls_is_not_panic() {
253        let min = PrimitiveArray::from_option_iter::<i32, _>([None, None, None, None])
254            .statistics()
255            .compute_as::<i64>(Stat::Min);
256
257        assert_eq!(min, None);
258    }
259
260    #[test]
261    fn has_same_dtype_as_array() {
262        assert!(Stat::Min.has_same_dtype_as_array());
263        assert!(Stat::Max.has_same_dtype_as_array());
264        for stat in all::<Stat>().filter(|s| !matches!(s, Stat::Min | Stat::Max)) {
265            assert!(!stat.has_same_dtype_as_array());
266        }
267    }
268}
vortex_array/expr/stats/mod.rs

vortex_array/expr/stats/
mod.rs
