vortex_btrblocks/integer/
dictionary.rs

1// SPDX-License-Identifier: Apache-2.0
2// SPDX-FileCopyrightText: Copyright the Vortex contributors
3
4//! Dictionary compressor that reuses the unique values in the `IntegerStats`.
5
6use vortex_array::IntoArray;
7use vortex_array::arrays::PrimitiveArray;
8use vortex_array::validity::Validity;
9use vortex_array::vtable::ValidityHelper;
10use vortex_buffer::Buffer;
11use vortex_dict::DictArray;
12use vortex_error::VortexResult;
13
14use crate::integer::IntegerStats;
15use crate::integer::stats::ErasedStats;
16
17macro_rules! typed_encode {
18    ($stats:ident, $typed:ident, $validity:ident, $typ:ty) => {{
19        let values: Buffer<$typ> = $typed.distinct_values.keys().map(|x| x.0).collect();
20
21        let max_code = values.len();
22        let codes = if max_code <= u8::MAX as usize {
23            let buf =
24                <DictEncoder as Encode<$typ, u8>>::encode(&values, $stats.src.as_slice::<$typ>());
25            PrimitiveArray::new(buf, $validity.clone()).into_array()
26        } else if max_code <= u16::MAX as usize {
27            let buf =
28                <DictEncoder as Encode<$typ, u16>>::encode(&values, $stats.src.as_slice::<$typ>());
29            PrimitiveArray::new(buf, $validity.clone()).into_array()
30        } else {
31            let buf =
32                <DictEncoder as Encode<$typ, u32>>::encode(&values, $stats.src.as_slice::<$typ>());
33            PrimitiveArray::new(buf, $validity.clone()).into_array()
34        };
35
36        let values_validity = match $validity {
37            Validity::NonNullable => Validity::NonNullable,
38            _ => Validity::AllValid,
39        };
40
41        let values = PrimitiveArray::new(values, values_validity).into_array();
42        DictArray::try_new(codes, values)
43    }};
44}
45
46#[allow(clippy::cognitive_complexity)]
47pub fn dictionary_encode(stats: &IntegerStats) -> VortexResult<DictArray> {
48    // We need to preserve the nullability somehow from the original
49    let src_validity = stats.src.validity();
50
51    match &stats.typed {
52        ErasedStats::U8(typed) => typed_encode!(stats, typed, src_validity, u8),
53        ErasedStats::U16(typed) => typed_encode!(stats, typed, src_validity, u16),
54        ErasedStats::U32(typed) => typed_encode!(stats, typed, src_validity, u32),
55        ErasedStats::U64(typed) => typed_encode!(stats, typed, src_validity, u64),
56        ErasedStats::I8(typed) => typed_encode!(stats, typed, src_validity, i8),
57        ErasedStats::I16(typed) => typed_encode!(stats, typed, src_validity, i16),
58        ErasedStats::I32(typed) => typed_encode!(stats, typed, src_validity, i32),
59        ErasedStats::I64(typed) => typed_encode!(stats, typed, src_validity, i64),
60    }
61}
62
63struct DictEncoder;
64
65trait Encode<T, I> {
66    /// Using the distinct value set, turn the values into a set of codes.
67    fn encode(distinct: &[T], values: &[T]) -> Buffer<I>;
68}
69
70macro_rules! impl_encode {
71    ($typ:ty) => { impl_encode!($typ, u8, u16, u32); };
72    ($typ:ty, $($ityp:ty),+) => {
73        $(
74        impl Encode<$typ, $ityp> for DictEncoder {
75            #[allow(clippy::cast_possible_truncation)]
76            fn encode(distinct: &[$typ], values: &[$typ]) -> Buffer<$ityp> {
77                let mut codes =
78                    vortex_utils::aliases::hash_map::HashMap::<$typ, $ityp>::with_capacity(
79                        distinct.len(),
80                    );
81                for (code, &value) in distinct.iter().enumerate() {
82                    codes.insert(value, code as $ityp);
83                }
84
85                let mut output = vortex_buffer::BufferMut::with_capacity(values.len());
86                for value in values {
87                    // Any code lookups which fail are for nulls, so their value
88                    // does not matter.
89                    // SAFETY: we have exactly sized output to be as large as values.
90                    unsafe { output.push_unchecked(codes.get(value).copied().unwrap_or_default()) };
91                }
92
93                return output.freeze();
94            }
95        }
96        )*
97    };
98}
99
100impl_encode!(u8);
101impl_encode!(u16);
102impl_encode!(u32);
103impl_encode!(u64);
104impl_encode!(i8);
105impl_encode!(i16);
106impl_encode!(i32);
107impl_encode!(i64);
108
109#[cfg(test)]
110mod tests {
111    use vortex_array::arrays::{BoolArray, PrimitiveArray};
112    use vortex_array::validity::Validity;
113    use vortex_array::{Array, IntoArray, ToCanonical};
114    use vortex_buffer::buffer;
115
116    use crate::CompressorStats;
117    use crate::integer::IntegerStats;
118    use crate::integer::dictionary::dictionary_encode;
119
120    #[test]
121    fn test_dict_encode_integer_stats() {
122        // Create an array that has some nulls
123        let data = buffer![100i32, 200, 100, 0, 100];
124        let validity =
125            Validity::Array(BoolArray::from_iter([true, true, true, false, true]).into_array());
126        let array = PrimitiveArray::new(data, validity);
127
128        let stats = IntegerStats::generate(&array);
129        let dict_array = dictionary_encode(&stats).unwrap();
130        assert_eq!(dict_array.values().len(), 2);
131        assert_eq!(dict_array.codes().len(), 5);
132
133        let undict = dict_array.to_primitive().unwrap();
134
135        // We just use code zero, but it doesn't really matter.
136        // We can just shove a whole validity buffer in there instead.
137        assert_eq!(undict.as_slice::<i32>(), &[100i32, 200, 100, 100, 100]);
138    }
139}
vortex_btrblocks/integer/dictionary.rs

vortex_btrblocks/integer/
dictionary.rs
