Skip to main content

vortex_compressor/builtins/dict/
string.rs

1// SPDX-License-Identifier: Apache-2.0
2// SPDX-FileCopyrightText: Copyright the Vortex contributors
3
4//! UTF8-specific dictionary encoding implementation.
5//!
6//! Vortex encoders must always produce unsigned integer codes; signed codes are only accepted
7//! for external compatibility.
8
9use vortex_array::ArrayRef;
10use vortex_array::Canonical;
11use vortex_array::IntoArray;
12use vortex_array::arrays::DictArray;
13use vortex_array::arrays::PrimitiveArray;
14use vortex_array::arrays::dict::DictArrayExt;
15use vortex_array::arrays::dict::DictArraySlotsExt;
16use vortex_array::arrays::primitive::PrimitiveArrayExt;
17use vortex_array::builders::dict::dict_encode;
18use vortex_error::VortexExpect;
19use vortex_error::VortexResult;
20
21use crate::CascadingCompressor;
22use crate::builtins::IntDictScheme;
23use crate::builtins::StringDictScheme;
24use crate::builtins::is_utf8_string;
25use crate::ctx::CompressorContext;
26use crate::estimate::CompressionEstimate;
27use crate::scheme::ChildSelection;
28use crate::scheme::DescendantExclusion;
29use crate::scheme::Scheme;
30use crate::scheme::SchemeExt;
31use crate::stats::ArrayAndStats;
32use crate::stats::GenerateStatsOptions;
33
34impl Scheme for StringDictScheme {
35    fn scheme_name(&self) -> &'static str {
36        "vortex.string.dict"
37    }
38
39    fn matches(&self, canonical: &Canonical) -> bool {
40        is_utf8_string(canonical)
41    }
42
43    fn stats_options(&self) -> GenerateStatsOptions {
44        GenerateStatsOptions {
45            count_distinct_values: true,
46        }
47    }
48
49    /// Children: values=0, codes=1.
50    fn num_children(&self) -> usize {
51        2
52    }
53
54    /// String dict codes (child 1) are compact unsigned integers that should not be dict-encoded
55    /// again.
56    ///
57    /// Additional exclusions for codes (IntSequenceScheme, FoRScheme, ZigZagScheme, SparseScheme,
58    /// RunEndScheme, RLE, etc.) are expressed as pull rules on those schemes in `vortex-btrblocks`.
59    fn descendant_exclusions(&self) -> Vec<DescendantExclusion> {
60        vec![DescendantExclusion {
61            excluded: IntDictScheme.id(),
62            children: ChildSelection::One(1),
63        }]
64    }
65
66    fn expected_compression_ratio(
67        &self,
68        data: &mut ArrayAndStats,
69        _ctx: CompressorContext,
70    ) -> CompressionEstimate {
71        let stats = data.string_stats();
72
73        if stats.value_count() == 0 {
74            return CompressionEstimate::Skip;
75        }
76
77        let estimated_distinct_values_count = stats.estimated_distinct_count().vortex_expect(
78            "this must be present since `DictScheme` declared that we need distinct values",
79        );
80
81        // If > 50% of the values are distinct, skip dictionary scheme.
82        if estimated_distinct_values_count > stats.value_count() / 2 {
83            return CompressionEstimate::Skip;
84        }
85
86        // Let sampling determine the expected ratio.
87        CompressionEstimate::Sample
88    }
89
90    fn compress(
91        &self,
92        compressor: &CascadingCompressor,
93        data: &mut ArrayAndStats,
94        ctx: CompressorContext,
95    ) -> VortexResult<ArrayRef> {
96        let dict = dict_encode(data.array())?;
97
98        // Values = child 0.
99        let compressed_values = compressor.compress_child(dict.values(), &ctx, self.id(), 0)?;
100
101        // Codes = child 1.
102        let narrowed_codes = dict
103            .codes()
104            .clone()
105            .execute::<PrimitiveArray>(&mut compressor.execution_ctx())?
106            .narrow()?
107            .into_array();
108        let compressed_codes = compressor.compress_child(&narrowed_codes, &ctx, self.id(), 1)?;
109
110        // SAFETY: compressing codes or values does not alter the invariants.
111        unsafe {
112            Ok(
113                DictArray::new_unchecked(compressed_codes, compressed_values)
114                    .set_all_values_referenced(dict.has_all_values_referenced())
115                    .into_array(),
116            )
117        }
118    }
119}