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::ExecutionCtx;
12use vortex_array::IntoArray;
13use vortex_array::arrays::DictArray;
14use vortex_array::arrays::PrimitiveArray;
15use vortex_array::arrays::dict::DictArrayExt;
16use vortex_array::arrays::dict::DictArraySlotsExt;
17use vortex_array::arrays::primitive::PrimitiveArrayExt;
18use vortex_array::builders::dict::dict_encode;
19use vortex_error::VortexExpect;
20use vortex_error::VortexResult;
21
22use crate::CascadingCompressor;
23use crate::builtins::IntDictScheme;
24use crate::builtins::StringDictScheme;
25use crate::builtins::is_utf8_string;
26use crate::ctx::CompressorContext;
27use crate::estimate::CompressionEstimate;
28use crate::estimate::DeferredEstimate;
29use crate::estimate::EstimateVerdict;
30use crate::scheme::ChildSelection;
31use crate::scheme::DescendantExclusion;
32use crate::scheme::Scheme;
33use crate::scheme::SchemeExt;
34use crate::stats::ArrayAndStats;
35use crate::stats::GenerateStatsOptions;
36
37impl Scheme for StringDictScheme {
38    fn scheme_name(&self) -> &'static str {
39        "vortex.string.dict"
40    }
41
42    fn matches(&self, canonical: &Canonical) -> bool {
43        is_utf8_string(canonical)
44    }
45
46    fn stats_options(&self) -> GenerateStatsOptions {
47        GenerateStatsOptions {
48            count_distinct_values: true,
49        }
50    }
51
52    /// Children: values=0, codes=1.
53    fn num_children(&self) -> usize {
54        2
55    }
56
57    /// String dict codes (child 1) are compact unsigned integers that should not be dict-encoded
58    /// again.
59    ///
60    /// Additional exclusions for codes (IntSequenceScheme, FoRScheme, ZigZagScheme, SparseScheme,
61    /// RunEndScheme, RLE, etc.) are expressed as pull rules on those schemes in `vortex-btrblocks`.
62    fn descendant_exclusions(&self) -> Vec<DescendantExclusion> {
63        vec![DescendantExclusion {
64            excluded: IntDictScheme.id(),
65            children: ChildSelection::One(1),
66        }]
67    }
68
69    fn expected_compression_ratio(
70        &self,
71        data: &ArrayAndStats,
72        _compress_ctx: CompressorContext,
73        exec_ctx: &mut ExecutionCtx,
74    ) -> CompressionEstimate {
75        let stats = data.string_stats(exec_ctx);
76
77        if stats.value_count() == 0 {
78            return CompressionEstimate::Verdict(EstimateVerdict::Skip);
79        }
80
81        let estimated_distinct_values_count = stats.estimated_distinct_count().vortex_expect(
82            "this must be present since `DictScheme` declared that we need distinct values",
83        );
84
85        // If > 50% of the values are distinct, skip dictionary scheme.
86        if estimated_distinct_values_count > stats.value_count() / 2 {
87            return CompressionEstimate::Verdict(EstimateVerdict::Skip);
88        }
89
90        // Let sampling determine the expected ratio.
91        CompressionEstimate::Deferred(DeferredEstimate::Sample)
92    }
93
94    fn compress(
95        &self,
96        compressor: &CascadingCompressor,
97        data: &ArrayAndStats,
98        compress_ctx: CompressorContext,
99        exec_ctx: &mut ExecutionCtx,
100    ) -> VortexResult<ArrayRef> {
101        let dict = dict_encode(data.array())?;
102
103        // Values = child 0.
104        let compressed_values =
105            compressor.compress_child(dict.values(), &compress_ctx, self.id(), 0, exec_ctx)?;
106
107        // Codes = child 1.
108        let narrowed_codes = dict
109            .codes()
110            .clone()
111            .execute::<PrimitiveArray>(exec_ctx)?
112            .narrow()?
113            .into_array();
114        let compressed_codes =
115            compressor.compress_child(&narrowed_codes, &compress_ctx, self.id(), 1, exec_ctx)?;
116
117        // SAFETY: compressing codes or values does not alter the invariants.
118        unsafe {
119            Ok(
120                DictArray::new_unchecked(compressed_codes, compressed_values)
121                    .set_all_values_referenced(dict.has_all_values_referenced())
122                    .into_array(),
123            )
124        }
125    }
126}