vortex_array/arrays/list/vtable/
mod.rs1use std::hash::Hasher;
5use std::sync::Arc;
6
7use prost::Message;
8use vortex_error::VortexExpect;
9use vortex_error::VortexResult;
10use vortex_error::vortex_bail;
11use vortex_error::vortex_ensure;
12use vortex_error::vortex_panic;
13use vortex_session::VortexSession;
14use vortex_session::registry::CachedId;
15
16use crate::ArrayEq;
17use crate::ArrayHash;
18use crate::ArrayRef;
19use crate::EqMode;
20use crate::ExecutionCtx;
21use crate::ExecutionResult;
22use crate::IntoArray;
23use crate::array::Array;
24use crate::array::ArrayId;
25use crate::array::ArrayParts;
26use crate::array::ArrayView;
27use crate::array::VTable;
28use crate::array::with_empty_buffers;
29use crate::arrays::list::ListArraySlotsExt;
30use crate::arrays::list::ListData;
31use crate::arrays::list::ListSlots;
32use crate::arrays::list::compute::rules::PARENT_RULES;
33use crate::arrays::listview::list_view_from_list;
34use crate::buffer::BufferHandle;
35use crate::builders::ArrayBuilder;
36use crate::dtype::DType;
37use crate::dtype::Nullability;
38use crate::dtype::PType;
39use crate::match_each_list_builder;
40use crate::serde::ArrayChildren;
41use crate::validity::Validity;
42mod operations;
43mod validity;
44pub type ListArray = Array<List>;
46
47#[derive(Clone, prost::Message)]
48pub struct ListMetadata {
49 #[prost(uint64, tag = "1")]
50 elements_len: u64,
51 #[prost(enumeration = "PType", tag = "2")]
52 offset_ptype: i32,
53}
54
55impl ArrayHash for ListData {
56 fn array_hash<H: Hasher>(&self, _state: &mut H, _accuracy: EqMode) {}
57}
58
59impl ArrayEq for ListData {
60 fn array_eq(&self, _other: &Self, _accuracy: EqMode) -> bool {
61 true
62 }
63}
64
65impl VTable for List {
66 type TypedArrayData = ListData;
67
68 type OperationsVTable = Self;
69 type ValidityVTable = Self;
70 fn id(&self) -> ArrayId {
71 static ID: CachedId = CachedId::new("vortex.list");
72 *ID
73 }
74
75 fn nbuffers(_array: ArrayView<'_, Self>) -> usize {
76 0
77 }
78
79 fn buffer(_array: ArrayView<'_, Self>, idx: usize) -> BufferHandle {
80 vortex_panic!("ListArray buffer index {idx} out of bounds")
81 }
82
83 fn buffer_name(_array: ArrayView<'_, Self>, idx: usize) -> Option<String> {
84 vortex_panic!("ListArray buffer_name index {idx} out of bounds")
85 }
86
87 fn with_buffers(
88 &self,
89 array: ArrayView<'_, Self>,
90 buffers: &[BufferHandle],
91 ) -> VortexResult<ArrayParts<Self>> {
92 with_empty_buffers(self, array, buffers)
93 }
94
95 fn reduce_parent(
96 array: ArrayView<'_, Self>,
97 parent: &ArrayRef,
98 child_idx: usize,
99 ) -> VortexResult<Option<ArrayRef>> {
100 PARENT_RULES.evaluate(array, parent, child_idx)
101 }
102
103 fn serialize(
104 array: ArrayView<'_, Self>,
105 _session: &VortexSession,
106 ) -> VortexResult<Option<Vec<u8>>> {
107 Ok(Some(
108 ListMetadata {
109 elements_len: array.elements().len() as u64,
110 offset_ptype: PType::try_from(array.offsets().dtype())? as i32,
111 }
112 .encode_to_vec(),
113 ))
114 }
115
116 fn validate(
117 &self,
118 _data: &ListData,
119 dtype: &DType,
120 len: usize,
121 slots: &[Option<ArrayRef>],
122 ) -> VortexResult<()> {
123 vortex_ensure!(
124 slots.len() == ListSlots::COUNT,
125 "ListArray expected {} slots, found {}",
126 ListSlots::COUNT,
127 slots.len()
128 );
129 let elements = slots[ListSlots::ELEMENTS]
130 .as_ref()
131 .vortex_expect("ListArray elements slot");
132 let offsets = slots[ListSlots::OFFSETS]
133 .as_ref()
134 .vortex_expect("ListArray offsets slot");
135 vortex_ensure!(
136 offsets.len().saturating_sub(1) == len,
137 "ListArray length {} does not match outer length {}",
138 offsets.len().saturating_sub(1),
139 len
140 );
141
142 let actual_dtype = DType::List(Arc::new(elements.dtype().clone()), dtype.nullability());
143 vortex_ensure!(
144 &actual_dtype == dtype,
145 "ListArray dtype {} does not match outer dtype {}",
146 actual_dtype,
147 dtype
148 );
149
150 Ok(())
151 }
152
153 fn deserialize(
154 &self,
155 dtype: &DType,
156 len: usize,
157 metadata: &[u8],
158
159 _buffers: &[BufferHandle],
160 children: &dyn ArrayChildren,
161 _session: &VortexSession,
162 ) -> VortexResult<ArrayParts<Self>> {
163 let metadata = ListMetadata::decode(metadata)?;
164 let validity = if children.len() == 2 {
165 Validity::from(dtype.nullability())
166 } else if children.len() == 3 {
167 let validity = children.get(2, &Validity::DTYPE, len)?;
168 Validity::Array(validity)
169 } else {
170 vortex_bail!("Expected 2 or 3 children, got {}", children.len());
171 };
172
173 let DType::List(element_dtype, _) = &dtype else {
174 vortex_bail!("Expected List dtype, got {:?}", dtype);
175 };
176 let elements = children.get(
177 0,
178 element_dtype.as_ref(),
179 usize::try_from(metadata.elements_len)?,
180 )?;
181
182 let offsets = children.get(
183 1,
184 &DType::Primitive(metadata.offset_ptype(), Nullability::NonNullable),
185 len + 1,
186 )?;
187
188 let data = ListData::try_build(elements.clone(), offsets.clone(), validity.clone())?;
189 let slots = ListData::make_slots(&elements, &offsets, &validity, len);
190 Ok(ArrayParts::new(self.clone(), dtype.clone(), len, data).with_slots(slots))
191 }
192
193 fn slot_name(_array: ArrayView<'_, Self>, idx: usize) -> String {
194 ListSlots::NAMES[idx].to_string()
195 }
196
197 fn execute(array: Array<Self>, ctx: &mut ExecutionCtx) -> VortexResult<ExecutionResult> {
198 Ok(ExecutionResult::done(
199 list_view_from_list(array, ctx)?.into_array(),
200 ))
201 }
202
203 #[expect(clippy::cognitive_complexity)]
205 fn append_to_builder(
206 array: ArrayView<'_, Self>,
207 builder: &mut dyn ArrayBuilder,
208 ctx: &mut ExecutionCtx,
209 ) -> VortexResult<()> {
210 match match_each_list_builder!(&mut *builder, |b| b.append_list_array(array, ctx)) {
211 Some(result) => result,
212 None => vortex_bail!(
213 "cannot append a List array of dtype {} to a {} builder",
214 array.dtype(),
215 builder.dtype()
216 ),
217 }
218 }
219}
220
221#[derive(Clone, Debug)]
222pub struct List;