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_eq;
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::VortexSessionExecute;
24use crate::array::Array;
25use crate::array::ArrayId;
26use crate::array::ArrayParts;
27use crate::array::ArrayView;
28use crate::array::VTable;
29use crate::array::with_empty_buffers;
30use crate::arrays::list::ListArraySlotsExt;
31use crate::arrays::list::ListData;
32use crate::arrays::list::ListSlots;
33use crate::arrays::list::compute::rules::PARENT_RULES;
34use crate::arrays::listview::list_view_from_list;
35use crate::buffer::BufferHandle;
36use crate::builders::ArrayBuilder;
37use crate::dtype::DType;
38use crate::dtype::Nullability;
39use crate::dtype::PType;
40use crate::match_each_list_builder;
41use crate::serde::ArrayChildren;
42use crate::validity::Validity;
43mod operations;
44mod validity;
45pub type ListArray = Array<List>;
47
48#[derive(Clone, prost::Message)]
49pub struct ListMetadata {
50 #[prost(uint64, tag = "1")]
51 elements_len: u64,
52 #[prost(enumeration = "PType", tag = "2")]
53 offset_ptype: i32,
54}
55
56impl ArrayHash for ListData {
57 fn array_hash<H: Hasher>(&self, _state: &mut H, _accuracy: EqMode) {}
58}
59
60impl ArrayEq for ListData {
61 fn array_eq(&self, _other: &Self, _accuracy: EqMode) -> bool {
62 true
63 }
64}
65
66impl VTable for List {
67 type TypedArrayData = ListData;
68
69 type OperationsVTable = Self;
70 type ValidityVTable = Self;
71 fn id(&self) -> ArrayId {
72 static ID: CachedId = CachedId::new("vortex.list");
73 *ID
74 }
75
76 fn nbuffers(_array: ArrayView<'_, Self>) -> usize {
77 0
78 }
79
80 fn buffer(_array: ArrayView<'_, Self>, idx: usize) -> BufferHandle {
81 vortex_panic!("ListArray buffer index {idx} out of bounds")
82 }
83
84 fn buffer_name(_array: ArrayView<'_, Self>, idx: usize) -> Option<String> {
85 vortex_panic!("ListArray buffer_name index {idx} out of bounds")
86 }
87
88 fn with_buffers(
89 &self,
90 array: ArrayView<'_, Self>,
91 buffers: &[BufferHandle],
92 ) -> VortexResult<ArrayParts<Self>> {
93 with_empty_buffers(self, array, buffers)
94 }
95
96 fn reduce_parent(
97 array: ArrayView<'_, Self>,
98 parent: &ArrayRef,
99 child_idx: usize,
100 ) -> VortexResult<Option<ArrayRef>> {
101 PARENT_RULES.evaluate(array, parent, child_idx)
102 }
103
104 fn serialize(
105 array: ArrayView<'_, Self>,
106 _session: &VortexSession,
107 ) -> VortexResult<Option<Vec<u8>>> {
108 Ok(Some(
109 ListMetadata {
110 elements_len: array.elements().len() as u64,
111 offset_ptype: PType::try_from(array.offsets().dtype())? as i32,
112 }
113 .encode_to_vec(),
114 ))
115 }
116
117 fn validate(
118 &self,
119 _data: &ListData,
120 dtype: &DType,
121 len: usize,
122 slots: &[Option<ArrayRef>],
123 ) -> VortexResult<()> {
124 vortex_ensure_eq!(slots.len(), ListSlots::COUNT);
125 let elements = slots[ListSlots::ELEMENTS]
126 .as_ref()
127 .vortex_expect("ListArray elements slot");
128 let offsets = slots[ListSlots::OFFSETS]
129 .as_ref()
130 .vortex_expect("ListArray offsets slot");
131 vortex_ensure_eq!(
132 offsets.len().saturating_sub(1),
133 len,
134 "ListArray length does not match outer length",
135 );
136
137 let actual_dtype = DType::List(Arc::new(elements.dtype().clone()), dtype.nullability());
138 vortex_ensure_eq!(
139 &actual_dtype,
140 dtype,
141 "ListArray dtype does not match outer dtype",
142 );
143
144 Ok(())
145 }
146
147 fn deserialize(
148 &self,
149 dtype: &DType,
150 len: usize,
151 metadata: &[u8],
152
153 _buffers: &[BufferHandle],
154 children: &dyn ArrayChildren,
155 session: &VortexSession,
156 ) -> VortexResult<ArrayParts<Self>> {
157 let metadata = ListMetadata::decode(metadata)?;
158 let validity = if children.len() == 2 {
159 Validity::from(dtype.nullability())
160 } else if children.len() == 3 {
161 let validity = children.get(2, &Validity::DTYPE, len)?;
162 Validity::Array(validity)
163 } else {
164 vortex_bail!("Expected 2 or 3 children, got {}", children.len());
165 };
166
167 let DType::List(element_dtype, _) = &dtype else {
168 vortex_bail!("Expected List dtype, got {:?}", dtype);
169 };
170 let elements = children.get(
171 0,
172 element_dtype.as_ref(),
173 usize::try_from(metadata.elements_len)?,
174 )?;
175
176 let offsets = children.get(
177 1,
178 &DType::Primitive(metadata.offset_ptype(), Nullability::NonNullable),
179 len + 1,
180 )?;
181
182 let data = ListData::try_build(
183 elements.clone(),
184 offsets.clone(),
185 validity.clone(),
186 &mut session.create_execution_ctx(),
187 )?;
188 let slots = ListData::make_slots(&elements, &offsets, &validity, len);
189 Ok(ArrayParts::new(
190 self.clone(),
191 dtype.clone(),
192 len,
193 data,
194 slots,
195 ))
196 }
197
198 fn slot_name(_array: ArrayView<'_, Self>, idx: usize) -> String {
199 ListSlots::NAMES[idx].to_string()
200 }
201
202 fn execute(array: Array<Self>, ctx: &mut ExecutionCtx) -> VortexResult<ExecutionResult> {
203 Ok(ExecutionResult::done(
204 list_view_from_list(array, ctx)?.into_array(),
205 ))
206 }
207
208 #[expect(clippy::cognitive_complexity)]
210 fn append_to_builder(
211 array: ArrayView<'_, Self>,
212 builder: &mut dyn ArrayBuilder,
213 ctx: &mut ExecutionCtx,
214 ) -> VortexResult<()> {
215 match match_each_list_builder!(&mut *builder, |b| b.append_list_array(array, ctx)) {
216 Some(result) => result,
217 None => vortex_bail!(
218 "cannot append a List array of dtype {} to a {} builder",
219 array.dtype(),
220 builder.dtype()
221 ),
222 }
223 }
224}
225
226#[derive(Clone, Debug)]
227pub struct List;