1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
|
// ----------------------------------------------------------------------------
// - Open3D: www.open3d.org -
// ----------------------------------------------------------------------------
// Copyright (c) 2018-2024 www.open3d.org
// SPDX-License-Identifier: MIT
// ----------------------------------------------------------------------------
#include "open3d/t/geometry/TensorMap.h"
#include "open3d/t/geometry/PointCloud.h"
#include "pybind/docstring.h"
#include "pybind/t/geometry/geometry.h"
namespace open3d {
namespace t {
namespace geometry {
// This is a copy of py::bind_map function, with `__delitem__` function
// removed. The same is defined in `pybind_tensormap` to use
// `TensorMap::Erase(key)`.
template <typename Map,
typename holder_type = std::unique_ptr<Map>,
typename... Args>
static py::class_<Map, holder_type> bind_tensor_map(py::handle scope,
const std::string &name,
Args &&... args) {
using KeyType = typename Map::key_type;
using MappedType = typename Map::mapped_type;
using Class_ = py::class_<Map, holder_type>;
// If either type is a non-module-local bound type then make the map binding
// non-local as well; otherwise (e.g. both types are either module-local or
// converting) the map will be module-local.
auto tinfo = py::detail::get_type_info(typeid(MappedType));
bool local = !tinfo || tinfo->module_local;
if (local) {
tinfo = py::detail::get_type_info(typeid(KeyType));
local = !tinfo || tinfo->module_local;
}
Class_ cl(scope, name.c_str(), pybind11::module_local(local),
std::forward<Args>(args)...);
cl.def(py::init<>());
// Register stream insertion operator (if possible)
py::detail::map_if_insertion_operator<Map, Class_>(cl, name);
cl.def(
"__bool__", [](const Map &m) -> bool { return !m.empty(); },
"Check whether the map is nonempty");
// Essential: keep list alive while iterator exists
cl.def(
"__iter__",
[](Map &m) { return py::make_key_iterator(m.begin(), m.end()); },
py::keep_alive<0, 1>()
);
// Essential: keep list alive while iterator exists
cl.def(
"items",
[](Map &m) { return py::make_iterator(m.begin(), m.end()); },
py::keep_alive<0, 1>());
cl.def(
"__getitem__",
[](Map &m, const KeyType &k) -> MappedType & {
auto it = m.find(k);
if (it == m.end())
throw py::key_error(
fmt::format("Key {} not found in TensorMap", k));
return it->second;
},
// py::return_value_policy::copy is used as the safest option.
// The goal is to make TensorMap works similarly as putting Tensors
// into a python dict, i.e., {"a": Tensor(xx), "b": Tensor(XX)}.
// Accessing a value in the map will return a shallow copy of the
// tensor that shares the same underlying memory.
//
// - automatic : works, different id
// - automatic_reference: works, different id
// - take_ownership : doesn't work, segfault
// - copy : works, different id
// - move : doesn't work, blob is null
// - reference : doesn't work, when a key is deleted, the
// alias becomes invalid
// - reference_internal : doesn't work, value in map overwritten
// when assigning to alias
py::return_value_policy::copy);
cl.def("__setitem__", [](Map &m, const KeyType &k, const MappedType &v) {
if (!TensorMap::GetReservedKeys().count(k)) {
m[k] = v;
} else {
throw py::key_error(
fmt::format("Cannot assign to reserved key \"{}\"", k));
}
});
cl.def("__contains__", [](Map &m, const KeyType &k) -> bool {
auto it = m.find(k);
if (it == m.end()) return false;
return true;
});
// Assignment provided only if the type is copyable
py::detail::map_assignment<Map, Class_>(cl);
// Deleted the "__delitem__" function.
// This will be implemented in `pybind_tensormap()`.
cl.def("__len__", [](const Map &m) -> size_t { return m.size(); });
return cl;
}
void pybind_tensormap_declarations(py::module &m) {
auto tm = bind_tensor_map<TensorMap>(
m, "TensorMap", "Map of String to Tensor with a primary key.");
}
void pybind_tensormap_definitions(py::module &m) {
// Bind to the generic dictionary interface such that it works the same as a
// regular dictionary in Python, except that types are enforced. Supported
// functions include `__bool__`, `__iter__`, `items`, `__getitem__`,
// `__contains__`, `__len__` and map assignment.
// The `__delitem__` function is removed from bind_map, in bind_tensor_map,
// and defined in this function, to use TensorMap::Erase, in order to
// protect users from deleting the `private_key`.
auto tm = static_cast<py::class_<TensorMap, std::unique_ptr<TensorMap>>>(
m.attr("TensorMap"));
tm.def("__delitem__",
[](TensorMap &m, const std::string &k) { return m.Erase(k); });
tm.def("erase",
[](TensorMap &m, const std::string &k) { return m.Erase(k); });
// Constructors.
tm.def(py::init<const std::string &>(), "primary_key"_a);
tm.def(py::init<const std::string &,
const std::unordered_map<std::string, core::Tensor> &>(),
"primary_key"_a, "map_keys_to_tensors"_a);
// Member functions. Some C++ functions are ignored since the
// functionalities are already covered in the generic dictionary interface.
tm.def_property_readonly("primary_key", &TensorMap::GetPrimaryKey);
tm.def("is_size_synchronized", &TensorMap::IsSizeSynchronized);
tm.def("assert_size_synchronized", &TensorMap::AssertSizeSynchronized);
// Pickle support.
tm.def(py::pickle(
[](const TensorMap &m) {
// __getstate__
std::unordered_map<std::string, core::Tensor> map;
for (const auto &kv : m) {
map[kv.first] = kv.second;
}
return py::make_tuple(m.GetPrimaryKey(), map);
},
[](py::tuple t) {
// __setstate__
if (t.size() != 2) {
utility::LogError(
"Cannot unpickle TensorMap! Expecting a tuple of "
"size 2.");
}
return TensorMap(t[0].cast<std::string>(),
t[1].cast<std::unordered_map<std::string,
core::Tensor>>());
}));
tm.def("__setattr__",
[](TensorMap &m, const std::string &key, const core::Tensor &val) {
if (!TensorMap::GetReservedKeys().count(key)) {
m[key] = val;
} else {
throw py::key_error(fmt::format(
"Cannot assign to reserved key \"{}\"", key));
}
});
tm.def("__getattr__",
[](TensorMap &m, const std::string &key) -> core::Tensor {
auto it = m.find(key);
if (it == m.end()) {
throw py::key_error(
fmt::format("Key {} not found in TensorMap", key));
}
return it->second;
});
tm.def("__delattr__", [](TensorMap &m, const std::string &key) {
auto it = m.find(key);
if (it == m.end()) {
throw py::key_error(
fmt::format("Key {} not found in TensorMap", key));
}
return m.Erase(key);
});
tm.def("__str__", &TensorMap::ToString);
tm.def("__repr__", &TensorMap::ToString);
tm.def("__dir__", [](TensorMap &m) {
auto keys = py::list();
for (const auto &kv : m) {
keys.append(kv.first);
}
return keys;
});
}
} // namespace geometry
} // namespace t
} // namespace open3d
|
