1 | //===-- PythonDataObjects.cpp ---------------------------------------------===// |
2 | // |
3 | // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. |
4 | // See https://llvm.org/LICENSE.txt for license information. |
5 | // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception |
6 | // |
7 | //===----------------------------------------------------------------------===// |
8 | |
9 | #include "lldb/Host/Config.h" |
10 | |
11 | #if LLDB_ENABLE_PYTHON |
12 | |
13 | #include "PythonDataObjects.h" |
14 | #include "ScriptInterpreterPython.h" |
15 | |
16 | #include "lldb/Host/File.h" |
17 | #include "lldb/Host/FileSystem.h" |
18 | #include "lldb/Interpreter/ScriptInterpreter.h" |
19 | #include "lldb/Utility/LLDBLog.h" |
20 | #include "lldb/Utility/Log.h" |
21 | #include "lldb/Utility/Stream.h" |
22 | |
23 | #include "llvm/Support/Casting.h" |
24 | #include "llvm/Support/ConvertUTF.h" |
25 | #include "llvm/Support/Errno.h" |
26 | |
27 | #include <cstdio> |
28 | #include <variant> |
29 | |
30 | using namespace lldb_private; |
31 | using namespace lldb; |
32 | using namespace lldb_private::python; |
33 | using llvm::cantFail; |
34 | using llvm::Error; |
35 | using llvm::Expected; |
36 | using llvm::Twine; |
37 | |
38 | template <> Expected<bool> python::As<bool>(Expected<PythonObject> &&obj) { |
39 | if (!obj) |
40 | return obj.takeError(); |
41 | return obj.get().IsTrue(); |
42 | } |
43 | |
44 | template <> |
45 | Expected<long long> python::As<long long>(Expected<PythonObject> &&obj) { |
46 | if (!obj) |
47 | return obj.takeError(); |
48 | return obj->AsLongLong(); |
49 | } |
50 | |
51 | template <> |
52 | Expected<unsigned long long> |
53 | python::As<unsigned long long>(Expected<PythonObject> &&obj) { |
54 | if (!obj) |
55 | return obj.takeError(); |
56 | return obj->AsUnsignedLongLong(); |
57 | } |
58 | |
59 | template <> |
60 | Expected<std::string> python::As<std::string>(Expected<PythonObject> &&obj) { |
61 | if (!obj) |
62 | return obj.takeError(); |
63 | PyObject *str_obj = PyObject_Str(obj.get().get()); |
64 | if (!str_obj) |
65 | return llvm::make_error<PythonException>(); |
66 | auto str = Take<PythonString>(obj: str_obj); |
67 | auto utf8 = str.AsUTF8(); |
68 | if (!utf8) |
69 | return utf8.takeError(); |
70 | return std::string(utf8.get()); |
71 | } |
72 | |
73 | static bool python_is_finalizing() { |
74 | #if (PY_MAJOR_VERSION == 3 && PY_MINOR_VERSION >= 13) || (PY_MAJOR_VERSION > 3) |
75 | return Py_IsFinalizing(); |
76 | #elif PY_MAJOR_VERSION == 3 && PY_MINOR_VERSION < 7 |
77 | return _Py_Finalizing != nullptr; |
78 | #else |
79 | return _Py_IsFinalizing(); |
80 | #endif |
81 | } |
82 | |
83 | void PythonObject::Reset() { |
84 | if (m_py_obj && Py_IsInitialized()) { |
85 | if (python_is_finalizing()) { |
86 | // Leak m_py_obj rather than crashing the process. |
87 | // https://docs.python.org/3/c-api/init.html#c.PyGILState_Ensure |
88 | } else { |
89 | PyGILState_STATE state = PyGILState_Ensure(); |
90 | Py_DECREF(m_py_obj); |
91 | PyGILState_Release(state); |
92 | } |
93 | } |
94 | m_py_obj = nullptr; |
95 | } |
96 | |
97 | Expected<long long> PythonObject::AsLongLong() const { |
98 | if (!m_py_obj) |
99 | return nullDeref(); |
100 | assert(!PyErr_Occurred()); |
101 | long long r = PyLong_AsLongLong(m_py_obj); |
102 | if (PyErr_Occurred()) |
103 | return exception(); |
104 | return r; |
105 | } |
106 | |
107 | Expected<unsigned long long> PythonObject::AsUnsignedLongLong() const { |
108 | if (!m_py_obj) |
109 | return nullDeref(); |
110 | assert(!PyErr_Occurred()); |
111 | long long r = PyLong_AsUnsignedLongLong(m_py_obj); |
112 | if (PyErr_Occurred()) |
113 | return exception(); |
114 | return r; |
115 | } |
116 | |
117 | // wraps on overflow, instead of raising an error. |
118 | Expected<unsigned long long> PythonObject::AsModuloUnsignedLongLong() const { |
119 | if (!m_py_obj) |
120 | return nullDeref(); |
121 | assert(!PyErr_Occurred()); |
122 | unsigned long long r = PyLong_AsUnsignedLongLongMask(m_py_obj); |
123 | // FIXME: We should fetch the exception message and hoist it. |
124 | if (PyErr_Occurred()) |
125 | return exception(); |
126 | return r; |
127 | } |
128 | |
129 | void StructuredPythonObject::Serialize(llvm::json::OStream &s) const { |
130 | s.value(V: llvm::formatv(Fmt: "Python Obj: {0:X}" , Vals: GetValue()).str()); |
131 | } |
132 | |
133 | // PythonObject |
134 | |
135 | void PythonObject::Dump(Stream &strm) const { |
136 | if (m_py_obj) { |
137 | FILE *file = llvm::sys::RetryAfterSignal(Fail: nullptr, F&: ::tmpfile); |
138 | if (file) { |
139 | ::PyObject_Print(m_py_obj, file, 0); |
140 | const long length = ftell(stream: file); |
141 | if (length) { |
142 | ::rewind(stream: file); |
143 | std::vector<char> file_contents(length, '\0'); |
144 | const size_t length_read = |
145 | ::fread(ptr: file_contents.data(), size: 1, n: file_contents.size(), stream: file); |
146 | if (length_read > 0) |
147 | strm.Write(src: file_contents.data(), src_len: length_read); |
148 | } |
149 | ::fclose(stream: file); |
150 | } |
151 | } else |
152 | strm.PutCString(cstr: "NULL" ); |
153 | } |
154 | |
155 | PyObjectType PythonObject::GetObjectType() const { |
156 | if (!IsAllocated()) |
157 | return PyObjectType::None; |
158 | |
159 | if (PythonModule::Check(py_obj: m_py_obj)) |
160 | return PyObjectType::Module; |
161 | if (PythonList::Check(py_obj: m_py_obj)) |
162 | return PyObjectType::List; |
163 | if (PythonTuple::Check(py_obj: m_py_obj)) |
164 | return PyObjectType::Tuple; |
165 | if (PythonDictionary::Check(py_obj: m_py_obj)) |
166 | return PyObjectType::Dictionary; |
167 | if (PythonString::Check(py_obj: m_py_obj)) |
168 | return PyObjectType::String; |
169 | if (PythonBytes::Check(py_obj: m_py_obj)) |
170 | return PyObjectType::Bytes; |
171 | if (PythonByteArray::Check(py_obj: m_py_obj)) |
172 | return PyObjectType::ByteArray; |
173 | if (PythonBoolean::Check(py_obj: m_py_obj)) |
174 | return PyObjectType::Boolean; |
175 | if (PythonInteger::Check(py_obj: m_py_obj)) |
176 | return PyObjectType::Integer; |
177 | if (PythonFile::Check(py_obj: m_py_obj)) |
178 | return PyObjectType::File; |
179 | if (PythonCallable::Check(py_obj: m_py_obj)) |
180 | return PyObjectType::Callable; |
181 | return PyObjectType::Unknown; |
182 | } |
183 | |
184 | PythonString PythonObject::Repr() const { |
185 | if (!m_py_obj) |
186 | return PythonString(); |
187 | PyObject *repr = PyObject_Repr(m_py_obj); |
188 | if (!repr) |
189 | return PythonString(); |
190 | return PythonString(PyRefType::Owned, repr); |
191 | } |
192 | |
193 | PythonString PythonObject::Str() const { |
194 | if (!m_py_obj) |
195 | return PythonString(); |
196 | PyObject *str = PyObject_Str(m_py_obj); |
197 | if (!str) |
198 | return PythonString(); |
199 | return PythonString(PyRefType::Owned, str); |
200 | } |
201 | |
202 | PythonObject |
203 | PythonObject::ResolveNameWithDictionary(llvm::StringRef name, |
204 | const PythonDictionary &dict) { |
205 | size_t dot_pos = name.find(C: '.'); |
206 | llvm::StringRef piece = name.substr(Start: 0, N: dot_pos); |
207 | PythonObject result = dict.GetItemForKey(key: PythonString(piece)); |
208 | if (dot_pos == llvm::StringRef::npos) { |
209 | // There was no dot, we're done. |
210 | return result; |
211 | } |
212 | |
213 | // There was a dot. The remaining portion of the name should be looked up in |
214 | // the context of the object that was found in the dictionary. |
215 | return result.ResolveName(name: name.substr(Start: dot_pos + 1)); |
216 | } |
217 | |
218 | PythonObject PythonObject::ResolveName(llvm::StringRef name) const { |
219 | // Resolve the name in the context of the specified object. If, for example, |
220 | // `this` refers to a PyModule, then this will look for `name` in this |
221 | // module. If `this` refers to a PyType, then it will resolve `name` as an |
222 | // attribute of that type. If `this` refers to an instance of an object, |
223 | // then it will resolve `name` as the value of the specified field. |
224 | // |
225 | // This function handles dotted names so that, for example, if `m_py_obj` |
226 | // refers to the `sys` module, and `name` == "path.append", then it will find |
227 | // the function `sys.path.append`. |
228 | |
229 | size_t dot_pos = name.find(C: '.'); |
230 | if (dot_pos == llvm::StringRef::npos) { |
231 | // No dots in the name, we should be able to find the value immediately as |
232 | // an attribute of `m_py_obj`. |
233 | return GetAttributeValue(attribute: name); |
234 | } |
235 | |
236 | // Look up the first piece of the name, and resolve the rest as a child of |
237 | // that. |
238 | PythonObject parent = ResolveName(name: name.substr(Start: 0, N: dot_pos)); |
239 | if (!parent.IsAllocated()) |
240 | return PythonObject(); |
241 | |
242 | // Tail recursion.. should be optimized by the compiler |
243 | return parent.ResolveName(name: name.substr(Start: dot_pos + 1)); |
244 | } |
245 | |
246 | bool PythonObject::HasAttribute(llvm::StringRef attr) const { |
247 | if (!IsValid()) |
248 | return false; |
249 | PythonString py_attr(attr); |
250 | return !!PyObject_HasAttr(m_py_obj, py_attr.get()); |
251 | } |
252 | |
253 | PythonObject PythonObject::GetAttributeValue(llvm::StringRef attr) const { |
254 | if (!IsValid()) |
255 | return PythonObject(); |
256 | |
257 | PythonString py_attr(attr); |
258 | if (!PyObject_HasAttr(m_py_obj, py_attr.get())) |
259 | return PythonObject(); |
260 | |
261 | return PythonObject(PyRefType::Owned, |
262 | PyObject_GetAttr(m_py_obj, py_attr.get())); |
263 | } |
264 | |
265 | StructuredData::ObjectSP PythonObject::CreateStructuredObject() const { |
266 | assert(PyGILState_Check()); |
267 | switch (GetObjectType()) { |
268 | case PyObjectType::Dictionary: |
269 | return PythonDictionary(PyRefType::Borrowed, m_py_obj) |
270 | .CreateStructuredDictionary(); |
271 | case PyObjectType::Boolean: |
272 | return PythonBoolean(PyRefType::Borrowed, m_py_obj) |
273 | .CreateStructuredBoolean(); |
274 | case PyObjectType::Integer: { |
275 | StructuredData::IntegerSP int_sp = |
276 | PythonInteger(PyRefType::Borrowed, m_py_obj).CreateStructuredInteger(); |
277 | if (std::holds_alternative<StructuredData::UnsignedIntegerSP>(v: int_sp)) |
278 | return std::get<StructuredData::UnsignedIntegerSP>(v&: int_sp); |
279 | if (std::holds_alternative<StructuredData::SignedIntegerSP>(v: int_sp)) |
280 | return std::get<StructuredData::SignedIntegerSP>(v&: int_sp); |
281 | return nullptr; |
282 | }; |
283 | case PyObjectType::List: |
284 | return PythonList(PyRefType::Borrowed, m_py_obj).CreateStructuredArray(); |
285 | case PyObjectType::String: |
286 | return PythonString(PyRefType::Borrowed, m_py_obj).CreateStructuredString(); |
287 | case PyObjectType::Bytes: |
288 | return PythonBytes(PyRefType::Borrowed, m_py_obj).CreateStructuredString(); |
289 | case PyObjectType::ByteArray: |
290 | return PythonByteArray(PyRefType::Borrowed, m_py_obj) |
291 | .CreateStructuredString(); |
292 | case PyObjectType::None: |
293 | return StructuredData::ObjectSP(); |
294 | default: |
295 | return StructuredData::ObjectSP(new StructuredPythonObject( |
296 | PythonObject(PyRefType::Borrowed, m_py_obj))); |
297 | } |
298 | } |
299 | |
300 | // PythonString |
301 | |
302 | PythonBytes::PythonBytes(llvm::ArrayRef<uint8_t> bytes) { SetBytes(bytes); } |
303 | |
304 | PythonBytes::PythonBytes(const uint8_t *bytes, size_t length) { |
305 | SetBytes(llvm::ArrayRef<uint8_t>(bytes, length)); |
306 | } |
307 | |
308 | bool PythonBytes::Check(PyObject *py_obj) { |
309 | if (!py_obj) |
310 | return false; |
311 | return PyBytes_Check(py_obj); |
312 | } |
313 | |
314 | llvm::ArrayRef<uint8_t> PythonBytes::GetBytes() const { |
315 | if (!IsValid()) |
316 | return llvm::ArrayRef<uint8_t>(); |
317 | |
318 | Py_ssize_t size; |
319 | char *c; |
320 | |
321 | PyBytes_AsStringAndSize(obj: m_py_obj, s: &c, len: &size); |
322 | return llvm::ArrayRef<uint8_t>(reinterpret_cast<uint8_t *>(c), size); |
323 | } |
324 | |
325 | size_t PythonBytes::GetSize() const { |
326 | if (!IsValid()) |
327 | return 0; |
328 | return PyBytes_Size(m_py_obj); |
329 | } |
330 | |
331 | void PythonBytes::SetBytes(llvm::ArrayRef<uint8_t> bytes) { |
332 | const char *data = reinterpret_cast<const char *>(bytes.data()); |
333 | *this = Take<PythonBytes>(obj: PyBytes_FromStringAndSize(data, bytes.size())); |
334 | } |
335 | |
336 | StructuredData::StringSP PythonBytes::CreateStructuredString() const { |
337 | StructuredData::StringSP result(new StructuredData::String); |
338 | Py_ssize_t size; |
339 | char *c; |
340 | PyBytes_AsStringAndSize(obj: m_py_obj, s: &c, len: &size); |
341 | result->SetValue(std::string(c, size)); |
342 | return result; |
343 | } |
344 | |
345 | PythonByteArray::PythonByteArray(llvm::ArrayRef<uint8_t> bytes) |
346 | : PythonByteArray(bytes.data(), bytes.size()) {} |
347 | |
348 | PythonByteArray::PythonByteArray(const uint8_t *bytes, size_t length) { |
349 | const char *str = reinterpret_cast<const char *>(bytes); |
350 | *this = Take<PythonByteArray>(obj: PyByteArray_FromStringAndSize(str, length)); |
351 | } |
352 | |
353 | bool PythonByteArray::Check(PyObject *py_obj) { |
354 | if (!py_obj) |
355 | return false; |
356 | return PyByteArray_Check(py_obj); |
357 | } |
358 | |
359 | llvm::ArrayRef<uint8_t> PythonByteArray::GetBytes() const { |
360 | if (!IsValid()) |
361 | return llvm::ArrayRef<uint8_t>(); |
362 | |
363 | char *c = PyByteArray_AsString(m_py_obj); |
364 | size_t size = GetSize(); |
365 | return llvm::ArrayRef<uint8_t>(reinterpret_cast<uint8_t *>(c), size); |
366 | } |
367 | |
368 | size_t PythonByteArray::GetSize() const { |
369 | if (!IsValid()) |
370 | return 0; |
371 | |
372 | return PyByteArray_Size(m_py_obj); |
373 | } |
374 | |
375 | StructuredData::StringSP PythonByteArray::CreateStructuredString() const { |
376 | StructuredData::StringSP result(new StructuredData::String); |
377 | llvm::ArrayRef<uint8_t> bytes = GetBytes(); |
378 | const char *str = reinterpret_cast<const char *>(bytes.data()); |
379 | result->SetValue(std::string(str, bytes.size())); |
380 | return result; |
381 | } |
382 | |
383 | // PythonString |
384 | |
385 | Expected<PythonString> PythonString::FromUTF8(llvm::StringRef string) { |
386 | PyObject *str = PyUnicode_FromStringAndSize(u: string.data(), size: string.size()); |
387 | if (!str) |
388 | return llvm::make_error<PythonException>(); |
389 | return Take<PythonString>(obj: str); |
390 | } |
391 | |
392 | PythonString::PythonString(llvm::StringRef string) { SetString(string); } |
393 | |
394 | bool PythonString::Check(PyObject *py_obj) { |
395 | if (!py_obj) |
396 | return false; |
397 | |
398 | if (PyUnicode_Check(py_obj)) |
399 | return true; |
400 | return false; |
401 | } |
402 | |
403 | llvm::StringRef PythonString::GetString() const { |
404 | auto s = AsUTF8(); |
405 | if (!s) { |
406 | llvm::consumeError(Err: s.takeError()); |
407 | return llvm::StringRef("" ); |
408 | } |
409 | return s.get(); |
410 | } |
411 | |
412 | Expected<llvm::StringRef> PythonString::AsUTF8() const { |
413 | if (!IsValid()) |
414 | return nullDeref(); |
415 | |
416 | Py_ssize_t size; |
417 | const char *data; |
418 | |
419 | data = PyUnicode_AsUTF8AndSize(unicode: m_py_obj, size: &size); |
420 | |
421 | if (!data) |
422 | return exception(); |
423 | |
424 | return llvm::StringRef(data, size); |
425 | } |
426 | |
427 | size_t PythonString::GetSize() const { |
428 | if (IsValid()) { |
429 | #if PY_MINOR_VERSION >= 3 |
430 | return PyUnicode_GetLength(unicode: m_py_obj); |
431 | #else |
432 | return PyUnicode_GetSize(m_py_obj); |
433 | #endif |
434 | } |
435 | return 0; |
436 | } |
437 | |
438 | void PythonString::SetString(llvm::StringRef string) { |
439 | auto s = FromUTF8(string); |
440 | if (!s) { |
441 | llvm::consumeError(Err: s.takeError()); |
442 | Reset(); |
443 | } else { |
444 | *this = std::move(s.get()); |
445 | } |
446 | } |
447 | |
448 | StructuredData::StringSP PythonString::CreateStructuredString() const { |
449 | StructuredData::StringSP result(new StructuredData::String); |
450 | result->SetValue(GetString()); |
451 | return result; |
452 | } |
453 | |
454 | // PythonInteger |
455 | |
456 | PythonInteger::PythonInteger(int64_t value) { SetInteger(value); } |
457 | |
458 | bool PythonInteger::Check(PyObject *py_obj) { |
459 | if (!py_obj) |
460 | return false; |
461 | |
462 | // Python 3 does not have PyInt_Check. There is only one type of integral |
463 | // value, long. |
464 | return PyLong_Check(py_obj); |
465 | } |
466 | |
467 | void PythonInteger::SetInteger(int64_t value) { |
468 | *this = Take<PythonInteger>(obj: PyLong_FromLongLong(value)); |
469 | } |
470 | |
471 | StructuredData::IntegerSP PythonInteger::CreateStructuredInteger() const { |
472 | StructuredData::UnsignedIntegerSP uint_sp = CreateStructuredUnsignedInteger(); |
473 | return uint_sp ? StructuredData::IntegerSP(uint_sp) |
474 | : CreateStructuredSignedInteger(); |
475 | } |
476 | |
477 | StructuredData::UnsignedIntegerSP |
478 | PythonInteger::CreateStructuredUnsignedInteger() const { |
479 | StructuredData::UnsignedIntegerSP result = nullptr; |
480 | llvm::Expected<unsigned long long> value = AsUnsignedLongLong(); |
481 | if (!value) |
482 | llvm::consumeError(Err: value.takeError()); |
483 | else |
484 | result = std::make_shared<StructuredData::UnsignedInteger>(args&: value.get()); |
485 | |
486 | return result; |
487 | } |
488 | |
489 | StructuredData::SignedIntegerSP |
490 | PythonInteger::CreateStructuredSignedInteger() const { |
491 | StructuredData::SignedIntegerSP result = nullptr; |
492 | llvm::Expected<long long> value = AsLongLong(); |
493 | if (!value) |
494 | llvm::consumeError(Err: value.takeError()); |
495 | else |
496 | result = std::make_shared<StructuredData::SignedInteger>(args&: value.get()); |
497 | |
498 | return result; |
499 | } |
500 | |
501 | // PythonBoolean |
502 | |
503 | PythonBoolean::PythonBoolean(bool value) { |
504 | SetValue(value); |
505 | } |
506 | |
507 | bool PythonBoolean::Check(PyObject *py_obj) { |
508 | return py_obj ? PyBool_Check(py_obj) : false; |
509 | } |
510 | |
511 | bool PythonBoolean::GetValue() const { |
512 | return m_py_obj ? PyObject_IsTrue(m_py_obj) : false; |
513 | } |
514 | |
515 | void PythonBoolean::SetValue(bool value) { |
516 | *this = Take<PythonBoolean>(obj: PyBool_FromLong(value)); |
517 | } |
518 | |
519 | StructuredData::BooleanSP PythonBoolean::CreateStructuredBoolean() const { |
520 | StructuredData::BooleanSP result(new StructuredData::Boolean); |
521 | result->SetValue(GetValue()); |
522 | return result; |
523 | } |
524 | |
525 | // PythonList |
526 | |
527 | PythonList::PythonList(PyInitialValue value) { |
528 | if (value == PyInitialValue::Empty) |
529 | *this = Take<PythonList>(obj: PyList_New(size: 0)); |
530 | } |
531 | |
532 | PythonList::PythonList(int list_size) { |
533 | *this = Take<PythonList>(obj: PyList_New(size: list_size)); |
534 | } |
535 | |
536 | bool PythonList::Check(PyObject *py_obj) { |
537 | if (!py_obj) |
538 | return false; |
539 | return PyList_Check(py_obj); |
540 | } |
541 | |
542 | uint32_t PythonList::GetSize() const { |
543 | if (IsValid()) |
544 | return PyList_GET_SIZE(m_py_obj); |
545 | return 0; |
546 | } |
547 | |
548 | PythonObject PythonList::GetItemAtIndex(uint32_t index) const { |
549 | if (IsValid()) |
550 | return PythonObject(PyRefType::Borrowed, PyList_GetItem(m_py_obj, index)); |
551 | return PythonObject(); |
552 | } |
553 | |
554 | void PythonList::SetItemAtIndex(uint32_t index, const PythonObject &object) { |
555 | if (IsAllocated() && object.IsValid()) { |
556 | // PyList_SetItem is documented to "steal" a reference, so we need to |
557 | // convert it to an owned reference by incrementing it. |
558 | Py_INCREF(object.get()); |
559 | PyList_SetItem(m_py_obj, index, object.get()); |
560 | } |
561 | } |
562 | |
563 | void PythonList::AppendItem(const PythonObject &object) { |
564 | if (IsAllocated() && object.IsValid()) { |
565 | // `PyList_Append` does *not* steal a reference, so do not call `Py_INCREF` |
566 | // here like we do with `PyList_SetItem`. |
567 | PyList_Append(m_py_obj, object.get()); |
568 | } |
569 | } |
570 | |
571 | StructuredData::ArraySP PythonList::CreateStructuredArray() const { |
572 | StructuredData::ArraySP result(new StructuredData::Array); |
573 | uint32_t count = GetSize(); |
574 | for (uint32_t i = 0; i < count; ++i) { |
575 | PythonObject obj = GetItemAtIndex(index: i); |
576 | result->AddItem(item: obj.CreateStructuredObject()); |
577 | } |
578 | return result; |
579 | } |
580 | |
581 | // PythonTuple |
582 | |
583 | PythonTuple::PythonTuple(PyInitialValue value) { |
584 | if (value == PyInitialValue::Empty) |
585 | *this = Take<PythonTuple>(obj: PyTuple_New(size: 0)); |
586 | } |
587 | |
588 | PythonTuple::PythonTuple(int tuple_size) { |
589 | *this = Take<PythonTuple>(obj: PyTuple_New(size: tuple_size)); |
590 | } |
591 | |
592 | PythonTuple::PythonTuple(std::initializer_list<PythonObject> objects) { |
593 | m_py_obj = PyTuple_New(size: objects.size()); |
594 | |
595 | uint32_t idx = 0; |
596 | for (auto object : objects) { |
597 | if (object.IsValid()) |
598 | SetItemAtIndex(index: idx, object); |
599 | idx++; |
600 | } |
601 | } |
602 | |
603 | PythonTuple::PythonTuple(std::initializer_list<PyObject *> objects) { |
604 | m_py_obj = PyTuple_New(size: objects.size()); |
605 | |
606 | uint32_t idx = 0; |
607 | for (auto py_object : objects) { |
608 | PythonObject object(PyRefType::Borrowed, py_object); |
609 | if (object.IsValid()) |
610 | SetItemAtIndex(index: idx, object); |
611 | idx++; |
612 | } |
613 | } |
614 | |
615 | bool PythonTuple::Check(PyObject *py_obj) { |
616 | if (!py_obj) |
617 | return false; |
618 | return PyTuple_Check(py_obj); |
619 | } |
620 | |
621 | uint32_t PythonTuple::GetSize() const { |
622 | if (IsValid()) |
623 | return PyTuple_GET_SIZE(m_py_obj); |
624 | return 0; |
625 | } |
626 | |
627 | PythonObject PythonTuple::GetItemAtIndex(uint32_t index) const { |
628 | if (IsValid()) |
629 | return PythonObject(PyRefType::Borrowed, PyTuple_GetItem(m_py_obj, index)); |
630 | return PythonObject(); |
631 | } |
632 | |
633 | void PythonTuple::SetItemAtIndex(uint32_t index, const PythonObject &object) { |
634 | if (IsAllocated() && object.IsValid()) { |
635 | // PyTuple_SetItem is documented to "steal" a reference, so we need to |
636 | // convert it to an owned reference by incrementing it. |
637 | Py_INCREF(object.get()); |
638 | PyTuple_SetItem(m_py_obj, index, object.get()); |
639 | } |
640 | } |
641 | |
642 | StructuredData::ArraySP PythonTuple::CreateStructuredArray() const { |
643 | StructuredData::ArraySP result(new StructuredData::Array); |
644 | uint32_t count = GetSize(); |
645 | for (uint32_t i = 0; i < count; ++i) { |
646 | PythonObject obj = GetItemAtIndex(index: i); |
647 | result->AddItem(item: obj.CreateStructuredObject()); |
648 | } |
649 | return result; |
650 | } |
651 | |
652 | // PythonDictionary |
653 | |
654 | PythonDictionary::PythonDictionary(PyInitialValue value) { |
655 | if (value == PyInitialValue::Empty) |
656 | *this = Take<PythonDictionary>(obj: PyDict_New()); |
657 | } |
658 | |
659 | bool PythonDictionary::Check(PyObject *py_obj) { |
660 | if (!py_obj) |
661 | return false; |
662 | |
663 | return PyDict_Check(py_obj); |
664 | } |
665 | |
666 | bool PythonDictionary::HasKey(const llvm::Twine &key) const { |
667 | if (!IsValid()) |
668 | return false; |
669 | |
670 | PythonString key_object(key.isSingleStringRef() ? key.getSingleStringRef() |
671 | : key.str()); |
672 | |
673 | if (int res = PyDict_Contains(mp: m_py_obj, key: key_object.get()) > 0) |
674 | return res; |
675 | |
676 | PyErr_Print(); |
677 | return false; |
678 | } |
679 | |
680 | uint32_t PythonDictionary::GetSize() const { |
681 | if (IsValid()) |
682 | return PyDict_Size(mp: m_py_obj); |
683 | return 0; |
684 | } |
685 | |
686 | PythonList PythonDictionary::GetKeys() const { |
687 | if (IsValid()) |
688 | return PythonList(PyRefType::Owned, PyDict_Keys(mp: m_py_obj)); |
689 | return PythonList(PyInitialValue::Invalid); |
690 | } |
691 | |
692 | PythonObject PythonDictionary::GetItemForKey(const PythonObject &key) const { |
693 | auto item = GetItem(key); |
694 | if (!item) { |
695 | llvm::consumeError(Err: item.takeError()); |
696 | return PythonObject(); |
697 | } |
698 | return std::move(item.get()); |
699 | } |
700 | |
701 | Expected<PythonObject> |
702 | PythonDictionary::GetItem(const PythonObject &key) const { |
703 | if (!IsValid()) |
704 | return nullDeref(); |
705 | PyObject *o = PyDict_GetItemWithError(mp: m_py_obj, key: key.get()); |
706 | if (PyErr_Occurred()) |
707 | return exception(); |
708 | if (!o) |
709 | return keyError(); |
710 | return Retain<PythonObject>(obj: o); |
711 | } |
712 | |
713 | Expected<PythonObject> PythonDictionary::GetItem(const Twine &key) const { |
714 | if (!IsValid()) |
715 | return nullDeref(); |
716 | PyObject *o = PyDict_GetItemString(dp: m_py_obj, key: NullTerminated(key)); |
717 | if (PyErr_Occurred()) |
718 | return exception(); |
719 | if (!o) |
720 | return keyError(); |
721 | return Retain<PythonObject>(obj: o); |
722 | } |
723 | |
724 | Error PythonDictionary::SetItem(const PythonObject &key, |
725 | const PythonObject &value) const { |
726 | if (!IsValid() || !value.IsValid()) |
727 | return nullDeref(); |
728 | int r = PyDict_SetItem(mp: m_py_obj, key: key.get(), item: value.get()); |
729 | if (r < 0) |
730 | return exception(); |
731 | return Error::success(); |
732 | } |
733 | |
734 | Error PythonDictionary::SetItem(const Twine &key, |
735 | const PythonObject &value) const { |
736 | if (!IsValid() || !value.IsValid()) |
737 | return nullDeref(); |
738 | int r = PyDict_SetItemString(dp: m_py_obj, key: NullTerminated(key), item: value.get()); |
739 | if (r < 0) |
740 | return exception(); |
741 | return Error::success(); |
742 | } |
743 | |
744 | void PythonDictionary::SetItemForKey(const PythonObject &key, |
745 | const PythonObject &value) { |
746 | Error error = SetItem(key, value); |
747 | if (error) |
748 | llvm::consumeError(Err: std::move(error)); |
749 | } |
750 | |
751 | StructuredData::DictionarySP |
752 | PythonDictionary::CreateStructuredDictionary() const { |
753 | StructuredData::DictionarySP result(new StructuredData::Dictionary); |
754 | PythonList keys(GetKeys()); |
755 | uint32_t num_keys = keys.GetSize(); |
756 | for (uint32_t i = 0; i < num_keys; ++i) { |
757 | PythonObject key = keys.GetItemAtIndex(index: i); |
758 | PythonObject value = GetItemForKey(key); |
759 | StructuredData::ObjectSP structured_value = value.CreateStructuredObject(); |
760 | result->AddItem(key: key.Str().GetString(), value_sp: structured_value); |
761 | } |
762 | return result; |
763 | } |
764 | |
765 | PythonModule PythonModule::BuiltinsModule() { return AddModule(module: "builtins" ); } |
766 | |
767 | PythonModule PythonModule::MainModule() { return AddModule(module: "__main__" ); } |
768 | |
769 | PythonModule PythonModule::AddModule(llvm::StringRef module) { |
770 | std::string str = module.str(); |
771 | return PythonModule(PyRefType::Borrowed, PyImport_AddModule(name: str.c_str())); |
772 | } |
773 | |
774 | Expected<PythonModule> PythonModule::Import(const Twine &name) { |
775 | PyObject *mod = PyImport_ImportModule(name: NullTerminated(name)); |
776 | if (!mod) |
777 | return exception(); |
778 | return Take<PythonModule>(obj: mod); |
779 | } |
780 | |
781 | Expected<PythonObject> PythonModule::Get(const Twine &name) { |
782 | if (!IsValid()) |
783 | return nullDeref(); |
784 | PyObject *dict = PyModule_GetDict(m_py_obj); |
785 | if (!dict) |
786 | return exception(); |
787 | PyObject *item = PyDict_GetItemString(dp: dict, key: NullTerminated(name)); |
788 | if (!item) |
789 | return exception(); |
790 | return Retain<PythonObject>(obj: item); |
791 | } |
792 | |
793 | bool PythonModule::Check(PyObject *py_obj) { |
794 | if (!py_obj) |
795 | return false; |
796 | |
797 | return PyModule_Check(py_obj); |
798 | } |
799 | |
800 | PythonDictionary PythonModule::GetDictionary() const { |
801 | if (!IsValid()) |
802 | return PythonDictionary(); |
803 | return Retain<PythonDictionary>(obj: PyModule_GetDict(m_py_obj)); |
804 | } |
805 | |
806 | bool PythonCallable::Check(PyObject *py_obj) { |
807 | if (!py_obj) |
808 | return false; |
809 | |
810 | return PyCallable_Check(py_obj); |
811 | } |
812 | |
813 | #if PY_MAJOR_VERSION >= 3 && PY_MINOR_VERSION >= 3 |
814 | static const char get_arg_info_script[] = R"( |
815 | from inspect import signature, Parameter, ismethod |
816 | from collections import namedtuple |
817 | ArgInfo = namedtuple('ArgInfo', ['count', 'has_varargs']) |
818 | def main(f): |
819 | count = 0 |
820 | varargs = False |
821 | for parameter in signature(f).parameters.values(): |
822 | kind = parameter.kind |
823 | if kind in (Parameter.POSITIONAL_ONLY, |
824 | Parameter.POSITIONAL_OR_KEYWORD): |
825 | count += 1 |
826 | elif kind == Parameter.VAR_POSITIONAL: |
827 | varargs = True |
828 | elif kind in (Parameter.KEYWORD_ONLY, |
829 | Parameter.VAR_KEYWORD): |
830 | pass |
831 | else: |
832 | raise Exception(f'unknown parameter kind: {kind}') |
833 | return ArgInfo(count, varargs) |
834 | )" ; |
835 | #endif |
836 | |
837 | Expected<PythonCallable::ArgInfo> PythonCallable::GetArgInfo() const { |
838 | ArgInfo result = {}; |
839 | if (!IsValid()) |
840 | return nullDeref(); |
841 | |
842 | #if PY_MAJOR_VERSION >= 3 && PY_MINOR_VERSION >= 3 |
843 | |
844 | // no need to synchronize access to this global, we already have the GIL |
845 | static PythonScript get_arg_info(get_arg_info_script); |
846 | Expected<PythonObject> pyarginfo = get_arg_info(*this); |
847 | if (!pyarginfo) |
848 | return pyarginfo.takeError(); |
849 | long long count = |
850 | cantFail(ValOrErr: As<long long>(obj: pyarginfo.get().GetAttribute(name: "count" ))); |
851 | bool has_varargs = |
852 | cantFail(ValOrErr: As<bool>(obj: pyarginfo.get().GetAttribute(name: "has_varargs" ))); |
853 | result.max_positional_args = has_varargs ? ArgInfo::UNBOUNDED : count; |
854 | |
855 | #else |
856 | PyObject *py_func_obj; |
857 | bool is_bound_method = false; |
858 | bool is_class = false; |
859 | |
860 | if (PyType_Check(m_py_obj) || PyClass_Check(m_py_obj)) { |
861 | auto init = GetAttribute("__init__" ); |
862 | if (!init) |
863 | return init.takeError(); |
864 | py_func_obj = init.get().get(); |
865 | is_class = true; |
866 | } else { |
867 | py_func_obj = m_py_obj; |
868 | } |
869 | |
870 | if (PyMethod_Check(py_func_obj)) { |
871 | py_func_obj = PyMethod_GET_FUNCTION(py_func_obj); |
872 | PythonObject im_self = GetAttributeValue("im_self" ); |
873 | if (im_self.IsValid() && !im_self.IsNone()) |
874 | is_bound_method = true; |
875 | } else { |
876 | // see if this is a callable object with an __call__ method |
877 | if (!PyFunction_Check(py_func_obj)) { |
878 | PythonObject __call__ = GetAttributeValue("__call__" ); |
879 | if (__call__.IsValid()) { |
880 | auto __callable__ = __call__.AsType<PythonCallable>(); |
881 | if (__callable__.IsValid()) { |
882 | py_func_obj = PyMethod_GET_FUNCTION(__callable__.get()); |
883 | PythonObject im_self = __callable__.GetAttributeValue("im_self" ); |
884 | if (im_self.IsValid() && !im_self.IsNone()) |
885 | is_bound_method = true; |
886 | } |
887 | } |
888 | } |
889 | } |
890 | |
891 | if (!py_func_obj) |
892 | return result; |
893 | |
894 | PyCodeObject *code = (PyCodeObject *)PyFunction_GET_CODE(py_func_obj); |
895 | if (!code) |
896 | return result; |
897 | |
898 | auto count = code->co_argcount; |
899 | bool has_varargs = !!(code->co_flags & CO_VARARGS); |
900 | result.max_positional_args = |
901 | has_varargs ? ArgInfo::UNBOUNDED |
902 | : (count - (int)is_bound_method) - (int)is_class; |
903 | |
904 | #endif |
905 | |
906 | return result; |
907 | } |
908 | |
909 | constexpr unsigned |
910 | PythonCallable::ArgInfo::UNBOUNDED; // FIXME delete after c++17 |
911 | |
912 | PythonObject PythonCallable::operator()() { |
913 | return PythonObject(PyRefType::Owned, PyObject_CallObject(callable: m_py_obj, args: nullptr)); |
914 | } |
915 | |
916 | PythonObject PythonCallable:: |
917 | operator()(std::initializer_list<PyObject *> args) { |
918 | PythonTuple arg_tuple(args); |
919 | return PythonObject(PyRefType::Owned, |
920 | PyObject_CallObject(callable: m_py_obj, args: arg_tuple.get())); |
921 | } |
922 | |
923 | PythonObject PythonCallable:: |
924 | operator()(std::initializer_list<PythonObject> args) { |
925 | PythonTuple arg_tuple(args); |
926 | return PythonObject(PyRefType::Owned, |
927 | PyObject_CallObject(callable: m_py_obj, args: arg_tuple.get())); |
928 | } |
929 | |
930 | bool PythonFile::Check(PyObject *py_obj) { |
931 | if (!py_obj) |
932 | return false; |
933 | // In Python 3, there is no `PyFile_Check`, and in fact PyFile is not even a |
934 | // first-class object type anymore. `PyFile_FromFd` is just a thin wrapper |
935 | // over `io.open()`, which returns some object derived from `io.IOBase`. As a |
936 | // result, the only way to detect a file in Python 3 is to check whether it |
937 | // inherits from `io.IOBase`. |
938 | auto io_module = PythonModule::Import(name: "io" ); |
939 | if (!io_module) { |
940 | llvm::consumeError(Err: io_module.takeError()); |
941 | return false; |
942 | } |
943 | auto iobase = io_module.get().Get(name: "IOBase" ); |
944 | if (!iobase) { |
945 | llvm::consumeError(Err: iobase.takeError()); |
946 | return false; |
947 | } |
948 | int r = PyObject_IsInstance(object: py_obj, typeorclass: iobase.get().get()); |
949 | if (r < 0) { |
950 | llvm::consumeError(Err: exception()); // clear the exception and log it. |
951 | return false; |
952 | } |
953 | return !!r; |
954 | } |
955 | |
956 | const char *PythonException::toCString() const { |
957 | if (!m_repr_bytes) |
958 | return "unknown exception" ; |
959 | return PyBytes_AS_STRING(m_repr_bytes); |
960 | } |
961 | |
962 | PythonException::PythonException(const char *caller) { |
963 | assert(PyErr_Occurred()); |
964 | m_exception_type = m_exception = m_traceback = m_repr_bytes = nullptr; |
965 | PyErr_Fetch(&m_exception_type, &m_exception, &m_traceback); |
966 | PyErr_NormalizeException(&m_exception_type, &m_exception, &m_traceback); |
967 | PyErr_Clear(); |
968 | if (m_exception) { |
969 | PyObject *repr = PyObject_Repr(m_exception); |
970 | if (repr) { |
971 | m_repr_bytes = PyUnicode_AsEncodedString(unicode: repr, encoding: "utf-8" , errors: nullptr); |
972 | if (!m_repr_bytes) { |
973 | PyErr_Clear(); |
974 | } |
975 | Py_XDECREF(repr); |
976 | } else { |
977 | PyErr_Clear(); |
978 | } |
979 | } |
980 | Log *log = GetLog(mask: LLDBLog::Script); |
981 | if (caller) |
982 | LLDB_LOGF(log, "%s failed with exception: %s" , caller, toCString()); |
983 | else |
984 | LLDB_LOGF(log, "python exception: %s" , toCString()); |
985 | } |
986 | void PythonException::Restore() { |
987 | if (m_exception_type && m_exception) { |
988 | PyErr_Restore(m_exception_type, m_exception, m_traceback); |
989 | } else { |
990 | PyErr_SetString(exception: PyExc_Exception, string: toCString()); |
991 | } |
992 | m_exception_type = m_exception = m_traceback = nullptr; |
993 | } |
994 | |
995 | PythonException::~PythonException() { |
996 | Py_XDECREF(m_exception_type); |
997 | Py_XDECREF(m_exception); |
998 | Py_XDECREF(m_traceback); |
999 | Py_XDECREF(m_repr_bytes); |
1000 | } |
1001 | |
1002 | void PythonException::log(llvm::raw_ostream &OS) const { OS << toCString(); } |
1003 | |
1004 | std::error_code PythonException::convertToErrorCode() const { |
1005 | return llvm::inconvertibleErrorCode(); |
1006 | } |
1007 | |
1008 | bool PythonException::Matches(PyObject *exc) const { |
1009 | return PyErr_GivenExceptionMatches(m_exception_type, exc); |
1010 | } |
1011 | |
1012 | const char read_exception_script[] = R"( |
1013 | import sys |
1014 | from traceback import print_exception |
1015 | if sys.version_info.major < 3: |
1016 | from StringIO import StringIO |
1017 | else: |
1018 | from io import StringIO |
1019 | def main(exc_type, exc_value, tb): |
1020 | f = StringIO() |
1021 | print_exception(exc_type, exc_value, tb, file=f) |
1022 | return f.getvalue() |
1023 | )" ; |
1024 | |
1025 | std::string PythonException::ReadBacktrace() const { |
1026 | |
1027 | if (!m_traceback) |
1028 | return toCString(); |
1029 | |
1030 | // no need to synchronize access to this global, we already have the GIL |
1031 | static PythonScript read_exception(read_exception_script); |
1032 | |
1033 | Expected<std::string> backtrace = As<std::string>( |
1034 | obj: read_exception(m_exception_type, m_exception, m_traceback)); |
1035 | |
1036 | if (!backtrace) { |
1037 | std::string message = |
1038 | std::string(toCString()) + "\n" + |
1039 | "Traceback unavailable, an error occurred while reading it:\n" ; |
1040 | return (message + llvm::toString(E: backtrace.takeError())); |
1041 | } |
1042 | |
1043 | return std::move(backtrace.get()); |
1044 | } |
1045 | |
1046 | char PythonException::ID = 0; |
1047 | |
1048 | llvm::Expected<File::OpenOptions> |
1049 | GetOptionsForPyObject(const PythonObject &obj) { |
1050 | auto options = File::OpenOptions(0); |
1051 | auto readable = As<bool>(obj: obj.CallMethod(name: "readable" )); |
1052 | if (!readable) |
1053 | return readable.takeError(); |
1054 | auto writable = As<bool>(obj: obj.CallMethod(name: "writable" )); |
1055 | if (!writable) |
1056 | return writable.takeError(); |
1057 | if (readable.get() && writable.get()) |
1058 | options |= File::eOpenOptionReadWrite; |
1059 | else if (writable.get()) |
1060 | options |= File::eOpenOptionWriteOnly; |
1061 | else if (readable.get()) |
1062 | options |= File::eOpenOptionReadOnly; |
1063 | return options; |
1064 | } |
1065 | |
1066 | // Base class template for python files. All it knows how to do |
1067 | // is hold a reference to the python object and close or flush it |
1068 | // when the File is closed. |
1069 | namespace { |
1070 | template <typename Base> class OwnedPythonFile : public Base { |
1071 | public: |
1072 | template <typename... Args> |
1073 | OwnedPythonFile(const PythonFile &file, bool borrowed, Args... args) |
1074 | : Base(args...), m_py_obj(file), m_borrowed(borrowed) { |
1075 | assert(m_py_obj); |
1076 | } |
1077 | |
1078 | ~OwnedPythonFile() override { |
1079 | assert(m_py_obj); |
1080 | GIL takeGIL; |
1081 | Close(); |
1082 | // we need to ensure the python object is released while we still |
1083 | // hold the GIL |
1084 | m_py_obj.Reset(); |
1085 | } |
1086 | |
1087 | bool IsPythonSideValid() const { |
1088 | GIL takeGIL; |
1089 | auto closed = As<bool>(obj: m_py_obj.GetAttribute(name: "closed" )); |
1090 | if (!closed) { |
1091 | llvm::consumeError(Err: closed.takeError()); |
1092 | return false; |
1093 | } |
1094 | return !closed.get(); |
1095 | } |
1096 | |
1097 | bool IsValid() const override { |
1098 | return IsPythonSideValid() && Base::IsValid(); |
1099 | } |
1100 | |
1101 | Status Close() override { |
1102 | assert(m_py_obj); |
1103 | Status py_error, base_error; |
1104 | GIL takeGIL; |
1105 | if (!m_borrowed) { |
1106 | auto r = m_py_obj.CallMethod(name: "close" ); |
1107 | if (!r) |
1108 | py_error = Status(r.takeError()); |
1109 | } |
1110 | base_error = Base::Close(); |
1111 | if (py_error.Fail()) |
1112 | return py_error; |
1113 | return base_error; |
1114 | }; |
1115 | |
1116 | PyObject *GetPythonObject() const { |
1117 | assert(m_py_obj.IsValid()); |
1118 | return m_py_obj.get(); |
1119 | } |
1120 | |
1121 | static bool classof(const File *file) = delete; |
1122 | |
1123 | protected: |
1124 | PythonFile m_py_obj; |
1125 | bool m_borrowed; |
1126 | }; |
1127 | } // namespace |
1128 | |
1129 | // A SimplePythonFile is a OwnedPythonFile that just does all I/O as |
1130 | // a NativeFile |
1131 | namespace { |
1132 | class SimplePythonFile : public OwnedPythonFile<NativeFile> { |
1133 | public: |
1134 | SimplePythonFile(const PythonFile &file, bool borrowed, int fd, |
1135 | File::OpenOptions options) |
1136 | : OwnedPythonFile(file, borrowed, fd, options, false) {} |
1137 | |
1138 | static char ID; |
1139 | bool isA(const void *classID) const override { |
1140 | return classID == &ID || NativeFile::isA(classID); |
1141 | } |
1142 | static bool classof(const File *file) { return file->isA(classID: &ID); } |
1143 | }; |
1144 | char SimplePythonFile::ID = 0; |
1145 | } // namespace |
1146 | |
1147 | namespace { |
1148 | class PythonBuffer { |
1149 | public: |
1150 | PythonBuffer &operator=(const PythonBuffer &) = delete; |
1151 | PythonBuffer(const PythonBuffer &) = delete; |
1152 | |
1153 | static Expected<PythonBuffer> Create(PythonObject &obj, |
1154 | int flags = PyBUF_SIMPLE) { |
1155 | Py_buffer py_buffer = {}; |
1156 | PyObject_GetBuffer(obj: obj.get(), view: &py_buffer, flags); |
1157 | if (!py_buffer.obj) |
1158 | return llvm::make_error<PythonException>(); |
1159 | return PythonBuffer(py_buffer); |
1160 | } |
1161 | |
1162 | PythonBuffer(PythonBuffer &&other) { |
1163 | m_buffer = other.m_buffer; |
1164 | other.m_buffer.obj = nullptr; |
1165 | } |
1166 | |
1167 | ~PythonBuffer() { |
1168 | if (m_buffer.obj) |
1169 | PyBuffer_Release(view: &m_buffer); |
1170 | } |
1171 | |
1172 | Py_buffer &get() { return m_buffer; } |
1173 | |
1174 | private: |
1175 | // takes ownership of the buffer. |
1176 | PythonBuffer(const Py_buffer &py_buffer) : m_buffer(py_buffer) {} |
1177 | Py_buffer m_buffer; |
1178 | }; |
1179 | } // namespace |
1180 | |
1181 | // Shared methods between TextPythonFile and BinaryPythonFile |
1182 | namespace { |
1183 | class PythonIOFile : public OwnedPythonFile<File> { |
1184 | public: |
1185 | PythonIOFile(const PythonFile &file, bool borrowed) |
1186 | : OwnedPythonFile(file, borrowed) {} |
1187 | |
1188 | ~PythonIOFile() override { Close(); } |
1189 | |
1190 | bool IsValid() const override { return IsPythonSideValid(); } |
1191 | |
1192 | Status Close() override { |
1193 | assert(m_py_obj); |
1194 | GIL takeGIL; |
1195 | if (m_borrowed) |
1196 | return Flush(); |
1197 | auto r = m_py_obj.CallMethod(name: "close" ); |
1198 | if (!r) |
1199 | return Status(r.takeError()); |
1200 | return Status(); |
1201 | } |
1202 | |
1203 | Status Flush() override { |
1204 | GIL takeGIL; |
1205 | auto r = m_py_obj.CallMethod(name: "flush" ); |
1206 | if (!r) |
1207 | return Status(r.takeError()); |
1208 | return Status(); |
1209 | } |
1210 | |
1211 | Expected<File::OpenOptions> GetOptions() const override { |
1212 | GIL takeGIL; |
1213 | return GetOptionsForPyObject(obj: m_py_obj); |
1214 | } |
1215 | |
1216 | static char ID; |
1217 | bool isA(const void *classID) const override { |
1218 | return classID == &ID || File::isA(classID); |
1219 | } |
1220 | static bool classof(const File *file) { return file->isA(classID: &ID); } |
1221 | }; |
1222 | char PythonIOFile::ID = 0; |
1223 | } // namespace |
1224 | |
1225 | namespace { |
1226 | class BinaryPythonFile : public PythonIOFile { |
1227 | protected: |
1228 | int m_descriptor; |
1229 | |
1230 | public: |
1231 | BinaryPythonFile(int fd, const PythonFile &file, bool borrowed) |
1232 | : PythonIOFile(file, borrowed), |
1233 | m_descriptor(File::DescriptorIsValid(descriptor: fd) ? fd |
1234 | : File::kInvalidDescriptor) {} |
1235 | |
1236 | int GetDescriptor() const override { return m_descriptor; } |
1237 | |
1238 | Status Write(const void *buf, size_t &num_bytes) override { |
1239 | GIL takeGIL; |
1240 | PyObject *pybuffer_p = PyMemoryView_FromMemory( |
1241 | mem: const_cast<char *>((const char *)buf), size: num_bytes, PyBUF_READ); |
1242 | if (!pybuffer_p) |
1243 | return Status(llvm::make_error<PythonException>()); |
1244 | auto pybuffer = Take<PythonObject>(obj: pybuffer_p); |
1245 | num_bytes = 0; |
1246 | auto bytes_written = As<long long>(obj: m_py_obj.CallMethod(name: "write" , t: pybuffer)); |
1247 | if (!bytes_written) |
1248 | return Status(bytes_written.takeError()); |
1249 | if (bytes_written.get() < 0) |
1250 | return Status(".write() method returned a negative number!" ); |
1251 | static_assert(sizeof(long long) >= sizeof(size_t), "overflow" ); |
1252 | num_bytes = bytes_written.get(); |
1253 | return Status(); |
1254 | } |
1255 | |
1256 | Status Read(void *buf, size_t &num_bytes) override { |
1257 | GIL takeGIL; |
1258 | static_assert(sizeof(long long) >= sizeof(size_t), "overflow" ); |
1259 | auto pybuffer_obj = |
1260 | m_py_obj.CallMethod(name: "read" , t: (unsigned long long)num_bytes); |
1261 | if (!pybuffer_obj) |
1262 | return Status(pybuffer_obj.takeError()); |
1263 | num_bytes = 0; |
1264 | if (pybuffer_obj.get().IsNone()) { |
1265 | // EOF |
1266 | num_bytes = 0; |
1267 | return Status(); |
1268 | } |
1269 | auto pybuffer = PythonBuffer::Create(obj&: pybuffer_obj.get()); |
1270 | if (!pybuffer) |
1271 | return Status(pybuffer.takeError()); |
1272 | memcpy(dest: buf, src: pybuffer.get().get().buf, n: pybuffer.get().get().len); |
1273 | num_bytes = pybuffer.get().get().len; |
1274 | return Status(); |
1275 | } |
1276 | }; |
1277 | } // namespace |
1278 | |
1279 | namespace { |
1280 | class TextPythonFile : public PythonIOFile { |
1281 | protected: |
1282 | int m_descriptor; |
1283 | |
1284 | public: |
1285 | TextPythonFile(int fd, const PythonFile &file, bool borrowed) |
1286 | : PythonIOFile(file, borrowed), |
1287 | m_descriptor(File::DescriptorIsValid(descriptor: fd) ? fd |
1288 | : File::kInvalidDescriptor) {} |
1289 | |
1290 | int GetDescriptor() const override { return m_descriptor; } |
1291 | |
1292 | Status Write(const void *buf, size_t &num_bytes) override { |
1293 | GIL takeGIL; |
1294 | auto pystring = |
1295 | PythonString::FromUTF8(string: llvm::StringRef((const char *)buf, num_bytes)); |
1296 | if (!pystring) |
1297 | return Status(pystring.takeError()); |
1298 | num_bytes = 0; |
1299 | auto bytes_written = |
1300 | As<long long>(obj: m_py_obj.CallMethod(name: "write" , t: pystring.get())); |
1301 | if (!bytes_written) |
1302 | return Status(bytes_written.takeError()); |
1303 | if (bytes_written.get() < 0) |
1304 | return Status(".write() method returned a negative number!" ); |
1305 | static_assert(sizeof(long long) >= sizeof(size_t), "overflow" ); |
1306 | num_bytes = bytes_written.get(); |
1307 | return Status(); |
1308 | } |
1309 | |
1310 | Status Read(void *buf, size_t &num_bytes) override { |
1311 | GIL takeGIL; |
1312 | size_t num_chars = num_bytes / 6; |
1313 | size_t orig_num_bytes = num_bytes; |
1314 | num_bytes = 0; |
1315 | if (orig_num_bytes < 6) { |
1316 | return Status("can't read less than 6 bytes from a utf8 text stream" ); |
1317 | } |
1318 | auto pystring = As<PythonString>( |
1319 | obj: m_py_obj.CallMethod(name: "read" , t: (unsigned long long)num_chars)); |
1320 | if (!pystring) |
1321 | return Status(pystring.takeError()); |
1322 | if (pystring.get().IsNone()) { |
1323 | // EOF |
1324 | return Status(); |
1325 | } |
1326 | auto stringref = pystring.get().AsUTF8(); |
1327 | if (!stringref) |
1328 | return Status(stringref.takeError()); |
1329 | num_bytes = stringref.get().size(); |
1330 | memcpy(dest: buf, src: stringref.get().begin(), n: num_bytes); |
1331 | return Status(); |
1332 | } |
1333 | }; |
1334 | } // namespace |
1335 | |
1336 | llvm::Expected<FileSP> PythonFile::ConvertToFile(bool borrowed) { |
1337 | if (!IsValid()) |
1338 | return llvm::createStringError(EC: llvm::inconvertibleErrorCode(), |
1339 | Msg: "invalid PythonFile" ); |
1340 | |
1341 | int fd = PyObject_AsFileDescriptor(m_py_obj); |
1342 | if (fd < 0) { |
1343 | PyErr_Clear(); |
1344 | return ConvertToFileForcingUseOfScriptingIOMethods(borrowed); |
1345 | } |
1346 | auto options = GetOptionsForPyObject(obj: *this); |
1347 | if (!options) |
1348 | return options.takeError(); |
1349 | |
1350 | File::OpenOptions rw = |
1351 | options.get() & (File::eOpenOptionReadOnly | File::eOpenOptionWriteOnly | |
1352 | File::eOpenOptionReadWrite); |
1353 | if (rw == File::eOpenOptionWriteOnly || rw == File::eOpenOptionReadWrite) { |
1354 | // LLDB and python will not share I/O buffers. We should probably |
1355 | // flush the python buffers now. |
1356 | auto r = CallMethod(name: "flush" ); |
1357 | if (!r) |
1358 | return r.takeError(); |
1359 | } |
1360 | |
1361 | FileSP file_sp; |
1362 | if (borrowed) { |
1363 | // In this case we don't need to retain the python |
1364 | // object at all. |
1365 | file_sp = std::make_shared<NativeFile>(args&: fd, args&: options.get(), args: false); |
1366 | } else { |
1367 | file_sp = std::static_pointer_cast<File>( |
1368 | r: std::make_shared<SimplePythonFile>(args&: *this, args&: borrowed, args&: fd, args&: options.get())); |
1369 | } |
1370 | if (!file_sp->IsValid()) |
1371 | return llvm::createStringError(EC: llvm::inconvertibleErrorCode(), |
1372 | Msg: "invalid File" ); |
1373 | |
1374 | return file_sp; |
1375 | } |
1376 | |
1377 | llvm::Expected<FileSP> |
1378 | PythonFile::ConvertToFileForcingUseOfScriptingIOMethods(bool borrowed) { |
1379 | |
1380 | assert(!PyErr_Occurred()); |
1381 | |
1382 | if (!IsValid()) |
1383 | return llvm::createStringError(EC: llvm::inconvertibleErrorCode(), |
1384 | Msg: "invalid PythonFile" ); |
1385 | |
1386 | int fd = PyObject_AsFileDescriptor(m_py_obj); |
1387 | if (fd < 0) { |
1388 | PyErr_Clear(); |
1389 | fd = File::kInvalidDescriptor; |
1390 | } |
1391 | |
1392 | auto io_module = PythonModule::Import(name: "io" ); |
1393 | if (!io_module) |
1394 | return io_module.takeError(); |
1395 | auto textIOBase = io_module.get().Get(name: "TextIOBase" ); |
1396 | if (!textIOBase) |
1397 | return textIOBase.takeError(); |
1398 | auto rawIOBase = io_module.get().Get(name: "RawIOBase" ); |
1399 | if (!rawIOBase) |
1400 | return rawIOBase.takeError(); |
1401 | auto bufferedIOBase = io_module.get().Get(name: "BufferedIOBase" ); |
1402 | if (!bufferedIOBase) |
1403 | return bufferedIOBase.takeError(); |
1404 | |
1405 | FileSP file_sp; |
1406 | |
1407 | auto isTextIO = IsInstance(cls: textIOBase.get()); |
1408 | if (!isTextIO) |
1409 | return isTextIO.takeError(); |
1410 | if (isTextIO.get()) |
1411 | file_sp = std::static_pointer_cast<File>( |
1412 | r: std::make_shared<TextPythonFile>(args&: fd, args&: *this, args&: borrowed)); |
1413 | |
1414 | auto isRawIO = IsInstance(cls: rawIOBase.get()); |
1415 | if (!isRawIO) |
1416 | return isRawIO.takeError(); |
1417 | auto isBufferedIO = IsInstance(cls: bufferedIOBase.get()); |
1418 | if (!isBufferedIO) |
1419 | return isBufferedIO.takeError(); |
1420 | |
1421 | if (isRawIO.get() || isBufferedIO.get()) { |
1422 | file_sp = std::static_pointer_cast<File>( |
1423 | r: std::make_shared<BinaryPythonFile>(args&: fd, args&: *this, args&: borrowed)); |
1424 | } |
1425 | |
1426 | if (!file_sp) |
1427 | return llvm::createStringError(EC: llvm::inconvertibleErrorCode(), |
1428 | Msg: "python file is neither text nor binary" ); |
1429 | |
1430 | if (!file_sp->IsValid()) |
1431 | return llvm::createStringError(EC: llvm::inconvertibleErrorCode(), |
1432 | Msg: "invalid File" ); |
1433 | |
1434 | return file_sp; |
1435 | } |
1436 | |
1437 | Expected<PythonFile> PythonFile::FromFile(File &file, const char *mode) { |
1438 | if (!file.IsValid()) |
1439 | return llvm::createStringError(EC: llvm::inconvertibleErrorCode(), |
1440 | Msg: "invalid file" ); |
1441 | |
1442 | if (auto *simple = llvm::dyn_cast<SimplePythonFile>(Val: &file)) |
1443 | return Retain<PythonFile>(obj: simple->GetPythonObject()); |
1444 | if (auto *pythonio = llvm::dyn_cast<PythonIOFile>(Val: &file)) |
1445 | return Retain<PythonFile>(obj: pythonio->GetPythonObject()); |
1446 | |
1447 | if (!mode) { |
1448 | auto m = file.GetOpenMode(); |
1449 | if (!m) |
1450 | return m.takeError(); |
1451 | mode = m.get(); |
1452 | } |
1453 | |
1454 | PyObject *file_obj; |
1455 | file_obj = PyFile_FromFd(file.GetDescriptor(), nullptr, mode, -1, nullptr, |
1456 | "ignore" , nullptr, /*closefd=*/0); |
1457 | |
1458 | if (!file_obj) |
1459 | return exception(); |
1460 | |
1461 | return Take<PythonFile>(obj: file_obj); |
1462 | } |
1463 | |
1464 | Error PythonScript::Init() { |
1465 | if (function.IsValid()) |
1466 | return Error::success(); |
1467 | |
1468 | PythonDictionary globals(PyInitialValue::Empty); |
1469 | auto builtins = PythonModule::BuiltinsModule(); |
1470 | if (Error error = globals.SetItem(key: "__builtins__" , value: builtins)) |
1471 | return error; |
1472 | PyObject *o = |
1473 | PyRun_String(script, Py_file_input, globals.get(), globals.get()); |
1474 | if (!o) |
1475 | return exception(); |
1476 | Take<PythonObject>(obj: o); |
1477 | auto f = As<PythonCallable>(obj: globals.GetItem(key: "main" )); |
1478 | if (!f) |
1479 | return f.takeError(); |
1480 | function = std::move(f.get()); |
1481 | |
1482 | return Error::success(); |
1483 | } |
1484 | |
1485 | llvm::Expected<PythonObject> |
1486 | python::runStringOneLine(const llvm::Twine &string, |
1487 | const PythonDictionary &globals, |
1488 | const PythonDictionary &locals) { |
1489 | if (!globals.IsValid() || !locals.IsValid()) |
1490 | return nullDeref(); |
1491 | |
1492 | PyObject *code = |
1493 | Py_CompileString(NullTerminated(string), "<string>" , Py_eval_input); |
1494 | if (!code) { |
1495 | PyErr_Clear(); |
1496 | code = |
1497 | Py_CompileString(NullTerminated(string), "<string>" , Py_single_input); |
1498 | } |
1499 | if (!code) |
1500 | return exception(); |
1501 | auto code_ref = Take<PythonObject>(obj: code); |
1502 | |
1503 | PyObject *result = PyEval_EvalCode(code, globals.get(), locals.get()); |
1504 | |
1505 | if (!result) |
1506 | return exception(); |
1507 | |
1508 | return Take<PythonObject>(obj: result); |
1509 | } |
1510 | |
1511 | llvm::Expected<PythonObject> |
1512 | python::runStringMultiLine(const llvm::Twine &string, |
1513 | const PythonDictionary &globals, |
1514 | const PythonDictionary &locals) { |
1515 | if (!globals.IsValid() || !locals.IsValid()) |
1516 | return nullDeref(); |
1517 | PyObject *result = PyRun_String(NullTerminated(string), Py_file_input, |
1518 | globals.get(), locals.get()); |
1519 | if (!result) |
1520 | return exception(); |
1521 | return Take<PythonObject>(obj: result); |
1522 | } |
1523 | |
1524 | #endif |
1525 | |