| 1 | """
|
| 2 | mylib.py: Python stubs/interfaces that are reimplemented in C++, not directly
|
| 3 | translated.
|
| 4 | """
|
| 5 | from __future__ import print_function
|
| 6 |
|
| 7 | try:
|
| 8 | import cStringIO
|
| 9 | except ImportError:
|
| 10 | # Python 3 doesn't have cStringIO. Our yaks/ demo currently uses
|
| 11 | # mycpp/mylib.py with Python 3.
|
| 12 | cStringIO = None
|
| 13 | import io
|
| 14 |
|
| 15 | import sys
|
| 16 |
|
| 17 | from pylib import collections_
|
| 18 | try:
|
| 19 | import posix_ as posix
|
| 20 | except ImportError:
|
| 21 | # Hack for tangled dependencies.
|
| 22 | import os
|
| 23 | posix = os
|
| 24 |
|
| 25 | from typing import (Tuple, List, Dict, Optional, Iterator, Any, TypeVar,
|
| 26 | Generic, cast, TYPE_CHECKING)
|
| 27 | if TYPE_CHECKING:
|
| 28 | from mycpp import mops
|
| 29 |
|
| 30 | # For conditional translation
|
| 31 | CPP = False
|
| 32 | PYTHON = True
|
| 33 |
|
| 34 | # Use POSIX name directly
|
| 35 | STDIN_FILENO = 0
|
| 36 |
|
| 37 |
|
| 38 | def MaybeCollect():
|
| 39 | # type: () -> None
|
| 40 | pass
|
| 41 |
|
| 42 |
|
| 43 | def PrintGcStats():
|
| 44 | # type: () -> None
|
| 45 | pass
|
| 46 |
|
| 47 |
|
| 48 | def NewDict():
|
| 49 | # type: () -> Dict[str, Any]
|
| 50 | """Make dictionaries ordered in Python, e.g. for JSON.
|
| 51 |
|
| 52 | In C++, our Dict implementation should be ordered.
|
| 53 | """
|
| 54 | return collections_.OrderedDict()
|
| 55 |
|
| 56 |
|
| 57 | def log(msg, *args):
|
| 58 | # type: (str, *Any) -> None
|
| 59 | """Print debug output to stderr."""
|
| 60 | if args:
|
| 61 | msg = msg % args
|
| 62 | print(msg, file=sys.stderr)
|
| 63 |
|
| 64 |
|
| 65 | def print_stderr(s):
|
| 66 | # type: (str) -> None
|
| 67 | """Print a message to stderr for the user.
|
| 68 |
|
| 69 | This should be used sparingly, since it doesn't have location info, like
|
| 70 | ui.ErrorFormatter does. We use it to print fatal I/O errors that were only
|
| 71 | caught at the top level.
|
| 72 | """
|
| 73 | print(s, file=sys.stderr)
|
| 74 |
|
| 75 |
|
| 76 | #
|
| 77 | # Byte Operations avoid excessive allocations with string algorithms
|
| 78 | #
|
| 79 |
|
| 80 |
|
| 81 | def ByteAt(s, i):
|
| 82 | # type: (str, int) -> int
|
| 83 | """i must be in bounds."""
|
| 84 |
|
| 85 | # This simplifies the C++ implementation
|
| 86 | assert 0 <= i, 'No negative indices'
|
| 87 | assert i < len(s), 'No negative indices'
|
| 88 |
|
| 89 | return ord(s[i])
|
| 90 |
|
| 91 |
|
| 92 | def ByteEquals(byte, ch):
|
| 93 | # type: (int, str) -> bool
|
| 94 | assert len(ch) == 1, ch
|
| 95 | assert 0 <= byte < 256, byte
|
| 96 |
|
| 97 | return byte == ord(ch)
|
| 98 |
|
| 99 |
|
| 100 | def ByteInSet(byte, byte_set):
|
| 101 | # type: (int, str) -> bool
|
| 102 | assert 0 <= byte < 256, byte
|
| 103 |
|
| 104 | return chr(byte) in byte_set
|
| 105 |
|
| 106 |
|
| 107 | def JoinBytes(byte_list):
|
| 108 | # type: (List[int]) -> str
|
| 109 |
|
| 110 | return ''.join(chr(b) for b in byte_list)
|
| 111 |
|
| 112 |
|
| 113 | #
|
| 114 | # For SparseArray
|
| 115 | #
|
| 116 |
|
| 117 |
|
| 118 | def BigIntSort(keys):
|
| 119 | # type: (List[mops.BigInt]) -> None
|
| 120 | keys.sort(key=lambda big: big.i)
|
| 121 |
|
| 122 |
|
| 123 | #
|
| 124 | # Files
|
| 125 | #
|
| 126 |
|
| 127 |
|
| 128 | class LineReader:
|
| 129 |
|
| 130 | def readline(self):
|
| 131 | # type: () -> str
|
| 132 | raise NotImplementedError()
|
| 133 |
|
| 134 | def close(self):
|
| 135 | # type: () -> None
|
| 136 | raise NotImplementedError()
|
| 137 |
|
| 138 | def isatty(self):
|
| 139 | # type: () -> bool
|
| 140 | raise NotImplementedError()
|
| 141 |
|
| 142 |
|
| 143 | if TYPE_CHECKING:
|
| 144 |
|
| 145 | class BufLineReader(LineReader):
|
| 146 |
|
| 147 | def __init__(self, s):
|
| 148 | # type: (str) -> None
|
| 149 | raise NotImplementedError()
|
| 150 |
|
| 151 | def open(path):
|
| 152 | # type: (str) -> LineReader
|
| 153 |
|
| 154 | # TODO: should probably return mylib.File
|
| 155 | # mylib.open() is currently only used in yaks/yaks_main and
|
| 156 | # bin.osh_parse
|
| 157 | raise NotImplementedError()
|
| 158 |
|
| 159 | else:
|
| 160 | # Actual runtime
|
| 161 | if cStringIO:
|
| 162 | BufLineReader = cStringIO.StringIO
|
| 163 | else: # Python 3
|
| 164 | BufLineReader = io.StringIO
|
| 165 |
|
| 166 | open = open
|
| 167 |
|
| 168 |
|
| 169 | class Writer:
|
| 170 |
|
| 171 | def write(self, s):
|
| 172 | # type: (str) -> None
|
| 173 | raise NotImplementedError()
|
| 174 |
|
| 175 | def flush(self):
|
| 176 | # type: () -> None
|
| 177 | raise NotImplementedError()
|
| 178 |
|
| 179 | def isatty(self):
|
| 180 | # type: () -> bool
|
| 181 | raise NotImplementedError()
|
| 182 |
|
| 183 | def close(self):
|
| 184 | # type: () -> None
|
| 185 | raise NotImplementedError()
|
| 186 |
|
| 187 |
|
| 188 | class BufWriter(Writer):
|
| 189 | """Mimic StringIO API, but add clear() so we can reuse objects.
|
| 190 |
|
| 191 | We can also add accelerators for directly writing numbers, to avoid
|
| 192 | allocations when encoding JSON.
|
| 193 | """
|
| 194 |
|
| 195 | def __init__(self):
|
| 196 | # type: () -> None
|
| 197 | self.parts = []
|
| 198 |
|
| 199 | def write(self, s):
|
| 200 | # type: (str) -> None
|
| 201 | self.parts.append(s)
|
| 202 |
|
| 203 | def isatty(self):
|
| 204 | # type: () -> bool
|
| 205 | return False
|
| 206 |
|
| 207 | def write_spaces(self, n):
|
| 208 | # type: (int) -> None
|
| 209 | """For JSON indenting. Avoid intermediate allocations in C++."""
|
| 210 | self.parts.append(' ' * n)
|
| 211 |
|
| 212 | def getvalue(self):
|
| 213 | # type: () -> str
|
| 214 | return ''.join(self.parts)
|
| 215 |
|
| 216 | def clear(self):
|
| 217 | # type: () -> None
|
| 218 | del self.parts[:]
|
| 219 |
|
| 220 | def close(self):
|
| 221 | # type: () -> None
|
| 222 |
|
| 223 | # No-op for now - we could invalidate write()?
|
| 224 | pass
|
| 225 |
|
| 226 |
|
| 227 | def Stdout():
|
| 228 | # type: () -> Writer
|
| 229 | return sys.stdout
|
| 230 |
|
| 231 |
|
| 232 | def Stderr():
|
| 233 | # type: () -> Writer
|
| 234 | return sys.stderr
|
| 235 |
|
| 236 |
|
| 237 | def Stdin():
|
| 238 | # type: () -> LineReader
|
| 239 | return sys.stdin
|
| 240 |
|
| 241 |
|
| 242 | class switch(object):
|
| 243 | """Translates to C switch on int.
|
| 244 |
|
| 245 | with switch(i) as case:
|
| 246 | if case(42, 43):
|
| 247 | print('hi')
|
| 248 | elif case(99):
|
| 249 | print('two')
|
| 250 | else:
|
| 251 | print('neither')
|
| 252 | """
|
| 253 |
|
| 254 | def __init__(self, value):
|
| 255 | # type: (int) -> None
|
| 256 | self.value = value
|
| 257 |
|
| 258 | def __enter__(self):
|
| 259 | # type: () -> switch
|
| 260 | return self
|
| 261 |
|
| 262 | def __exit__(self, type, value, traceback):
|
| 263 | # type: (Any, Any, Any) -> bool
|
| 264 | return False # Allows a traceback to occur
|
| 265 |
|
| 266 | def __call__(self, *cases):
|
| 267 | # type: (*Any) -> bool
|
| 268 | return self.value in cases
|
| 269 |
|
| 270 |
|
| 271 | class str_switch(object):
|
| 272 | """Translates to fast dispatch on string length, then memcmp()."""
|
| 273 |
|
| 274 | def __init__(self, value):
|
| 275 | # type: (str) -> None
|
| 276 | self.value = value
|
| 277 |
|
| 278 | def __enter__(self):
|
| 279 | # type: () -> switch
|
| 280 | return self
|
| 281 |
|
| 282 | def __exit__(self, type, value, traceback):
|
| 283 | # type: (Any, Any, Any) -> bool
|
| 284 | return False # Allows a traceback to occur
|
| 285 |
|
| 286 | def __call__(self, *cases):
|
| 287 | # type: (*Any) -> bool
|
| 288 | return self.value in cases
|
| 289 |
|
| 290 |
|
| 291 | class tagswitch(object):
|
| 292 | """Translates to C switch(node->tag())"""
|
| 293 |
|
| 294 | def __init__(self, node):
|
| 295 | # type: (Any) -> None
|
| 296 | self.tag = node.tag()
|
| 297 |
|
| 298 | def __enter__(self):
|
| 299 | # type: () -> tagswitch
|
| 300 | return self
|
| 301 |
|
| 302 | def __exit__(self, type, value, traceback):
|
| 303 | # type: (Any, Any, Any) -> bool
|
| 304 | return False # Allows a traceback to occur
|
| 305 |
|
| 306 | def __call__(self, *cases):
|
| 307 | # type: (*Any) -> bool
|
| 308 | return self.tag in cases
|
| 309 |
|
| 310 |
|
| 311 | if TYPE_CHECKING:
|
| 312 | # Doesn't work
|
| 313 | T = TypeVar('T')
|
| 314 |
|
| 315 | class StackArray(Generic[T]):
|
| 316 |
|
| 317 | def __init__(self):
|
| 318 | self.items = [] # type: List[T]
|
| 319 |
|
| 320 | def append(self, item):
|
| 321 | # type: (T) -> None
|
| 322 | self.items.append(item)
|
| 323 |
|
| 324 | def pop(self):
|
| 325 | # type: () -> T
|
| 326 | return self.items.pop()
|
| 327 |
|
| 328 | # Doesn't work, this is only for primitive types
|
| 329 | #StackArray = NewType('StackArray', list)
|
| 330 |
|
| 331 |
|
| 332 | def MakeStackArray(item_type):
|
| 333 | # type: (TypeVar) -> StackArray[item_type]
|
| 334 | """
|
| 335 | Convenience "constructor" used like this:
|
| 336 |
|
| 337 | myarray = MakeStackArray(int)
|
| 338 |
|
| 339 | The idiom could also be
|
| 340 |
|
| 341 | myarray = cast('StackArray[int]', [])
|
| 342 |
|
| 343 | But that's uglier.
|
| 344 | """
|
| 345 | return cast('StackArray[item_type]', [])
|
| 346 |
|
| 347 |
|
| 348 | if TYPE_CHECKING:
|
| 349 | K = TypeVar('K')
|
| 350 | V = TypeVar('V')
|
| 351 |
|
| 352 |
|
| 353 | def iteritems(d):
|
| 354 | # type: (Dict[K, V]) -> Iterator[Tuple[K, V]]
|
| 355 | """Make translation a bit easier."""
|
| 356 | return d.iteritems()
|
| 357 |
|
| 358 |
|
| 359 | def split_once(s, delim):
|
| 360 | # type: (str, str) -> Tuple[str, Optional[str]]
|
| 361 | """Easier to call than split(s, 1) because of tuple unpacking."""
|
| 362 |
|
| 363 | parts = s.split(delim, 1)
|
| 364 | if len(parts) == 1:
|
| 365 | no_str = None # type: Optional[str]
|
| 366 | return s, no_str
|
| 367 | else:
|
| 368 | return parts[0], parts[1]
|
| 369 |
|
| 370 |
|
| 371 | def hex_lower(i):
|
| 372 | # type: (int) -> str
|
| 373 | return '%x' % i
|
| 374 |
|
| 375 |
|
| 376 | def dict_erase(d, key):
|
| 377 | # type: (Dict[Any, Any], Any) -> None
|
| 378 | """
|
| 379 | Ensure that a key isn't in the Dict d. This makes C++ translation easier.
|
| 380 | """
|
| 381 | try:
|
| 382 | del d[key]
|
| 383 | except KeyError:
|
| 384 | pass
|
| 385 |
|
| 386 |
|
| 387 | def str_cmp(s1, s2):
|
| 388 | # type: (str, str) -> int
|
| 389 | if s1 == s2:
|
| 390 | return 0
|
| 391 | if s1 < s2:
|
| 392 | return -1
|
| 393 | else:
|
| 394 | return 1
|
| 395 |
|
| 396 |
|
| 397 | class UniqueObjects(object):
|
| 398 | """A set of objects identified by their address in memory
|
| 399 |
|
| 400 | Python's id(obj) returns the address of any object. But we don't simply
|
| 401 | implement it, because it requires a uint64_t on 64-bit systems, while mycpp
|
| 402 | only supports 'int'.
|
| 403 |
|
| 404 | So we have a whole class.
|
| 405 |
|
| 406 | Should be used for:
|
| 407 |
|
| 408 | - Cycle detection when pretty printing, as Python's repr() does
|
| 409 | - See CPython's Objects/object.c PyObject_Repr()
|
| 410 | /* These methods are used to control infinite recursion in repr, str, print,
|
| 411 | etc. Container objects that may recursively contain themselves,
|
| 412 | e.g. builtin dictionaries and lists, should use Py_ReprEnter() and
|
| 413 | Py_ReprLeave() to avoid infinite recursion.
|
| 414 | */
|
| 415 | - e.g. dictobject.c dict_repr() calls Py_ReprEnter() to print {...}
|
| 416 | - In Python 2.7 a GLOBAL VAR is used
|
| 417 |
|
| 418 | - It also checks for STACK OVERFLOW
|
| 419 |
|
| 420 | - Packle serialization
|
| 421 | """
|
| 422 |
|
| 423 | def __init__(self):
|
| 424 | # 64-bit id() -> small integer ID
|
| 425 | self.addresses = {} # type: Dict[int, int]
|
| 426 |
|
| 427 | def Contains(self, obj):
|
| 428 | # type: (Any) -> bool
|
| 429 | """ Convenience? """
|
| 430 | return self.Get(obj) != -1
|
| 431 |
|
| 432 | def MaybeAdd(self, obj):
|
| 433 | # type: (Any) -> None
|
| 434 | """ Convenience? """
|
| 435 |
|
| 436 | # def AddNewObject(self, obj):
|
| 437 | def Add(self, obj):
|
| 438 | # type: (Any) -> None
|
| 439 | """
|
| 440 | Assert it isn't already there, and assign a new ID!
|
| 441 |
|
| 442 | # Lib/pickle does:
|
| 443 |
|
| 444 | self.memo[id(obj)] = memo_len, obj
|
| 445 |
|
| 446 | I guess that's the object ID and a void*
|
| 447 |
|
| 448 | Then it does:
|
| 449 |
|
| 450 | x = self.memo.get(id(obj))
|
| 451 |
|
| 452 | and
|
| 453 |
|
| 454 | # If the object is already in the memo, this means it is
|
| 455 | # recursive. In this case, throw away everything we put on the
|
| 456 | # stack, and fetch the object back from the memo.
|
| 457 | if id(obj) in self.memo:
|
| 458 | write(POP + self.get(self.memo[id(obj)][0]))
|
| 459 |
|
| 460 | BUT It only uses the numeric ID!
|
| 461 | """
|
| 462 | addr = id(obj)
|
| 463 | assert addr not in self.addresses
|
| 464 | self.addresses[addr] = len(self.addresses)
|
| 465 |
|
| 466 | def Get(self, obj):
|
| 467 | # type: (Any) -> int
|
| 468 | """
|
| 469 | Returns unique ID assigned
|
| 470 |
|
| 471 | Returns -1 if it doesn't exist?
|
| 472 | """
|
| 473 | addr = id(obj)
|
| 474 | return self.addresses.get(addr, -1)
|
| 475 |
|
| 476 | # Note: self.memo.clear() doesn't appear to be used
|
| 477 |
|
| 478 |
|
| 479 | def probe(provider, name, *args):
|
| 480 | # type: (str, str, Any) -> None
|
| 481 | """Create a probe for use with profilers like linux perf and ebpf or dtrace."""
|
| 482 | # Noop. Just a marker for mycpp to emit a DTRACE_PROBE()
|
| 483 | return
|
| 484 |
|
| 485 |
|
| 486 | class File:
|
| 487 | """
|
| 488 | TODO: This should define a read/write interface, and then LineReader() and
|
| 489 | Writer() can possibly inherit it, with runtime assertions
|
| 490 |
|
| 491 | Then we allow downcasting from File -> LineReader, like we currently do in
|
| 492 | C++ in gc_mylib.h.
|
| 493 |
|
| 494 | Inheritance can't express the structural Reader/Writer pattern of Go, which
|
| 495 | would be better. I suppose we could use File* everywhere, but having
|
| 496 | fine-grained types is nicer. And there will be very few casts.
|
| 497 | """
|
| 498 | pass
|