| 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 math
|
| 16 | import sys
|
| 17 | from stat import S_ISREG
|
| 18 |
|
| 19 | from pylib import collections_
|
| 20 | try:
|
| 21 | import posix_ as posix
|
| 22 | except ImportError:
|
| 23 | # Hack for tangled dependencies.
|
| 24 | import os
|
| 25 | posix = os
|
| 26 |
|
| 27 | from typing import (Tuple, List, Dict, Optional, Iterator, Any, TypeVar,
|
| 28 | Generic, cast, TYPE_CHECKING)
|
| 29 | if TYPE_CHECKING:
|
| 30 | from mycpp import mops
|
| 31 |
|
| 32 | # For conditional translation
|
| 33 | CPP = False
|
| 34 | PYTHON = True
|
| 35 |
|
| 36 | # Use POSIX name directly
|
| 37 | STDIN_FILENO = 0
|
| 38 |
|
| 39 |
|
| 40 | # Avoid name conflicts with C Macros
|
| 41 | def isinf_(x):
|
| 42 | # type: (float) -> bool
|
| 43 | return math.isinf(x)
|
| 44 |
|
| 45 |
|
| 46 | def isnan_(x):
|
| 47 | # type: (float) -> bool
|
| 48 | return math.isnan(x)
|
| 49 |
|
| 50 |
|
| 51 | def MaybeCollect():
|
| 52 | # type: () -> None
|
| 53 | pass
|
| 54 |
|
| 55 |
|
| 56 | def PrintGcStats():
|
| 57 | # type: () -> None
|
| 58 | pass
|
| 59 |
|
| 60 |
|
| 61 | def NewDict():
|
| 62 | # type: () -> Dict[str, Any]
|
| 63 | """Make dictionaries ordered in Python, e.g. for JSON.
|
| 64 |
|
| 65 | In C++, our Dict implementation should be ordered.
|
| 66 | """
|
| 67 | return collections_.OrderedDict()
|
| 68 |
|
| 69 |
|
| 70 | def log(msg, *args):
|
| 71 | # type: (str, *Any) -> None
|
| 72 | """Print debug output to stderr."""
|
| 73 | if args:
|
| 74 | msg = msg % args
|
| 75 | print(msg, file=sys.stderr)
|
| 76 |
|
| 77 |
|
| 78 | def print_stderr(s):
|
| 79 | # type: (str) -> None
|
| 80 | """Print a message to stderr for the user.
|
| 81 |
|
| 82 | This should be used sparingly, since it doesn't have location info, like
|
| 83 | ui.ErrorFormatter does. We use it to print fatal I/O errors that were only
|
| 84 | caught at the top level.
|
| 85 | """
|
| 86 | print(s, file=sys.stderr)
|
| 87 |
|
| 88 |
|
| 89 | #
|
| 90 | # Byte Operations avoid excessive allocations with string algorithms
|
| 91 | #
|
| 92 |
|
| 93 |
|
| 94 | def ByteAt(s, i):
|
| 95 | # type: (str, int) -> int
|
| 96 | """i must be in bounds."""
|
| 97 |
|
| 98 | # This simplifies the C++ implementation
|
| 99 | assert 0 <= i, 'No negative indices'
|
| 100 | assert i < len(s), 'No negative indices'
|
| 101 |
|
| 102 | return ord(s[i])
|
| 103 |
|
| 104 |
|
| 105 | def ByteEquals(byte, ch):
|
| 106 | # type: (int, str) -> bool
|
| 107 | assert len(ch) == 1, ch
|
| 108 | assert 0 <= byte < 256, byte
|
| 109 |
|
| 110 | return byte == ord(ch)
|
| 111 |
|
| 112 |
|
| 113 | def ByteInSet(byte, byte_set):
|
| 114 | # type: (int, str) -> bool
|
| 115 | assert 0 <= byte < 256, byte
|
| 116 |
|
| 117 | return chr(byte) in byte_set
|
| 118 |
|
| 119 |
|
| 120 | def JoinBytes(byte_list):
|
| 121 | # type: (List[int]) -> str
|
| 122 |
|
| 123 | return ''.join(chr(b) for b in byte_list)
|
| 124 |
|
| 125 |
|
| 126 | #
|
| 127 | # For BashArray
|
| 128 | #
|
| 129 |
|
| 130 |
|
| 131 | def BigIntSort(keys):
|
| 132 | # type: (List[mops.BigInt]) -> None
|
| 133 | keys.sort(key=lambda big: big.i)
|
| 134 |
|
| 135 |
|
| 136 | #
|
| 137 | # Files
|
| 138 | #
|
| 139 |
|
| 140 |
|
| 141 | class LineReader:
|
| 142 |
|
| 143 | def readline(self):
|
| 144 | # type: () -> str
|
| 145 | raise NotImplementedError()
|
| 146 |
|
| 147 | def close(self):
|
| 148 | # type: () -> None
|
| 149 | raise NotImplementedError()
|
| 150 |
|
| 151 | def isatty(self):
|
| 152 | # type: () -> bool
|
| 153 | raise NotImplementedError()
|
| 154 |
|
| 155 |
|
| 156 | if TYPE_CHECKING:
|
| 157 |
|
| 158 | class BufLineReader(LineReader):
|
| 159 |
|
| 160 | def __init__(self, s):
|
| 161 | # type: (str) -> None
|
| 162 | raise NotImplementedError()
|
| 163 |
|
| 164 | def open(path):
|
| 165 | # type: (str) -> LineReader
|
| 166 |
|
| 167 | # TODO: should probably return mylib.File
|
| 168 | # mylib.open() is currently only used in yaks/yaks_main and
|
| 169 | # bin.osh_parse
|
| 170 | raise NotImplementedError()
|
| 171 |
|
| 172 | else:
|
| 173 | # Actual runtime
|
| 174 | if cStringIO:
|
| 175 | BufLineReader = cStringIO.StringIO
|
| 176 | else: # Python 3
|
| 177 | BufLineReader = io.StringIO
|
| 178 |
|
| 179 | open = open
|
| 180 |
|
| 181 |
|
| 182 | class Writer:
|
| 183 |
|
| 184 | def write(self, s):
|
| 185 | # type: (str) -> None
|
| 186 | raise NotImplementedError()
|
| 187 |
|
| 188 | def flush(self):
|
| 189 | # type: () -> None
|
| 190 | raise NotImplementedError()
|
| 191 |
|
| 192 | def isatty(self):
|
| 193 | # type: () -> bool
|
| 194 | raise NotImplementedError()
|
| 195 |
|
| 196 | def close(self):
|
| 197 | # type: () -> None
|
| 198 | raise NotImplementedError()
|
| 199 |
|
| 200 |
|
| 201 | class BufWriter(Writer):
|
| 202 | """Mimic StringIO API, but add clear() so we can reuse objects.
|
| 203 |
|
| 204 | We can also add accelerators for directly writing numbers, to avoid
|
| 205 | allocations when encoding JSON.
|
| 206 | """
|
| 207 |
|
| 208 | def __init__(self):
|
| 209 | # type: () -> None
|
| 210 | self.parts = []
|
| 211 |
|
| 212 | def write(self, s):
|
| 213 | # type: (str) -> None
|
| 214 | self.parts.append(s)
|
| 215 |
|
| 216 | def isatty(self):
|
| 217 | # type: () -> bool
|
| 218 | return False
|
| 219 |
|
| 220 | def write_spaces(self, n):
|
| 221 | # type: (int) -> None
|
| 222 | """For JSON indenting. Avoid intermediate allocations in C++."""
|
| 223 | self.parts.append(' ' * n)
|
| 224 |
|
| 225 | def getvalue(self):
|
| 226 | # type: () -> str
|
| 227 | return ''.join(self.parts)
|
| 228 |
|
| 229 | def clear(self):
|
| 230 | # type: () -> None
|
| 231 | del self.parts[:]
|
| 232 |
|
| 233 | def close(self):
|
| 234 | # type: () -> None
|
| 235 |
|
| 236 | # No-op for now - we could invalidate write()?
|
| 237 | pass
|
| 238 |
|
| 239 |
|
| 240 | def Stdout():
|
| 241 | # type: () -> Writer
|
| 242 | return sys.stdout
|
| 243 |
|
| 244 |
|
| 245 | def Stderr():
|
| 246 | # type: () -> Writer
|
| 247 | return sys.stderr
|
| 248 |
|
| 249 |
|
| 250 | def Stdin():
|
| 251 | # type: () -> LineReader
|
| 252 | return sys.stdin
|
| 253 |
|
| 254 |
|
| 255 | class switch(object):
|
| 256 | """Translates to C switch on int.
|
| 257 |
|
| 258 | with switch(i) as case:
|
| 259 | if case(42, 43):
|
| 260 | print('hi')
|
| 261 | elif case(99):
|
| 262 | print('two')
|
| 263 | else:
|
| 264 | print('neither')
|
| 265 | """
|
| 266 |
|
| 267 | def __init__(self, value):
|
| 268 | # type: (int) -> None
|
| 269 | self.value = value
|
| 270 |
|
| 271 | def __enter__(self):
|
| 272 | # type: () -> switch
|
| 273 | return self
|
| 274 |
|
| 275 | def __exit__(self, type, value, traceback):
|
| 276 | # type: (Any, Any, Any) -> bool
|
| 277 | return False # Allows a traceback to occur
|
| 278 |
|
| 279 | def __call__(self, *cases):
|
| 280 | # type: (*Any) -> bool
|
| 281 | return self.value in cases
|
| 282 |
|
| 283 |
|
| 284 | class str_switch(object):
|
| 285 | """Translates to fast dispatch on string length, then memcmp()."""
|
| 286 |
|
| 287 | def __init__(self, value):
|
| 288 | # type: (str) -> None
|
| 289 | self.value = value
|
| 290 |
|
| 291 | def __enter__(self):
|
| 292 | # type: () -> switch
|
| 293 | return self
|
| 294 |
|
| 295 | def __exit__(self, type, value, traceback):
|
| 296 | # type: (Any, Any, Any) -> bool
|
| 297 | return False # Allows a traceback to occur
|
| 298 |
|
| 299 | def __call__(self, *cases):
|
| 300 | # type: (*Any) -> bool
|
| 301 | return self.value in cases
|
| 302 |
|
| 303 |
|
| 304 | class tagswitch(object):
|
| 305 | """Translates to C switch(node->tag())"""
|
| 306 |
|
| 307 | def __init__(self, node):
|
| 308 | # type: (Any) -> None
|
| 309 | self.tag = node.tag()
|
| 310 |
|
| 311 | def __enter__(self):
|
| 312 | # type: () -> tagswitch
|
| 313 | return self
|
| 314 |
|
| 315 | def __exit__(self, type, value, traceback):
|
| 316 | # type: (Any, Any, Any) -> bool
|
| 317 | return False # Allows a traceback to occur
|
| 318 |
|
| 319 | def __call__(self, *cases):
|
| 320 | # type: (*Any) -> bool
|
| 321 | return self.tag in cases
|
| 322 |
|
| 323 |
|
| 324 | if TYPE_CHECKING:
|
| 325 | # Doesn't work
|
| 326 | T = TypeVar('T')
|
| 327 |
|
| 328 | class StackArray(Generic[T]):
|
| 329 |
|
| 330 | def __init__(self):
|
| 331 | self.items = [] # type: List[T]
|
| 332 |
|
| 333 | def append(self, item):
|
| 334 | # type: (T) -> None
|
| 335 | self.items.append(item)
|
| 336 |
|
| 337 | def pop(self):
|
| 338 | # type: () -> T
|
| 339 | return self.items.pop()
|
| 340 |
|
| 341 | # Doesn't work, this is only for primitive types
|
| 342 | #StackArray = NewType('StackArray', list)
|
| 343 |
|
| 344 |
|
| 345 | def MakeStackArray(item_type):
|
| 346 | # type: (TypeVar) -> StackArray[item_type]
|
| 347 | """
|
| 348 | Convenience "constructor" used like this:
|
| 349 |
|
| 350 | myarray = MakeStackArray(int)
|
| 351 |
|
| 352 | The idiom could also be
|
| 353 |
|
| 354 | myarray = cast('StackArray[int]', [])
|
| 355 |
|
| 356 | But that's uglier.
|
| 357 | """
|
| 358 | return cast('StackArray[item_type]', [])
|
| 359 |
|
| 360 |
|
| 361 | if TYPE_CHECKING:
|
| 362 | K = TypeVar('K')
|
| 363 | V = TypeVar('V')
|
| 364 |
|
| 365 |
|
| 366 | def iteritems(d):
|
| 367 | # type: (Dict[K, V]) -> Iterator[Tuple[K, V]]
|
| 368 | """Make translation a bit easier."""
|
| 369 | return d.iteritems()
|
| 370 |
|
| 371 |
|
| 372 | def split_once(s, delim):
|
| 373 | # type: (str, str) -> Tuple[str, Optional[str]]
|
| 374 | """Easier to call than split(s, 1) because of tuple unpacking."""
|
| 375 |
|
| 376 | parts = s.split(delim, 1)
|
| 377 | if len(parts) == 1:
|
| 378 | no_str = None # type: Optional[str]
|
| 379 | return s, no_str
|
| 380 | else:
|
| 381 | return parts[0], parts[1]
|
| 382 |
|
| 383 |
|
| 384 | def hex_lower(i):
|
| 385 | # type: (int) -> str
|
| 386 | return '%x' % i
|
| 387 |
|
| 388 |
|
| 389 | def dict_erase(d, key):
|
| 390 | # type: (Dict[Any, Any], Any) -> None
|
| 391 | """
|
| 392 | Ensure that a key isn't in the Dict d. This makes C++ translation easier.
|
| 393 | """
|
| 394 | try:
|
| 395 | del d[key]
|
| 396 | except KeyError:
|
| 397 | pass
|
| 398 |
|
| 399 |
|
| 400 | def str_cmp(s1, s2):
|
| 401 | # type: (str, str) -> int
|
| 402 | if s1 == s2:
|
| 403 | return 0
|
| 404 | if s1 < s2:
|
| 405 | return -1
|
| 406 | else:
|
| 407 | return 1
|
| 408 |
|
| 409 |
|
| 410 | class UniqueObjects(object):
|
| 411 | """A set of objects identified by their address in memory
|
| 412 |
|
| 413 | Python's id(obj) returns the address of any object. But we don't simply
|
| 414 | implement it, because it requires a uint64_t on 64-bit systems, while mycpp
|
| 415 | only supports 'int'.
|
| 416 |
|
| 417 | So we have a whole class.
|
| 418 |
|
| 419 | Should be used for:
|
| 420 |
|
| 421 | - Cycle detection when pretty printing, as Python's repr() does
|
| 422 | - See CPython's Objects/object.c PyObject_Repr()
|
| 423 | /* These methods are used to control infinite recursion in repr, str, print,
|
| 424 | etc. Container objects that may recursively contain themselves,
|
| 425 | e.g. builtin dictionaries and lists, should use Py_ReprEnter() and
|
| 426 | Py_ReprLeave() to avoid infinite recursion.
|
| 427 | */
|
| 428 | - e.g. dictobject.c dict_repr() calls Py_ReprEnter() to print {...}
|
| 429 | - In Python 2.7 a GLOBAL VAR is used
|
| 430 |
|
| 431 | - It also checks for STACK OVERFLOW
|
| 432 |
|
| 433 | - Packle serialization
|
| 434 | """
|
| 435 |
|
| 436 | def __init__(self):
|
| 437 | # 64-bit id() -> small integer ID
|
| 438 | self.addresses = {} # type: Dict[int, int]
|
| 439 |
|
| 440 | def Contains(self, obj):
|
| 441 | # type: (Any) -> bool
|
| 442 | """ Convenience? """
|
| 443 | return self.Get(obj) != -1
|
| 444 |
|
| 445 | def MaybeAdd(self, obj):
|
| 446 | # type: (Any) -> None
|
| 447 | """ Convenience? """
|
| 448 |
|
| 449 | # def AddNewObject(self, obj):
|
| 450 | def Add(self, obj):
|
| 451 | # type: (Any) -> None
|
| 452 | """
|
| 453 | Assert it isn't already there, and assign a new ID!
|
| 454 |
|
| 455 | # Lib/pickle does:
|
| 456 |
|
| 457 | self.memo[id(obj)] = memo_len, obj
|
| 458 |
|
| 459 | I guess that's the object ID and a void*
|
| 460 |
|
| 461 | Then it does:
|
| 462 |
|
| 463 | x = self.memo.get(id(obj))
|
| 464 |
|
| 465 | and
|
| 466 |
|
| 467 | # If the object is already in the memo, this means it is
|
| 468 | # recursive. In this case, throw away everything we put on the
|
| 469 | # stack, and fetch the object back from the memo.
|
| 470 | if id(obj) in self.memo:
|
| 471 | write(POP + self.get(self.memo[id(obj)][0]))
|
| 472 |
|
| 473 | BUT It only uses the numeric ID!
|
| 474 | """
|
| 475 | addr = id(obj)
|
| 476 | assert addr not in self.addresses
|
| 477 | self.addresses[addr] = len(self.addresses)
|
| 478 |
|
| 479 | def Get(self, obj):
|
| 480 | # type: (Any) -> int
|
| 481 | """
|
| 482 | Returns unique ID assigned
|
| 483 |
|
| 484 | Returns -1 if it doesn't exist?
|
| 485 | """
|
| 486 | addr = id(obj)
|
| 487 | return self.addresses.get(addr, -1)
|
| 488 |
|
| 489 | # Note: self.memo.clear() doesn't appear to be used
|
| 490 |
|
| 491 |
|
| 492 | def probe(provider, name, *args):
|
| 493 | # type: (str, str, Any) -> None
|
| 494 | """Create a probe for use with profilers like linux perf and ebpf or dtrace."""
|
| 495 | # Noop. Just a marker for mycpp to emit a DTRACE_PROBE()
|
| 496 | return
|
| 497 |
|
| 498 |
|
| 499 | class File:
|
| 500 | """
|
| 501 | TODO: This should define a read/write interface, and then LineReader() and
|
| 502 | Writer() can possibly inherit it, with runtime assertions
|
| 503 |
|
| 504 | Then we allow downcasting from File -> LineReader, like we currently do in
|
| 505 | C++ in gc_mylib.h.
|
| 506 |
|
| 507 | Inheritance can't express the structural Reader/Writer pattern of Go, which
|
| 508 | would be better. I suppose we could use File* everywhere, but having
|
| 509 | fine-grained types is nicer. And there will be very few casts.
|
| 510 | """
|
| 511 | pass
|
| 512 |
|
| 513 |
|
| 514 | def stat(filename):
|
| 515 | # type: (str) -> StatResult
|
| 516 | try:
|
| 517 | stat_result = posix.stat(filename)
|
| 518 | except posix.error:
|
| 519 | return None
|
| 520 |
|
| 521 | return StatResult(stat_result)
|
| 522 |
|
| 523 |
|
| 524 | class StatResult:
|
| 525 | def __init__(self, stat_result):
|
| 526 | # type: (posix.stat_result) -> None
|
| 527 | self.stat_result = stat_result
|
| 528 |
|
| 529 | def isreg(self):
|
| 530 | # type: () -> bool
|
| 531 | return S_ISREG(self.stat_result.st_mode)
|