Jeffrey Yasskin | d7b0033 | 2008-01-15 07:46:24 +0000 | [diff] [blame] | 1 | # Originally contributed by Sjoerd Mullender. |
| 2 | # Significantly modified by Jeffrey Yasskin <jyasskin at gmail.com>. |
| 3 | |
| 4 | """Rational, infinite-precision, real numbers.""" |
| 5 | |
| 6 | from __future__ import division |
| 7 | import math |
| 8 | import numbers |
| 9 | import operator |
Jeffrey Yasskin | 45169fb | 2008-01-19 09:56:06 +0000 | [diff] [blame] | 10 | import re |
Jeffrey Yasskin | d7b0033 | 2008-01-15 07:46:24 +0000 | [diff] [blame] | 11 | |
| 12 | __all__ = ["Rational"] |
| 13 | |
| 14 | RationalAbc = numbers.Rational |
| 15 | |
| 16 | |
| 17 | def _gcd(a, b): |
| 18 | """Calculate the Greatest Common Divisor. |
| 19 | |
| 20 | Unless b==0, the result will have the same sign as b (so that when |
| 21 | b is divided by it, the result comes out positive). |
| 22 | """ |
| 23 | while b: |
| 24 | a, b = b, a%b |
| 25 | return a |
| 26 | |
| 27 | |
| 28 | def _binary_float_to_ratio(x): |
| 29 | """x -> (top, bot), a pair of ints s.t. x = top/bot. |
| 30 | |
| 31 | The conversion is done exactly, without rounding. |
| 32 | bot > 0 guaranteed. |
| 33 | Some form of binary fp is assumed. |
| 34 | Pass NaNs or infinities at your own risk. |
| 35 | |
| 36 | >>> _binary_float_to_ratio(10.0) |
| 37 | (10, 1) |
| 38 | >>> _binary_float_to_ratio(0.0) |
| 39 | (0, 1) |
| 40 | >>> _binary_float_to_ratio(-.25) |
| 41 | (-1, 4) |
| 42 | """ |
| 43 | |
| 44 | if x == 0: |
| 45 | return 0, 1 |
| 46 | f, e = math.frexp(x) |
| 47 | signbit = 1 |
| 48 | if f < 0: |
| 49 | f = -f |
| 50 | signbit = -1 |
| 51 | assert 0.5 <= f < 1.0 |
| 52 | # x = signbit * f * 2**e exactly |
| 53 | |
| 54 | # Suck up CHUNK bits at a time; 28 is enough so that we suck |
| 55 | # up all bits in 2 iterations for all known binary double- |
| 56 | # precision formats, and small enough to fit in an int. |
| 57 | CHUNK = 28 |
| 58 | top = 0 |
| 59 | # invariant: x = signbit * (top + f) * 2**e exactly |
| 60 | while f: |
| 61 | f = math.ldexp(f, CHUNK) |
| 62 | digit = trunc(f) |
| 63 | assert digit >> CHUNK == 0 |
| 64 | top = (top << CHUNK) | digit |
| 65 | f = f - digit |
| 66 | assert 0.0 <= f < 1.0 |
| 67 | e = e - CHUNK |
| 68 | assert top |
| 69 | |
| 70 | # Add in the sign bit. |
| 71 | top = signbit * top |
| 72 | |
| 73 | # now x = top * 2**e exactly; fold in 2**e |
| 74 | if e>0: |
| 75 | return (top * 2**e, 1) |
| 76 | else: |
| 77 | return (top, 2 ** -e) |
| 78 | |
| 79 | |
Jeffrey Yasskin | 45169fb | 2008-01-19 09:56:06 +0000 | [diff] [blame] | 80 | _RATIONAL_FORMAT = re.compile( |
| 81 | r'^\s*(?P<sign>[-+]?)(?P<num>\d+)(?:/(?P<denom>\d+))?\s*$') |
| 82 | |
| 83 | |
Jeffrey Yasskin | d7b0033 | 2008-01-15 07:46:24 +0000 | [diff] [blame] | 84 | class Rational(RationalAbc): |
| 85 | """This class implements rational numbers. |
| 86 | |
| 87 | Rational(8, 6) will produce a rational number equivalent to |
| 88 | 4/3. Both arguments must be Integral. The numerator defaults to 0 |
| 89 | and the denominator defaults to 1 so that Rational(3) == 3 and |
| 90 | Rational() == 0. |
| 91 | |
Jeffrey Yasskin | 45169fb | 2008-01-19 09:56:06 +0000 | [diff] [blame] | 92 | Rationals can also be constructed from strings of the form |
| 93 | '[-+]?[0-9]+(/[0-9]+)?', optionally surrounded by spaces. |
| 94 | |
Jeffrey Yasskin | d7b0033 | 2008-01-15 07:46:24 +0000 | [diff] [blame] | 95 | """ |
| 96 | |
Raymond Hettinger | 7a6eacd | 2008-01-24 18:05:54 +0000 | [diff] [blame] | 97 | __slots__ = ('numerator', 'denominator') |
Jeffrey Yasskin | d7b0033 | 2008-01-15 07:46:24 +0000 | [diff] [blame] | 98 | |
Jeffrey Yasskin | 45169fb | 2008-01-19 09:56:06 +0000 | [diff] [blame] | 99 | # We're immutable, so use __new__ not __init__ |
| 100 | def __new__(cls, numerator=0, denominator=1): |
| 101 | """Constructs a Rational. |
| 102 | |
| 103 | Takes a string, another Rational, or a numerator/denominator pair. |
| 104 | |
| 105 | """ |
| 106 | self = super(Rational, cls).__new__(cls) |
| 107 | |
| 108 | if denominator == 1: |
| 109 | if isinstance(numerator, basestring): |
| 110 | # Handle construction from strings. |
| 111 | input = numerator |
| 112 | m = _RATIONAL_FORMAT.match(input) |
| 113 | if m is None: |
| 114 | raise ValueError('Invalid literal for Rational: ' + input) |
| 115 | numerator = int(m.group('num')) |
| 116 | # Default denominator to 1. That's the only optional group. |
| 117 | denominator = int(m.group('denom') or 1) |
| 118 | if m.group('sign') == '-': |
| 119 | numerator = -numerator |
| 120 | |
| 121 | elif (not isinstance(numerator, numbers.Integral) and |
| 122 | isinstance(numerator, RationalAbc)): |
| 123 | # Handle copies from other rationals. |
| 124 | other_rational = numerator |
| 125 | numerator = other_rational.numerator |
| 126 | denominator = other_rational.denominator |
Jeffrey Yasskin | d7b0033 | 2008-01-15 07:46:24 +0000 | [diff] [blame] | 127 | |
| 128 | if (not isinstance(numerator, numbers.Integral) or |
| 129 | not isinstance(denominator, numbers.Integral)): |
| 130 | raise TypeError("Rational(%(numerator)s, %(denominator)s):" |
| 131 | " Both arguments must be integral." % locals()) |
| 132 | |
| 133 | if denominator == 0: |
| 134 | raise ZeroDivisionError('Rational(%s, 0)' % numerator) |
| 135 | |
| 136 | g = _gcd(numerator, denominator) |
Raymond Hettinger | 7a6eacd | 2008-01-24 18:05:54 +0000 | [diff] [blame] | 137 | self.numerator = int(numerator // g) |
| 138 | self.denominator = int(denominator // g) |
Jeffrey Yasskin | 45169fb | 2008-01-19 09:56:06 +0000 | [diff] [blame] | 139 | return self |
Jeffrey Yasskin | d7b0033 | 2008-01-15 07:46:24 +0000 | [diff] [blame] | 140 | |
| 141 | @classmethod |
| 142 | def from_float(cls, f): |
Jeffrey Yasskin | 45169fb | 2008-01-19 09:56:06 +0000 | [diff] [blame] | 143 | """Converts a finite float to a rational number, exactly. |
| 144 | |
| 145 | Beware that Rational.from_float(0.3) != Rational(3, 10). |
| 146 | |
| 147 | """ |
Jeffrey Yasskin | d7b0033 | 2008-01-15 07:46:24 +0000 | [diff] [blame] | 148 | if not isinstance(f, float): |
| 149 | raise TypeError("%s.from_float() only takes floats, not %r (%s)" % |
| 150 | (cls.__name__, f, type(f).__name__)) |
| 151 | if math.isnan(f) or math.isinf(f): |
| 152 | raise TypeError("Cannot convert %r to %s." % (f, cls.__name__)) |
| 153 | return cls(*_binary_float_to_ratio(f)) |
| 154 | |
Jeffrey Yasskin | 45169fb | 2008-01-19 09:56:06 +0000 | [diff] [blame] | 155 | @classmethod |
| 156 | def from_decimal(cls, dec): |
| 157 | """Converts a finite Decimal instance to a rational number, exactly.""" |
| 158 | from decimal import Decimal |
| 159 | if not isinstance(dec, Decimal): |
| 160 | raise TypeError( |
| 161 | "%s.from_decimal() only takes Decimals, not %r (%s)" % |
| 162 | (cls.__name__, dec, type(dec).__name__)) |
| 163 | if not dec.is_finite(): |
| 164 | # Catches infinities and nans. |
| 165 | raise TypeError("Cannot convert %s to %s." % (dec, cls.__name__)) |
| 166 | sign, digits, exp = dec.as_tuple() |
| 167 | digits = int(''.join(map(str, digits))) |
| 168 | if sign: |
| 169 | digits = -digits |
| 170 | if exp >= 0: |
| 171 | return cls(digits * 10 ** exp) |
| 172 | else: |
| 173 | return cls(digits, 10 ** -exp) |
| 174 | |
Raymond Hettinger | cf10926 | 2008-01-24 00:54:21 +0000 | [diff] [blame] | 175 | @classmethod |
| 176 | def from_continued_fraction(cls, seq): |
| 177 | 'Build a Rational from a continued fraction expessed as a sequence' |
| 178 | n, d = 1, 0 |
| 179 | for e in reversed(seq): |
| 180 | n, d = d, n |
| 181 | n += e * d |
Raymond Hettinger | f336c8b | 2008-01-24 02:05:06 +0000 | [diff] [blame] | 182 | return cls(n, d) if seq else cls(0) |
Raymond Hettinger | cf10926 | 2008-01-24 00:54:21 +0000 | [diff] [blame] | 183 | |
| 184 | def as_continued_fraction(self): |
| 185 | 'Return continued fraction expressed as a list' |
| 186 | n = self.numerator |
| 187 | d = self.denominator |
| 188 | cf = [] |
| 189 | while d: |
| 190 | e = int(n // d) |
| 191 | cf.append(e) |
| 192 | n -= e * d |
| 193 | n, d = d, n |
| 194 | return cf |
| 195 | |
| 196 | @classmethod |
| 197 | def approximate_from_float(cls, f, max_denominator): |
| 198 | 'Best rational approximation to f with a denominator <= max_denominator' |
| 199 | # XXX First cut at algorithm |
| 200 | # Still needs rounding rules as specified at |
| 201 | # http://en.wikipedia.org/wiki/Continued_fraction |
| 202 | cf = cls.from_float(f).as_continued_fraction() |
Raymond Hettinger | eb46190 | 2008-01-24 02:00:25 +0000 | [diff] [blame] | 203 | result = Rational(0) |
Raymond Hettinger | cf10926 | 2008-01-24 00:54:21 +0000 | [diff] [blame] | 204 | for i in range(1, len(cf)): |
| 205 | new = cls.from_continued_fraction(cf[:i]) |
| 206 | if new.denominator > max_denominator: |
| 207 | break |
| 208 | result = new |
| 209 | return result |
| 210 | |
Jeffrey Yasskin | d7b0033 | 2008-01-15 07:46:24 +0000 | [diff] [blame] | 211 | def __repr__(self): |
| 212 | """repr(self)""" |
Jeffrey Yasskin | 45169fb | 2008-01-19 09:56:06 +0000 | [diff] [blame] | 213 | return ('Rational(%r,%r)' % (self.numerator, self.denominator)) |
Jeffrey Yasskin | d7b0033 | 2008-01-15 07:46:24 +0000 | [diff] [blame] | 214 | |
| 215 | def __str__(self): |
| 216 | """str(self)""" |
| 217 | if self.denominator == 1: |
| 218 | return str(self.numerator) |
| 219 | else: |
Jeffrey Yasskin | 45169fb | 2008-01-19 09:56:06 +0000 | [diff] [blame] | 220 | return '%s/%s' % (self.numerator, self.denominator) |
Jeffrey Yasskin | d7b0033 | 2008-01-15 07:46:24 +0000 | [diff] [blame] | 221 | |
| 222 | def _operator_fallbacks(monomorphic_operator, fallback_operator): |
| 223 | """Generates forward and reverse operators given a purely-rational |
| 224 | operator and a function from the operator module. |
| 225 | |
| 226 | Use this like: |
| 227 | __op__, __rop__ = _operator_fallbacks(just_rational_op, operator.op) |
| 228 | |
| 229 | """ |
| 230 | def forward(a, b): |
| 231 | if isinstance(b, RationalAbc): |
| 232 | # Includes ints. |
| 233 | return monomorphic_operator(a, b) |
| 234 | elif isinstance(b, float): |
| 235 | return fallback_operator(float(a), b) |
| 236 | elif isinstance(b, complex): |
| 237 | return fallback_operator(complex(a), b) |
| 238 | else: |
| 239 | return NotImplemented |
| 240 | forward.__name__ = '__' + fallback_operator.__name__ + '__' |
| 241 | forward.__doc__ = monomorphic_operator.__doc__ |
| 242 | |
| 243 | def reverse(b, a): |
| 244 | if isinstance(a, RationalAbc): |
| 245 | # Includes ints. |
| 246 | return monomorphic_operator(a, b) |
| 247 | elif isinstance(a, numbers.Real): |
| 248 | return fallback_operator(float(a), float(b)) |
| 249 | elif isinstance(a, numbers.Complex): |
| 250 | return fallback_operator(complex(a), complex(b)) |
| 251 | else: |
| 252 | return NotImplemented |
| 253 | reverse.__name__ = '__r' + fallback_operator.__name__ + '__' |
| 254 | reverse.__doc__ = monomorphic_operator.__doc__ |
| 255 | |
| 256 | return forward, reverse |
| 257 | |
| 258 | def _add(a, b): |
| 259 | """a + b""" |
| 260 | return Rational(a.numerator * b.denominator + |
| 261 | b.numerator * a.denominator, |
| 262 | a.denominator * b.denominator) |
| 263 | |
| 264 | __add__, __radd__ = _operator_fallbacks(_add, operator.add) |
| 265 | |
| 266 | def _sub(a, b): |
| 267 | """a - b""" |
| 268 | return Rational(a.numerator * b.denominator - |
| 269 | b.numerator * a.denominator, |
| 270 | a.denominator * b.denominator) |
| 271 | |
| 272 | __sub__, __rsub__ = _operator_fallbacks(_sub, operator.sub) |
| 273 | |
| 274 | def _mul(a, b): |
| 275 | """a * b""" |
| 276 | return Rational(a.numerator * b.numerator, a.denominator * b.denominator) |
| 277 | |
| 278 | __mul__, __rmul__ = _operator_fallbacks(_mul, operator.mul) |
| 279 | |
| 280 | def _div(a, b): |
| 281 | """a / b""" |
| 282 | return Rational(a.numerator * b.denominator, |
| 283 | a.denominator * b.numerator) |
| 284 | |
| 285 | __truediv__, __rtruediv__ = _operator_fallbacks(_div, operator.truediv) |
| 286 | __div__, __rdiv__ = _operator_fallbacks(_div, operator.div) |
| 287 | |
Raymond Hettinger | 909e334 | 2008-01-24 23:50:26 +0000 | [diff] [blame^] | 288 | def __floordiv__(a, b): |
| 289 | """a // b""" |
| 290 | # Will be math.floor(a / b) in 3.0. |
Jeffrey Yasskin | d7b0033 | 2008-01-15 07:46:24 +0000 | [diff] [blame] | 291 | div = a / b |
| 292 | if isinstance(div, RationalAbc): |
| 293 | # trunc(math.floor(div)) doesn't work if the rational is |
| 294 | # more precise than a float because the intermediate |
| 295 | # rounding may cross an integer boundary. |
| 296 | return div.numerator // div.denominator |
| 297 | else: |
| 298 | return math.floor(div) |
| 299 | |
Jeffrey Yasskin | d7b0033 | 2008-01-15 07:46:24 +0000 | [diff] [blame] | 300 | def __rfloordiv__(b, a): |
| 301 | """a // b""" |
| 302 | # Will be math.floor(a / b) in 3.0. |
Raymond Hettinger | 909e334 | 2008-01-24 23:50:26 +0000 | [diff] [blame^] | 303 | div = a / b |
| 304 | if isinstance(div, RationalAbc): |
| 305 | # trunc(math.floor(div)) doesn't work if the rational is |
| 306 | # more precise than a float because the intermediate |
| 307 | # rounding may cross an integer boundary. |
| 308 | return div.numerator // div.denominator |
| 309 | else: |
| 310 | return math.floor(div) |
Jeffrey Yasskin | d7b0033 | 2008-01-15 07:46:24 +0000 | [diff] [blame] | 311 | |
| 312 | def __mod__(a, b): |
| 313 | """a % b""" |
Raymond Hettinger | 909e334 | 2008-01-24 23:50:26 +0000 | [diff] [blame^] | 314 | div = a // b |
| 315 | return a - b * div |
Jeffrey Yasskin | d7b0033 | 2008-01-15 07:46:24 +0000 | [diff] [blame] | 316 | |
| 317 | def __rmod__(b, a): |
| 318 | """a % b""" |
Raymond Hettinger | 909e334 | 2008-01-24 23:50:26 +0000 | [diff] [blame^] | 319 | div = a // b |
| 320 | return a - b * div |
Jeffrey Yasskin | d7b0033 | 2008-01-15 07:46:24 +0000 | [diff] [blame] | 321 | |
| 322 | def __pow__(a, b): |
| 323 | """a ** b |
| 324 | |
| 325 | If b is not an integer, the result will be a float or complex |
| 326 | since roots are generally irrational. If b is an integer, the |
| 327 | result will be rational. |
| 328 | |
| 329 | """ |
| 330 | if isinstance(b, RationalAbc): |
| 331 | if b.denominator == 1: |
| 332 | power = b.numerator |
| 333 | if power >= 0: |
| 334 | return Rational(a.numerator ** power, |
| 335 | a.denominator ** power) |
| 336 | else: |
| 337 | return Rational(a.denominator ** -power, |
| 338 | a.numerator ** -power) |
| 339 | else: |
| 340 | # A fractional power will generally produce an |
| 341 | # irrational number. |
| 342 | return float(a) ** float(b) |
| 343 | else: |
| 344 | return float(a) ** b |
| 345 | |
| 346 | def __rpow__(b, a): |
| 347 | """a ** b""" |
| 348 | if b.denominator == 1 and b.numerator >= 0: |
| 349 | # If a is an int, keep it that way if possible. |
| 350 | return a ** b.numerator |
| 351 | |
| 352 | if isinstance(a, RationalAbc): |
| 353 | return Rational(a.numerator, a.denominator) ** b |
| 354 | |
| 355 | if b.denominator == 1: |
| 356 | return a ** b.numerator |
| 357 | |
| 358 | return a ** float(b) |
| 359 | |
| 360 | def __pos__(a): |
| 361 | """+a: Coerces a subclass instance to Rational""" |
| 362 | return Rational(a.numerator, a.denominator) |
| 363 | |
| 364 | def __neg__(a): |
| 365 | """-a""" |
| 366 | return Rational(-a.numerator, a.denominator) |
| 367 | |
| 368 | def __abs__(a): |
| 369 | """abs(a)""" |
| 370 | return Rational(abs(a.numerator), a.denominator) |
| 371 | |
| 372 | def __trunc__(a): |
| 373 | """trunc(a)""" |
| 374 | if a.numerator < 0: |
| 375 | return -(-a.numerator // a.denominator) |
| 376 | else: |
| 377 | return a.numerator // a.denominator |
| 378 | |
Raymond Hettinger | 5b0e27e | 2008-01-24 19:30:19 +0000 | [diff] [blame] | 379 | __int__ = __trunc__ |
| 380 | |
Jeffrey Yasskin | d7b0033 | 2008-01-15 07:46:24 +0000 | [diff] [blame] | 381 | def __floor__(a): |
| 382 | """Will be math.floor(a) in 3.0.""" |
| 383 | return a.numerator // a.denominator |
| 384 | |
| 385 | def __ceil__(a): |
| 386 | """Will be math.ceil(a) in 3.0.""" |
| 387 | # The negations cleverly convince floordiv to return the ceiling. |
| 388 | return -(-a.numerator // a.denominator) |
| 389 | |
| 390 | def __round__(self, ndigits=None): |
| 391 | """Will be round(self, ndigits) in 3.0. |
| 392 | |
| 393 | Rounds half toward even. |
| 394 | """ |
| 395 | if ndigits is None: |
| 396 | floor, remainder = divmod(self.numerator, self.denominator) |
| 397 | if remainder * 2 < self.denominator: |
| 398 | return floor |
| 399 | elif remainder * 2 > self.denominator: |
| 400 | return floor + 1 |
| 401 | # Deal with the half case: |
| 402 | elif floor % 2 == 0: |
| 403 | return floor |
| 404 | else: |
| 405 | return floor + 1 |
| 406 | shift = 10**abs(ndigits) |
| 407 | # See _operator_fallbacks.forward to check that the results of |
| 408 | # these operations will always be Rational and therefore have |
| 409 | # __round__(). |
| 410 | if ndigits > 0: |
| 411 | return Rational((self * shift).__round__(), shift) |
| 412 | else: |
| 413 | return Rational((self / shift).__round__() * shift) |
| 414 | |
| 415 | def __hash__(self): |
| 416 | """hash(self) |
| 417 | |
| 418 | Tricky because values that are exactly representable as a |
| 419 | float must have the same hash as that float. |
| 420 | |
| 421 | """ |
| 422 | if self.denominator == 1: |
| 423 | # Get integers right. |
| 424 | return hash(self.numerator) |
| 425 | # Expensive check, but definitely correct. |
| 426 | if self == float(self): |
| 427 | return hash(float(self)) |
| 428 | else: |
| 429 | # Use tuple's hash to avoid a high collision rate on |
| 430 | # simple fractions. |
| 431 | return hash((self.numerator, self.denominator)) |
| 432 | |
| 433 | def __eq__(a, b): |
| 434 | """a == b""" |
| 435 | if isinstance(b, RationalAbc): |
| 436 | return (a.numerator == b.numerator and |
| 437 | a.denominator == b.denominator) |
| 438 | if isinstance(b, numbers.Complex) and b.imag == 0: |
| 439 | b = b.real |
| 440 | if isinstance(b, float): |
| 441 | return a == a.from_float(b) |
| 442 | else: |
| 443 | # XXX: If b.__eq__ is implemented like this method, it may |
| 444 | # give the wrong answer after float(a) changes a's |
| 445 | # value. Better ways of doing this are welcome. |
| 446 | return float(a) == b |
| 447 | |
| 448 | def _subtractAndCompareToZero(a, b, op): |
| 449 | """Helper function for comparison operators. |
| 450 | |
| 451 | Subtracts b from a, exactly if possible, and compares the |
| 452 | result with 0 using op, in such a way that the comparison |
| 453 | won't recurse. If the difference raises a TypeError, returns |
| 454 | NotImplemented instead. |
| 455 | |
| 456 | """ |
| 457 | if isinstance(b, numbers.Complex) and b.imag == 0: |
| 458 | b = b.real |
| 459 | if isinstance(b, float): |
| 460 | b = a.from_float(b) |
| 461 | try: |
| 462 | # XXX: If b <: Real but not <: RationalAbc, this is likely |
| 463 | # to fall back to a float. If the actual values differ by |
| 464 | # less than MIN_FLOAT, this could falsely call them equal, |
| 465 | # which would make <= inconsistent with ==. Better ways of |
| 466 | # doing this are welcome. |
| 467 | diff = a - b |
| 468 | except TypeError: |
| 469 | return NotImplemented |
| 470 | if isinstance(diff, RationalAbc): |
| 471 | return op(diff.numerator, 0) |
| 472 | return op(diff, 0) |
| 473 | |
| 474 | def __lt__(a, b): |
| 475 | """a < b""" |
| 476 | return a._subtractAndCompareToZero(b, operator.lt) |
| 477 | |
| 478 | def __gt__(a, b): |
| 479 | """a > b""" |
| 480 | return a._subtractAndCompareToZero(b, operator.gt) |
| 481 | |
| 482 | def __le__(a, b): |
| 483 | """a <= b""" |
| 484 | return a._subtractAndCompareToZero(b, operator.le) |
| 485 | |
| 486 | def __ge__(a, b): |
| 487 | """a >= b""" |
| 488 | return a._subtractAndCompareToZero(b, operator.ge) |
| 489 | |
| 490 | def __nonzero__(a): |
| 491 | """a != 0""" |
| 492 | return a.numerator != 0 |