NotificationsYou must be signed in to change notification settings
Fork11.7k
Star30.8k

Commit2b216b1

authored

BUG:numpy.random.* class runtime signatures (#30164)

This fixes `inspect.signature` for- `np.random.BitGenerator`- `np.random.Generator`- `np.random.MT19937`- `np.random.PCG64`- `np.random.PCG64DXSM`- `np.random.Philox`- `np.random.RandomState`- `np.random.SFC64`- `np.random.SeedSequence`- `np.random.bit_generator.SeedlessSeedSequence`This also fixes a typo in `bit_generator.pxd` that accidentally defined an empty unused class `np.random.bit_generator.SeedlessSequence`.Related to#30104,#30114,#30121,#30124,#30126,#30137,#30138,#30140,#30143,#30146,#30147, and#30155

1 parent125b5ce commit2b216b1Copy full SHA for 2b216b1

File tree

9 files changed

+129

-100

lines changed

numpy/random

9 files changed

+129

-100

lines changed

`‎numpy/random/_generator.pyx‎`

Lines changed: 59 additions & 59 deletions

Original file line number	Diff line number	Diff line change
`@@ -140,8 +140,8 @@ cdef bint _check_bit_generator(object bitgen):`
`140`	`140`
`141`	`141`
`142`	`142`	`cdefclassGenerator:`
`143`		`-"""`
`144`		`- Generator(bit_generator)`
	`143`	+# the first line is used to populate `__text_signature__`
	`144`	`+"""Generator(bit_generator)\n--`
`145`	`145`
`146`	`146`	`Container for the BitGenerators.`
`147`	`147`
`@@ -396,20 +396,20 @@ cdef class Generator:`
`396`	`396`	`Drawn samples from the parameterized beta distribution.`
`397`	`397`
`398`	`398`	`Examples`
`399`		`- --------`
`400`		- The beta distribution has mean a/(a+b). If ``a == b`` and both
	`399`	`+ --------`
	`400`	+ The beta distribution has mean a/(a+b). If ``a == b`` and both
`401`	`401`	`are > 1, the distribution is symmetric with mean 0.5.`
`402`	`402`
`403`	`403`	`>>> rng = np.random.default_rng()`
`404`	`404`	`>>> a, b, size = 2.0, 2.0, 10000`
`405`	`405`	`>>> sample = rng.beta(a=a, b=b, size=size)`
`406`	`406`	`>>> np.mean(sample)`
`407`	`407`	`0.5047328775385895 # may vary`
`408`		`-`
	`408`	`+`
`409`	`409`	`Otherwise the distribution is skewed left or right according to`
`410`	`410`	whether ``a`` or ``b`` is greater. The distribution is mirror
`411`	`411`	`symmetric. See for example:`
`412`		`-`
	`412`	`+`
`413`	`413`	`>>> a, b, size = 2, 7, 10000`
`414`	`414`	`>>> sample_left = rng.beta(a=a, b=b, size=size)`
`415`	`415`	`>>> sample_right = rng.beta(a=b, b=a, size=size)`
`@@ -422,12 +422,12 @@ cdef class Generator:`
`422`	`422`	`-0.0003163943736596009 # may vary`
`423`	`423`
`424`	`424`	`Display the histogram of the two samples:`
`425`		`-`
	`425`	`+`
`426`	`426`	`>>> import matplotlib.pyplot as plt`
`427`		`- >>> plt.hist([sample_left, sample_right],`
	`427`	`+ >>> plt.hist([sample_left, sample_right],`
`428`	`428`	`... 50, density=True, histtype='bar')`
`429`	`429`	`>>> plt.show()`
`430`		`-`
	`430`	`+`
`431`	`431`	`References`
`432`	`432`	`----------`
`433`	`433`	`.. [1] Wikipedia, "Beta distribution",`
`@@ -477,17 +477,17 @@ cdef class Generator:`
`477`	`477`
`478`	`478`	`Examples`
`479`	`479`	`--------`
`480`		`- Assume a company has 10000 customer support agents and the time`
`481`		`- between customer calls is exponentially distributed and that the`
	`480`	`+ Assume a company has 10000 customer support agents and the time`
	`481`	`+ between customer calls is exponentially distributed and that the`
`482`	`482`	`average time between customer calls is 4 minutes.`
`483`	`483`
`484`	`484`	`>>> scale, size = 4, 10000`
`485`	`485`	`>>> rng = np.random.default_rng()`
`486`	`486`	`>>> time_between_calls = rng.exponential(scale=scale, size=size)`
`487`	`487`
`488`		`- What is the probability that a customer will call in the next`
`489`		`- 4 to 5 minutes?`
`490`		`-`
	`488`	`+ What is the probability that a customer will call in the next`
	`489`	`+ 4 to 5 minutes?`
	`490`	`+`
`491`	`491`	`>>> x = ((time_between_calls < 5).sum())/size`
`492`	`492`	`>>> y = ((time_between_calls < 4).sum())/size`
`493`	`493`	`>>> x - y`
`@@ -718,10 +718,10 @@ cdef class Generator:`
`718`	`718`
`719`	`719`	`Notes`
`720`	`720`	`-----`
`721`		`- This function generates random bytes from a discrete uniform`
`722`		`- distribution. The generated bytes are independent from the CPU's`
	`721`	`+ This function generates random bytes from a discrete uniform`
	`722`	`+ distribution. The generated bytes are independent from the CPU's`
`723`	`723`	`native endianness.`
`724`		`-`
	`724`	`+`
`725`	`725`	`Examples`
`726`	`726`	`--------`
`727`	`727`	`>>> rng = np.random.default_rng()`
`@@ -1024,9 +1024,9 @@ cdef class Generator:`
`1024`	`1024`	`greater than or equal to low. The default value is 0.`
`1025`	`1025`	`high : float or array_like of floats`
`1026`	`1026`	`Upper boundary of the output interval. All values generated will be`
`1027`		`- less than high. The high limit may be included in the returned array of`
`1028`		`- floats due to floating-point rounding in the equation`
`1029`		- ``low + (high-low) * random_sample()``. high - low must be
	`1027`	`+ less than high. The high limit may be included in the returned array of`
	`1028`	`+ floats due to floating-point rounding in the equation`
	`1029`	+ ``low + (high-low) * random_sample()``. high - low must be
`1030`	`1030`	`non-negative. The default value is 1.0.`
`1031`	`1031`	`size : int or tuple of ints, optional`
`1032`	`1032`	Output shape. If the given shape is, e.g., ``(m, n, k)``, then
`@@ -1541,10 +1541,10 @@ cdef class Generator:`
`1541`	`1541`	`So there is about a 1% chance that the F statistic will exceed 7.62,`
`1542`	`1542`	`the measured value is 36, so the null hypothesis is rejected at the 1%`
`1543`	`1543`	`level.`
`1544`		`-`
`1545`		- The corresponding probability density function for ``n = 20``
	`1544`	`+`
	`1545`	+ The corresponding probability density function for ``n = 20``
`1546`	`1546`	and ``m = 20`` is:
`1547`		`-`
	`1547`	`+`
`1548`	`1548`	`>>> import matplotlib.pyplot as plt`
`1549`	`1549`	`>>> from scipy import stats`
`1550`	`1550`	`>>> dfnum, dfden, size = 20, 20, 10000`
`@@ -1554,7 +1554,7 @@ cdef class Generator:`
`1554`	`1554`	`>>> plt.plot(x, stats.f.pdf(x, dfnum, dfden))`
`1555`	`1555`	`>>> plt.xlim([0, 5])`
`1556`	`1556`	`>>> plt.show()`
`1557`		`-`
	`1557`	`+`
`1558`	`1558`	`"""`
`1559`	`1559`	`returncont(&random_f,&self._bitgen,size,self.lock,2,`
`1560`	`1560`	`dfnum,'dfnum',CONS_POSITIVE,`
`@@ -1701,7 +1701,7 @@ cdef class Generator:`
`1701`	`1701`
`1702`	`1702`	`The distribution of a chi-square random variable`
`1703`	`1703`	`with 20 degrees of freedom looks as follows:`
`1704`		`-`
	`1704`	`+`
`1705`	`1705`	`>>> import matplotlib.pyplot as plt`
`1706`	`1706`	`>>> import scipy.stats as stats`
`1707`	`1707`	`>>> s = rng.chisquare(20, 10000)`
`@@ -1921,14 +1921,14 @@ cdef class Generator:`
`1921`	`1921`	`Does their energy intake deviate systematically from the recommended`
`1922`	`1922`	`value of 7725 kJ? Our null hypothesis will be the absence of deviation,`
`1923`	`1923`	`and the alternate hypothesis will be the presence of an effect that could be`
`1924`		`- either positive or negative, hence making our test 2-tailed.`
	`1924`	`+ either positive or negative, hence making our test 2-tailed.`
`1925`	`1925`
`1926`	`1926`	`Because we are estimating the mean and we have N=11 values in our sample,`
`1927`		`- we have N-1=10 degrees of freedom. We set our significance level to 95% and`
`1928`		`- compute the t statistic using the empirical mean and empirical standard`
`1929`		`- deviation of our intake. We use a ddof of 1 to base the computation of our`
	`1927`	`+ we have N-1=10 degrees of freedom. We set our significance level to 95% and`
	`1928`	`+ compute the t statistic using the empirical mean and empirical standard`
	`1929`	`+ deviation of our intake. We use a ddof of 1 to base the computation of our`
`1930`	`1930`	`empirical standard deviation on an unbiased estimate of the variance (note:`
`1931`		`- the final estimate is not unbiased due to the concave nature of the square`
	`1931`	`+ the final estimate is not unbiased due to the concave nature of the square`
`1932`	`1932`	`root).`
`1933`	`1933`
`1934`	`1934`	`>>> np.mean(intake)`
`@@ -1947,18 +1947,18 @@ cdef class Generator:`
`1947`	`1947`	`>>> s = rng.standard_t(10, size=1000000)`
`1948`	`1948`	`>>> h = plt.hist(s, bins=100, density=True)`
`1949`	`1949`
`1950`		`- Does our t statistic land in one of the two critical regions found at`
	`1950`	`+ Does our t statistic land in one of the two critical regions found at`
`1951`	`1951`	`both tails of the distribution?`
`1952`	`1952`
`1953`	`1953`	`>>> np.sum(np.abs(t) < np.abs(s)) / float(len(s))`
`1954`	`1954`	`0.018318 #random < 0.05, statistic is in critical region`
`1955`	`1955`
`1956`		`- The probability value for this 2-tailed test is about 1.83%, which is`
`1957`		`- lower than the 5% pre-determined significance threshold.`
	`1956`	`+ The probability value for this 2-tailed test is about 1.83%, which is`
	`1957`	`+ lower than the 5% pre-determined significance threshold.`
`1958`	`1958`
`1959`	`1959`	`Therefore, the probability of observing values as extreme as our intake`
`1960`		`- conditionally on the null hypothesis being true is too low, and we reject`
`1961`		`- the null hypothesis of no deviation.`
	`1960`	`+ conditionally on the null hypothesis being true is too low, and we reject`
	`1961`	`+ the null hypothesis of no deviation.`
`1962`	`1962`
`1963`	`1963`	`"""`
`1964`	`1964`	`returncont(&random_standard_t,&self._bitgen,size,self.lock,1,`
`@@ -3034,21 +3034,21 @@ cdef class Generator:`
`3034`	`3034`	`Draw samples from the distribution:`
`3035`	`3035`
`3036`	`3036`	`>>> rng = np.random.default_rng()`
`3037`		`- >>> n, p, size = 10, .5, 10000`
	`3037`	`+ >>> n, p, size = 10, .5, 10000`
`3038`	`3038`	`>>> s = rng.binomial(n, p, 10000)`
`3039`	`3039`
`3040`	`3040`	`Assume a company drills 9 wild-cat oil exploration wells, each with`
`3041`		- an estimated probability of success of ``p=0.1``. All nine wells fail.
	`3041`	+ an estimated probability of success of ``p=0.1``. All nine wells fail.
`3042`	`3042`	`What is the probability of that happening?`
`3043`	`3043`
`3044`		- Over ``size = 20,000`` trials the probability of this happening
	`3044`	+ Over ``size = 20,000`` trials the probability of this happening
`3045`	`3045`	`is on average:`
`3046`	`3046`
`3047`	`3047`	`>>> n, p, size = 9, 0.1, 20000`
`3048`	`3048`	`>>> np.sum(rng.binomial(n=n, p=p, size=size) == 0)/size`
`3049`	`3049`	`0.39015 # may vary`
`3050`	`3050`
`3051`		- The following can be used to visualize a sample with ``n=100``,
	`3051`	+ The following can be used to visualize a sample with ``n=100``,
`3052`	`3052`	``p=0.4`` and the corresponding probability density function:
`3053`	`3053`
`3054`	`3054`	`>>> import matplotlib.pyplot as plt`
`@@ -3167,10 +3167,10 @@ cdef class Generator:`
`3167`	`3167`	`appear before the third "1" is a negative binomial distribution.`
`3168`	`3168`
`3169`	`3169`	Because this method internally calls ``Generator.poisson`` with an
`3170`		`- intermediate random value, a ValueError is raised when the choice of`
	`3170`	`+ intermediate random value, a ValueError is raised when the choice of`
`3171`	`3171`	:math:`n` and :math:`p` would result in the mean + 10 sigma of the sampled
`3172`		`- intermediate distribution exceeding the max acceptable value of the`
`3173`		- ``Generator.poisson`` method. This happens when :math:`p` is too low
	`3172`	`+ intermediate distribution exceeding the max acceptable value of the`
	`3173`	+ ``Generator.poisson`` method. This happens when :math:`p` is too low
`3174`	`3174`	(a lot of failures happen for every success) and :math:`n` is too big (
`3175`	`3175`	`a lot of successes are allowed).`
`3176`	`3176`	Therefore, the :math:`n` and :math:`p` values must satisfy the constraint:
`@@ -3302,7 +3302,7 @@ cdef class Generator:`
`3302`	`3302`	`>>> s = rng.poisson(lam=lam, size=size)`
`3303`	`3303`
`3304`	`3304`	Verify the mean and variance, which should be approximately ``lam``:
`3305`		`-`
	`3305`	`+`
`3306`	`3306`	`>>> s.mean(), s.var()`
`3307`	`3307`	`(4.9917 5.1088311) # may vary`
`3308`	`3308`
`@@ -3456,7 +3456,7 @@ cdef class Generator:`
`3456`	`3456`
`3457`	`3457`	`Examples`
`3458`	`3458`	`--------`
`3459`		`- Draw 10,000 values from the geometric distribution, with the`
	`3459`	`+ Draw 10,000 values from the geometric distribution, with the`
`3460`	`3460`	probability of an individual success equal to ``p = 0.35``:
`3461`	`3461`
`3462`	`3462`	`>>> p, size = 0.35, 10000`
`@@ -3475,7 +3475,7 @@ cdef class Generator:`
`3475`	`3475`	`>>> plt.plot(bins, (1-p)*(bins-1)p)`
`3476`	`3476`	`>>> plt.xlim([0, 25])`
`3477`	`3477`	`>>> plt.show()`
`3478`		`-`
	`3478`	`+`
`3479`	`3479`	`"""`
`3480`	`3480`	`returndisc(&random_geometric,&self._bitgen,size,self.lock,1,0,`
`3481`	`3481`	`p,'p',CONS_BOUNDED_GT_0_1,`
`@@ -4646,11 +4646,11 @@ cdef class Generator:`
`4646`	`4646`	`--------`
`4647`	`4647`	`shuffle`
`4648`	`4648`	`permutation`
`4649`		`-`
	`4649`	`+`
`4650`	`4650`	`Notes`
`4651`	`4651`	`-----`
`4652`		- An important distinction between methods ``shuffle`` and ``permuted`` is
`4653`		- how they both treat the ``axis`` parameter which can be found at
	`4652`	+ An important distinction between methods ``shuffle`` and ``permuted`` is
	`4653`	+ how they both treat the ``axis`` parameter which can be found at
`4654`	`4654`	:ref:`generator-handling-axis-parameter`.
`4655`	`4655`
`4656`	`4656`	`Examples`
`@@ -4722,7 +4722,7 @@ cdef class Generator:`
`4722`	`4722`	`ifaxisisNone:`
`4723`	`4723`	`ifx.ndim>1:`
`4724`	`4724`	`ifnot (np.PyArray_FLAGS(out)& (np.NPY_ARRAY_C_CONTIGUOUS\|`
`4725`		`-np.NPY_ARRAY_F_CONTIGUOUS)):`
	`4725`	`+np.NPY_ARRAY_F_CONTIGUOUS)):`
`4726`	`4726`	`flags= (np.NPY_ARRAY_C_CONTIGUOUS\|`
`4727`	`4727`	`NPY_ARRAY_WRITEBACKIFCOPY)`
`4728`	`4728`	`to_shuffle=PyArray_FromArray(<np.PyArrayObject*>out,`
`@@ -4802,8 +4802,8 @@ cdef class Generator:`
`4802`	`4802`
`4803`	`4803`	`Notes`
`4804`	`4804`	`-----`
`4805`		- An important distinction between methods ``shuffle`` and ``permuted`` is
`4806`		- how they both treat the ``axis`` parameter which can be found at
	`4805`	+ An important distinction between methods ``shuffle`` and ``permuted`` is
	`4806`	+ how they both treat the ``axis`` parameter which can be found at
`4807`	`4807`	:ref:`generator-handling-axis-parameter`.
`4808`	`4808`
`4809`	`4809`	`Examples`
`@@ -5011,11 +5011,11 @@ def default_rng(seed=None):`
`5011`	`5011`	`Examples`
`5012`	`5012`	`--------`
`5013`	`5013`	`default_rng` is the recommended constructor for the random number class
`5014`		- `Generator`. Here are several ways we can construct a random
`5015`		- number generator using `default_rng` and the `Generator` class.
	`5014`	+ `Generator`. Here are several ways we can construct a random
	`5015`	+ number generator using `default_rng` and the `Generator` class.
`5016`	`5016`
`5017`	`5017`	Here we use `default_rng` to generate a random float:
`5018`		`-`
	`5018`	`+`
`5019`	`5019`	`>>> import numpy as np`
`5020`	`5020`	`>>> rng = np.random.default_rng(12345)`
`5021`	`5021`	`>>> print(rng)`
`@@ -5025,20 +5025,20 @@ def default_rng(seed=None):`
`5025`	`5025`	`0.22733602246716966`
`5026`	`5026`	`>>> type(rfloat)`
`5027`	`5027`	`<class 'float'>`
`5028`		`-`
`5029`		- Here we use `default_rng` to generate 3 random integers between 0
	`5028`	`+`
	`5029`	+ Here we use `default_rng` to generate 3 random integers between 0
`5030`	`5030`	`(inclusive) and 10 (exclusive):`
`5031`		`-`
	`5031`	`+`
`5032`	`5032`	`>>> import numpy as np`
`5033`	`5033`	`>>> rng = np.random.default_rng(12345)`
`5034`	`5034`	`>>> rints = rng.integers(low=0, high=10, size=3)`
`5035`	`5035`	`>>> rints`
`5036`	`5036`	`array([6, 2, 7])`
`5037`	`5037`	`>>> type(rints[0])`
`5038`	`5038`	`<class 'numpy.int64'>`
`5039`		`-`
	`5039`	`+`
`5040`	`5040`	`Here we specify a seed so that we have reproducible results:`
`5041`		`-`
	`5041`	`+`
`5042`	`5042`	`>>> import numpy as np`
`5043`	`5043`	`>>> rng = np.random.default_rng(seed=42)`
`5044`	`5044`	`>>> print(rng)`

`‎numpy/random/_mt19937.pyx‎`

Lines changed: 2 additions & 2 deletions

Original file line number	Diff line number	Diff line change
`@@ -43,8 +43,8 @@ cdef uint64_t mt19937_raw(void *st) noexcept nogil:`
`43`	`43`	`return<uint64_t>mt19937_next32(<mt19937_state*> st)`
`44`	`44`
`45`	`45`	`cdefclass MT19937(BitGenerator):`
`46`		`-"""`
`47`		`- MT19937(seed=None)`
	`46`	+# the first line is used to populate `__text_signature__`
	`47`	`+"""MT19937(seed=None)\n--`
`48`	`48`
`49`	`49`	`Container for the Mersenne Twister pseudo-random number generator.`
`50`	`50`

`‎numpy/random/_pcg64.pyx‎`

Lines changed: 4 additions & 4 deletions

Original file line number	Diff line number	Diff line change
`@@ -51,8 +51,8 @@ cdef double pcg64_cm_double(void* st) noexcept nogil:`
`51`	`51`	`return uint64_to_double(pcg64_cm_next64(<pcg64_state*>st))`
`52`	`52`
`53`	`53`	`cdefclass PCG64(BitGenerator):`
`54`		`-"""`
`55`		`- PCG64(seed=None)`
	`54`	+# the first line is used to populate `__text_signature__`
	`55`	`+"""PCG64(seed=None)\n--`
`56`	`56`
`57`	`57`	`BitGenerator for the PCG-64 pseudo-random number generator.`
`58`	`58`
`@@ -284,8 +284,8 @@ cdef class PCG64(BitGenerator):`
`284`	`284`
`285`	`285`
`286`	`286`	`cdefclass PCG64DXSM(BitGenerator):`
`287`		`-"""`
`288`		`- PCG64DXSM(seed=None)`
	`287`	+# the first line is used to populate `__text_signature__`
	`288`	`+"""PCG64DXSM(seed=None)\n--`
`289`	`289`
`290`	`290`	`BitGenerator for the PCG-64 DXSM pseudo-random number generator.`
`291`	`291`

`‎numpy/random/_philox.pyx‎`

Lines changed: 3 additions & 3 deletions

Original file line number	Diff line number	Diff line change
`@@ -52,8 +52,8 @@ cdef double philox_double(void*st) noexcept nogil:`
`52`	`52`	`return uint64_to_double(philox_next64(<philox_state*> st))`
`53`	`53`
`54`	`54`	`cdefclass Philox(BitGenerator):`
`55`		`-"""`
`56`		`- Philox(seed=None, counter=None, key=None)`
	`55`	+# the first line is used to populate `__text_signature__`
	`56`	`+"""Philox(seed=None, counter=None, key=None)\n--`
`57`	`57`
`58`	`58`	`Container for the Philox (4x64) pseudo-random number generator.`
`59`	`59`
`@@ -194,7 +194,7 @@ cdef class Philox(BitGenerator):`
`194`	`194`
`195`	`195`	`cdef _reset_state_variables(self):`
`196`	`196`	`cdef philox_state*rng_state=&self.rng_state`
`197`		`-`
	`197`	`+`
`198`	`198`	`rng_state[0].has_uint32=0`
`199`	`199`	`rng_state[0].uinteger=0`
`200`	`200`	`rng_state[0].buffer_pos= PHILOX_BUFFER_SIZE`

`‎numpy/random/_sfc64.pyx‎`

Lines changed: 2 additions & 2 deletions

Original file line number	Diff line number	Diff line change
`@@ -34,8 +34,8 @@ cdef double sfc64_double(void* st) noexcept nogil:`
`34`	`34`
`35`	`35`
`36`	`36`	`cdefclass SFC64(BitGenerator):`
`37`		`-"""`
`38`		`- SFC64(seed=None)`
	`37`	+# the first line is used to populate `__text_signature__`
	`38`	`+"""SFC64(seed=None)\n--`
`39`	`39`
`40`	`40`	`BitGenerator for Chris Doty-Humphrey's Small Fast Chaotic PRNG.`
`41`	`41`

`‎numpy/random/bit_generator.pxd‎`

Lines changed: 1 addition & 1 deletion

Original file line number	Diff line number	Diff line change
`@@ -31,5 +31,5 @@ cdef class SeedSequence():`
`31`	`31`	`np.ndarray[np.npy_uint32, ndim=1] entropy_array)`
`32`	`32`	`cdef get_assembled_entropy(self)`
`33`	`33`
`34`		`-cdefclassSeedlessSequence():`
	`34`	`+cdefclassSeedlessSeedSequence:`
`35`	`35`	`pass`

0 commit comments

Comments

(0)

Movatterモバイル変換

Navigation Menu

Search code, repositories, users, issues, pull requests...

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Uh oh!

Commit2b216b1

File tree

9 files changed

9 files changed

`‎numpy/random/_generator.pyx‎`

`‎numpy/random/_mt19937.pyx‎`

`‎numpy/random/_pcg64.pyx‎`

`‎numpy/random/_philox.pyx‎`

`‎numpy/random/_sfc64.pyx‎`

`‎numpy/random/bit_generator.pxd‎`

0 commit comments