forked from QuantEcon/rvlib
-
Notifications
You must be signed in to change notification settings - Fork 0
/
build_interface.py
703 lines (522 loc) · 18.8 KB
/
build_interface.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
import glob
import os
import textwrap
import yaml
# =========================================
# Write out file for special functions
# =========================================
def _initiate_specials():
'''
Initiate python file for special functions which are present in
the Rmath.h file -- used mainly for characteristic functions
'''
pre_code = """\
\"""
Special functions used mainly to evaluate characteristic
functions of various distributions.
@authors : Daniel Csaba <daniel.csaba@nyu.edu>
Spencer Lyon <spencer.lyon@stern.nyu.edu>
@date : 2016-07-26
\"""
from . import _rmath_ffi
from numba import vectorize, jit
from numba import cffi_support
cffi_support.register_module(_rmath_ffi)
# ---------------
# gamma function
# ---------------
gammafn = _rmath_ffi.lib.gammafn
@vectorize(nopython=True)
def gamma(x):
return gammafn(x)
# ---------------------
# log of gamma function
# ---------------------
lgammafn = _rmath_ffi.lib.lgammafn
@vectorize(nopython=True)
def lgamma(x):
return lgammafn(x)
# ----------------
# digamma function
# ----------------
digammafn = _rmath_ffi.lib.digamma
@vectorize(nopython=True)
def digamma(x):
return digammafn(x)
# -------------
# beta funciton
# -------------
betafn = _rmath_ffi.lib.beta
@vectorize(nopython=True)
def beta(x, y):
return betafn(x, y)
# -------------------------------------------
# modified Bessel function of the second kind
# -------------------------------------------
bessel_k_fn = _rmath_ffi.lib.bessel_k
@vectorize(nopython=True)
def bessel_k(nu, x):
return bessel_k_fn(x, nu, 1)
# ----------------------------------
# seed setting for the random number
# generator of the Rmath library
# ----------------------------------
set_seed_rmath = _rmath_ffi.lib.set_seed
def set_seed(x, y):
return set_seed_rmath(x, y)
"""
with open(os.path.join("rvlib", "specials.py"), "w") as f:
f.write(textwrap.dedent(pre_code))
# =========================================
# write out preamble for univariate classes
# =========================================
def _initiate_univariate():
'''
Initiate python file which collects all the
classes for different univariate distributions.
'''
# Define code which appears for all classes
pre_code = """\
\"""
Univariate distributions.
@authors : Daniel Csaba <daniel.csaba@nyu.edu>
Spencer Lyon <spencer.lyon@stern.nyu.edu>
@date : 2016-07-26
\"""
from os.path import join, dirname, abspath
from numba import vectorize, jit, jitclass
from numba import int32, float32
import numpy as np
from .specials import gamma, lgamma, digamma, beta, bessel_k, set_seed
from . import _rmath_ffi
from numba import cffi_support
cffi_support.register_module(_rmath_ffi)
# shut down divide by zero warnings for now
import warnings
warnings.filterwarnings("ignore")
import yaml
fn = join(dirname(abspath(__file__)), "metadata.yaml")
with open(fn, 'r') as ymlfile:
mtdt = yaml.load(ymlfile)
# --------------------------------------------------
# docstring following Spencer Lyon's distcan package
# https://github.com/spencerlyon2/distcan.git
# --------------------------------------------------
univariate_class_docstr = r\"""
Construct a distribution representing {name_doc} random variables. The
probability density function of the distribution is given by
.. math::
{pdf_tex}
Parameters
----------
{param_list}
Attributes
----------
{param_attributes}
location: scalar(float)
location of the distribution
scale: scalar(float)
scale of the distribution
shape: scalar(float)
shape of the distribution
mean : scalar(float)
mean of the distribution
median: scalar(float)
median of the distribution
mode : scalar(float)
mode of the distribution
var : scalar(float)
variance of the distribution
std : scalar(float)
standard deviation of the distribution
skewness : scalar(float)
skewness of the distribution
kurtosis : scalar(float)
kurtosis of the distribution
isplatykurtic : Boolean
boolean indicating if kurtosis > 0
isleptokurtic : bool
boolean indicating if kurtosis < 0
ismesokurtic : bool
boolean indicating if kurtosis == 0
entropy : scalar(float)
entropy value of the distribution
\"""
param_str = "{name_doc} : {kind}\\n {descr}"
def _create_param_list_str(names, descrs, kinds="scalar(float)"):
names = (names, ) if isinstance(names, str) else names
names = (names, ) if isinstance(names, str) else names
if isinstance(kinds, (list, tuple)):
if len(names) != len(kinds):
raise ValueError("Must have same number of names and kinds")
if isinstance(kinds, str):
kinds = [kinds for i in range(len(names))]
if len(descrs) != len(names):
raise ValueError("Must have same number of names and descrs")
params = []
for i in range(len(names)):
n, k, d = names[i], kinds[i], descrs[i]
params.append(param_str.format(name_doc=n, kind=k, descr=d))
return str.join("\\n", params)
def _create_class_docstr(name_doc, param_names, param_descrs,
param_kinds="scalar(float)",
pdf_tex=r"\\text{not given}", **kwargs):
param_list = _create_param_list_str(param_names, param_descrs,
param_kinds)
param_attributes = str.join(", ", param_names) + " : See Parameters"
return univariate_class_docstr.format(**locals())
"""
with open(os.path.join("rvlib", "univariate.py"), "w") as f:
f.write(textwrap.dedent(pre_code))
# globals are called from the textwrapper
# with distribution specific content
def _import_rmath(rname, pyname, *pargs):
"""
Map the `_rmath.ffi.lib.*` function names to match the Julia API.
"""
# extract Rmath.h function names
dfun = "d{}".format(rname)
pfun = "p{}".format(rname)
qfun = "q{}".format(rname)
rfun = "r{}".format(rname)
# construct python function names
pdf = "{}_pdf".format(pyname)
cdf = "{}_cdf".format(pyname)
ccdf = "{}_ccdf".format(pyname)
logpdf = "{}_logpdf".format(pyname)
logcdf = "{}_logcdf".format(pyname)
logccdf = "{}_logccdf".format(pyname)
invcdf = "{}_invcdf".format(pyname)
invccdf = "{}_invccdf".format(pyname)
invlogcdf = "{}_invlogcdf".format(pyname)
invlogccdf = "{}_invlogccdf".format(pyname)
rand = "{}_rand".format(pyname)
# make sure all names are available
has_rand = True
if rname == "nbeta" or rname == "nf" or rname == "nt":
has_rand = False
p_args = ", ".join(pargs)
code = """\
# ============================= NEW DISTRIBUTION =================================
{dfun} = _rmath_ffi.lib.{dfun}
{pfun} = _rmath_ffi.lib.{pfun}
{qfun} = _rmath_ffi.lib.{qfun}
@vectorize(nopython=True)
def {pdf}({p_args}, x):
return {dfun}(x, {p_args}, 0)
@vectorize(nopython=True)
def {logpdf}({p_args}, x):
return {dfun}(x, {p_args}, 1)
@vectorize(nopython=True)
def {cdf}({p_args}, x):
return {pfun}(x, {p_args}, 1, 0)
@vectorize(nopython=True)
def {ccdf}({p_args}, x):
return {pfun}(x, {p_args}, 0, 0)
@vectorize(nopython=True)
def {logcdf}({p_args}, x):
return {pfun}(x, {p_args}, 1, 1)
@vectorize(nopython=True)
def {logccdf}({p_args}, x):
return {pfun}(x, {p_args}, 0, 1)
@vectorize(nopython=True)
def {invcdf}({p_args}, q):
return {qfun}(q, {p_args}, 1, 0)
@vectorize(nopython=True)
def {invccdf}({p_args}, q):
return {qfun}(q, {p_args}, 0, 0)
@vectorize(nopython=True)
def {invlogcdf}({p_args}, lq):
return {qfun}(lq, {p_args}, 1, 1)
@vectorize(nopython=True)
def {invlogccdf}({p_args}, lq):
return {qfun}(lq, {p_args}, 0, 1)
""".format(**locals())
# append code for specific class to main `univariate.py` file
with open(os.path.join("rvlib", "univariate.py"), "a") as f:
f.write(textwrap.dedent(code))
if not has_rand:
# end here if we don't have rand. I put it in a `not has_rand` block
# to the rand_code can be at the right indentation level below
return
rand_code = """\
{rfun} = _rmath_ffi.lib.{rfun}
@jit(nopython=True)
def {rand}({p_args}):
return {rfun}({p_args})
""".format(**locals())
with open(os.path.join("rvlib", "univariate.py"), "a") as f:
f.write(textwrap.dedent(rand_code))
# function to write out the distribution specific part
def _write_class_specific(metadata, *pargs):
"""
Write out the distribution specific part of the class
which is not related to the imported Rmath functions.
Builds on _metadata_DIST and some derived locals.
"""
p_args = ", ".join(pargs)
p_args_self = ", ".join(["".join(("self.", par)) for par in pargs])
data = locals()
data.update(metadata)
class_specific = """\
@vectorize(nopython=True)
def {pyname}_mgf({p_args}, x):
return {mgf}
@vectorize(nopython=True)
def {pyname}_cf({p_args}, x):
return {cf}
# -------------
# {name}
# -------------
spec = [
{spec}
]
@jitclass(spec)
class {name}():
# set docstring
__doc__ = _create_class_docstr(**mtdt['{name}'])
def __init__(self, {p_args}):
{p_args_self} = {p_args}
def __str__(self):
return "{string}" %(self.params)
def __repr__(self):
return self.__str__()
# ===================
# Parameter retrieval
# ===================
@property
def params(self):
\"""Return a tuple of parameters.\"""
return ({p_args_self})
@property
def location(self):
\"""Return location parameter if exists.\"""
return {loc}
@property
def scale(self):
\"""Return scale parameter if exists.\"""
return {scale}
@property
def shape(self):
\"""Return shape parameter if exists.\"""
return {shape}
# ==========
# Statistics
# ==========
@property
def mean(self):
\"""Return the mean.\"""
return {mean}
@property
def median(self):
\"""Return the median.\"""
return {median}
@property
def mode(self):
\"""Return the mode.\"""
return {mode}
@property
def var(self):
\"""Return the variance.\"""
return {var}
@property
def std(self):
\"""Return the standard deviation.\"""
return {std}
@property
def skewness(self):
\"""Return the skewness.\"""
return {skewness}
@property
def kurtosis(self):
\"""Return the kurtosis.\"""
return {kurtosis}
@property
def isplatykurtic(self):
\"""Kurtosis being greater than zero.\"""
return self.kurtosis > 0
@property
def isleptokurtic(self):
\"""Kurtosis being smaller than zero.\"""
return self.kurtosis < 0
@property
def ismesokurtic(self):
\"""Kurtosis being equal to zero.\"""
return self.kurtosis == 0.0
@property
def entropy(self):
\"""Return the entropy.\"""
return {entropy}
def mgf(self, x):
\"""Evaluate the moment generating function at x.\"""
return {pyname}_mgf({p_args_self}, x)
def cf(self, x):
\"""Evaluate the characteristic function at x.\"""
return {pyname}_cf({p_args_self}, x)
# ==========
# Evaluation
# ==========
def insupport(self, x):
\"""When x is a scalar, return whether x is within
the support of the distribution. When x is an array,
return whether every element of x is within
the support of the distribution.\"""
return {insupport}
""".format(**data)
# append distribution specific code to `univariate.py` file
with open(os.path.join("rvlib", "univariate.py"), "a") as f:
f.write(textwrap.dedent(class_specific))
def _write_class_rmath(rname, pyname, *pargs):
"""
Call top level @vectorized evaluation methods from Rmath.
"""
# construct distribution specific function names
pdf = "{}_pdf".format(pyname)
cdf = "{}_cdf".format(pyname)
ccdf = "{}_ccdf".format(pyname)
logpdf = "{}_logpdf".format(pyname)
logcdf = "{}_logcdf".format(pyname)
logccdf = "{}_logccdf".format(pyname)
invcdf = "{}_invcdf".format(pyname)
invccdf = "{}_invccdf".format(pyname)
invlogcdf = "{}_invlogcdf".format(pyname)
invlogccdf = "{}_invlogccdf".format(pyname)
rand = "{}_rand".format(pyname)
# make sure all names are available
has_rand = True
if rname == "nbeta" or rname == "nf" or rname == "nt":
has_rand = False
# append 'self.' at the beginning of each parameter
p_args = ", ".join(["".join(("self.", par)) for par in pargs])
loc_code = """\
def pdf(self, x):
\"""The pdf value(s) evaluated at x.\"""
return {pdf}({p_args}, x)
def logpdf(self, x):
\"""The logarithm of the pdf value(s) evaluated at x.\"""
return {logpdf}({p_args}, x)
def loglikelihood(self, x):
\"""The log-likelihood of the distribution w.r.t. all
samples contained in array x.\"""
return sum({logpdf}({p_args}, x))
def cdf(self, x):
\"""The cdf value(s) evaluated at x.\"""
return {cdf}({p_args}, x)
def ccdf(self, x):
\"""The complementary cdf evaluated at x, i.e. 1 - cdf(x).\"""
return {ccdf}({p_args}, x)
def logcdf(self, x):
\"""The logarithm of the cdf value(s) evaluated at x.\"""
return {logcdf}({p_args}, x)
def logccdf(self, x):
\"""The logarithm of the complementary cdf evaluated at x.\"""
return {logccdf}({p_args}, x)
def quantile(self, q):
\"""The quantile value evaluated at q.\"""
return {invcdf}({p_args}, q)
def cquantile(self, q):
\"""The complementary quantile value evaluated at q.\"""
return {invccdf}({p_args}, q)
def invlogcdf(self, lq):
\"""The inverse function of the logcdf.\"""
return {invlogcdf}({p_args}, lq)
def invlogccdf(self, lq):
\"""The inverse function of the logccdf.\"""
return {invlogccdf}({p_args}, lq)
""".format(**locals())
# append code for class to main `univariate.py` file
with open(os.path.join("rvlib", "univariate.py"), "a") as f:
f.write(loc_code)
if not has_rand:
# end here if we don't have rand. I put it in a `not has_rand` block
# to the rand_code can be at the right indentation level below
return
rand_code = """\
# ========
# Sampling
# ========
def rand(self, n):
\"""Generates a vector of n independent samples from the distribution.\"""
out = np.empty(n)
for i, _ in np.ndenumerate(out):
out[i] = {rand}({p_args})
return out
""".format(**locals())
with open(os.path.join("rvlib", "univariate.py"), "a") as f:
f.write(rand_code)
def main():
# Write out specials.py
_initiate_specials()
# Preamble for univariate.py
_initiate_univariate()
# Normal
_import_rmath("norm", "norm", "mu", "sigma")
_write_class_specific(mtdt["Normal"], "mu", "sigma")
_write_class_rmath("norm", "norm", "mu", "sigma")
# Chisq
_import_rmath("chisq", "chisq", "v")
_write_class_specific(mtdt["Chisq"], "v")
_write_class_rmath("chisq", "chisq", "v")
# Uniform
_import_rmath("unif", "unif", "a", "b")
_write_class_specific(mtdt["Uniform"], "a", "b")
_write_class_rmath("unif", "unif", "a", "b")
# T
_import_rmath("t", "tdist", "v")
_write_class_specific(mtdt["T"], "v")
_write_class_rmath("t", "tdist", "v")
# LogNormal
_import_rmath("lnorm", "lognormal", "mu", "sigma")
_write_class_specific(mtdt["LogNormal"], "mu", "sigma")
_write_class_rmath("lnorm", "lognormal", "mu", "sigma")
# F
_import_rmath("f", "fdist", "v1", "v2")
_write_class_specific(mtdt["F"], "v1", "v2")
_write_class_rmath("f", "fdist", "v1", "v2")
# Gamma
_import_rmath("gamma", "gamma", "alpha", "beta")
_write_class_specific(mtdt["Gamma"], "alpha", "beta")
_write_class_rmath("gamma", "gamma", "alpha", "beta")
# Beta
_import_rmath("beta", "beta", "alpha", "beta")
_write_class_specific(mtdt["Beta"], "alpha", "beta")
_write_class_rmath("beta", "beta", "alpha", "beta")
# Exponential
_import_rmath("exp", "exp", "theta")
_write_class_specific(mtdt["Exponential"], "theta")
_write_class_rmath("exp", "exp", "theta")
# Cauchy
_import_rmath("cauchy", "cauchy", "mu", "sigma")
_write_class_specific(mtdt["Cauchy"], "mu", "sigma")
_write_class_rmath("cauchy", "cauchy", "mu", "sigma")
# Poisson
_import_rmath("pois", "pois", "mu")
_write_class_specific(mtdt["Poisson"], "mu")
_write_class_rmath("pois", "pois", "mu")
# Geometric
_import_rmath("geom", "geom", "p")
_write_class_specific(mtdt["Geometric"], "p")
_write_class_rmath("geom", "geom", "p")
# Binomial
_import_rmath("binom", "binom", "n", "p")
_write_class_specific(mtdt["Binomial"], "n", "p")
_write_class_rmath("binom", "binom", "n", "p")
# Logistic
_import_rmath("logis", "logis", "mu", "theta")
_write_class_specific(mtdt["Logistic"], "mu", "theta")
_write_class_rmath("logis", "logis", "mu", "theta")
# Weibull
_import_rmath("weibull", "weibull", "alpha", "theta")
_write_class_specific(mtdt["Weibull"], "alpha", "theta")
_write_class_rmath("weibull", "weibull", "alpha", "theta")
# Hypergeometric
_import_rmath("hyper", "hyper", "s", "f", "n")
_write_class_specific(mtdt["Hypergeometric"], "s", "f", "n")
_write_class_rmath("hyper", "hyper", "s", "f", "n")
# NegativeBinomial
_import_rmath("nbinom", "nbinom", "r", "p")
_write_class_specific(mtdt["NegativeBinomial"], "r", "p")
_write_class_rmath("nbinom", "nbinom", "r", "p")
with open(os.path.join("rvlib", "metadata.yaml"), 'r') as ymlfile:
mtdt = yaml.load(ymlfile)
main()