File: fixedp.h

package info (click to toggle)
crawl 2%3A0.34.0-1
  • links: PTS, VCS
  • area: main
  • in suites: forky, sid
  • size: 100,188 kB
  • sloc: cpp: 363,709; ansic: 27,765; javascript: 9,516; python: 8,463; perl: 3,293; java: 3,132; xml: 2,380; makefile: 1,835; sh: 611; objc: 250; cs: 15; sed: 9; lisp: 3
file content (782 lines) | stat: -rw-r--r-- 27,224 bytes parent folder | download | duplicates (2)
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
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
#pragma once

#include <algorithm>
#include <cassert>
#include <cmath>
#include <limits>
#include <ostream>
#include <string>
#include <type_traits>

/**
 * Header-only decimal fixed point library for DCSS.
 * Author: Kyle Rawlins
 *
 * This class is a minimalist wrapper around the way that DCSS already
 * approaches fixed point calculations, namely by using an integer value
 * multiplied by some scale parameter. It's not a general-purpose
 * fixed-point library; see e.g. https://github.com/johnmcfarlane/fixed_point
 * and https://github.com/johnmcfarlane/cnl. Unfortunately the fixed point
 * ecosystem is in something of a state of flux, and I decided that the
 * existing libraries were just too heavy/complicated to use for our purposes.
 * This differs from the code it is replacing in one qualitative way:
 * it allows rounding, rather than truncation behavior in division and
 * multiplication. The fact that dcss has traditionally used decimal scales
 * also impacts this class: a more general fixed point library would use a
 * binary point. This class loses out on the various optimizations that you can
 * do easily with binary fixed point arithmetic (efficient rounding, shift for
 * divide, etc.).
 *
 * Down the line, perhaps that CNL class (or something else) will develop into
 * a C++ standard and this can be replaced.
 *
 * This implements most operators you would want, as well as many cmath-style
 * functions. Casts out of fixedp are all explicit, but this allows implicit
 * casts from int, long, and float.
 *
 * operators: +=, -=, *=, /=, +, -, *, /, <, <=, >, >=, ==, !=, <<, unary -,
 * ++ (prefix, postifx), -- (prefix, postfix).
 *
 * casts: explicit casts to int, long, bool, float. int and long use trancating
 * behavior.
 *
 * constructors: int, long, unsigned int, unsigned long, float, double. See
 * also the static factory functions from_scaled, from_basetype, and
 * from_fixedp (which allows converting between precisions).
 *
 * cmath-style functions: trunc, round, ceil, floor, abs, max, min
 *
 * By default, multiplication/division use rounding rather than truncation. You
 * can toggle this behavior with `set_rounding`.
 *
 * Warning: this has no overflow protection/detection, so if you use a big Scale
 * parameter, or multiply/divide by large numbers, be careful of how much space
 * you have. fixedp should work with any integral type so in a pinch you can use
 * bigger types.
 *
 * @tparam BaseType the base type to use for storing numbers. Must be integral.
 * @tparam Scale the scale value to use: a constant factor that determines where
 * the decimal is in the underlying representation.
 *
 * Examples:
 *
 * > auto AC = 4 + fixedp<>(experience_level) / 3
 * Safely (to .01 precision) compute a fractional AC value, for nagas. This
 * makes use of implicit constructors forced by arithmetic with a fixedp.
 *
 * > return 2 + fixedp<>(experience_level) * 2 / 5 + (max(0, fixedp<>(experience_level) - 7) * 2 / 5);
 * A more elaborate example with implicit constructors, in this case for
 * calculating gargoyle AC at a given XL. Makes use of max.
 *
 * See fixedp::test_cases for further examples.
 *
 * Using a Scale that is a power of 2 is the most efficient use of precision,
 * but using a Scale that is a power of 10 is much more transparent. This class
 * contains no particular optimization for the power of 2 case, and is mainly
 * intended for use with a decimal point.
 */
template <typename BaseType=int, int Scale=100>
class fixedp
{
    static_assert(std::is_integral<BaseType>(),
                                    "fixedp requires an integral base type.");
    // it's *much* more convenient to use `int` and have this assert, than to
    // use `unsigned`.
    static_assert(Scale > 0, "fixedp requires a positive scale.");

    // the following is very weak...
    // TODO: could instead require at least 1 bits of integral space?
    static_assert(Scale <= std::numeric_limits<BaseType>::max(),
                "fixedp requires that Scale not exceed the limit of BaseType!");

    // the next two functions use type traits/SFINAE to implement a version of
    // abs that is safe for unsigned types; it will return the unsigned as-is
    // and not trigger any warnings or errors.  (Using regular abs on an
    // unsigned leads to both warnings and errors in a template context.)
    //
    // This is used to calculate the fractional part safely for all types (and
    // not used by the actual abs, which intentionally has undefined behavior
    // when BaseType is unsigned).

    template<typename T=BaseType> T _safe_abs(T n,
            typename std::enable_if<std::is_signed<T>::value, T>::type=0) const
    {
        using std::abs;
        return static_cast<T>(abs(n));
    }

    template<typename T=BaseType> T constexpr _safe_abs(T n,
            typename std::enable_if<!std::is_signed<T>::value, T>::type=0) const
    {
        return n;
    }

    BaseType content;
    static bool rounding;
public:
    // default constructor for sanity's sake. This is different than builtin
    // numeric types, which have undefined behavior if not initialized!
    fixedp()
        : content(0)
    {
    }

    // copy constructor
    fixedp(const fixedp<BaseType, Scale> &rhs)
        : content(rhs.content)
    {
    }

    // TODO: overflow check in constructors? (|amethyst)
    constexpr fixedp(int n) : content(static_cast<BaseType>(n) *
                                                static_cast<BaseType>(Scale))
    {
    }
    constexpr fixedp(long n) : content(static_cast<BaseType>(n) *
                                                static_cast<BaseType>(Scale))
    {
    }
    constexpr fixedp(long long n) : content(static_cast<BaseType>(n) *
                                                static_cast<BaseType>(Scale))
    {
    }
    constexpr fixedp(unsigned int n) : content(static_cast<BaseType>(n) *
                                                static_cast<BaseType>(Scale))
    {
    }
    constexpr fixedp(unsigned long n) : content(static_cast<BaseType>(n) *
                                                static_cast<BaseType>(Scale))
    {
    }

    // is rounding here really a good idea?
    // TODO: per amalloy, should these really be implicit? It's awfully
    // convenient.
    constexpr fixedp(float n) : content(std::round(n *
                                                static_cast<float>(Scale)))
    {
    }

    constexpr fixedp(double n) : content(std::round(n *
                                                static_cast<double>(Scale)))
    {
    }

    /**
     * Construct a fixedp from an integer plus a fraction of Scale. This does
     * no sanity checking on frac, so use it carefully. In particular,
     * int_part and frac should have the same sign, and frac should be <
     * int_part.
     */
    constexpr fixedp(int int_part, int frac)
        : content(static_cast<BaseType>(int_part) *
                    static_cast<BaseType>(Scale) + static_cast<BaseType>(frac))
    {
    }

    template <int OtherScale> explicit fixedp (fixedp<BaseType, OtherScale> n)
        : content(from_fixedp<OtherScale>(n).to_scaled())
    {
    }


    /**
     * Factory function, not a constructor. We don't want this to participate
     * in arithmetic.
     *
     * @param n a value that has already been scaled by Scale to initialize the
     * fixedp with.
     */
    static fixedp from_scaled(const BaseType n)
    {
        // TODO: this could be constexpr on c++14
        fixedp ret(0);
        ret.content = n;
        return ret;
    }

    static fixedp from_basetype(const BaseType i, const BaseType frac)
    {
        return from_scaled(i * static_cast<BaseType>(Scale) + frac);
    }


    /**
     * Convert from a fixedp with a different Scale.
     *
     * This uses fixedp division when lowering precision, and so will inherit
     * its rounding behavior from the containing class.
     *
     * @tparam OtherScale the scale of the other fixedp.
     * @param n the fixedp to convert from.
     */
    template <int OtherScale>
        static fixedp from_fixedp(fixedp<BaseType, OtherScale> n)
    {
        // TODO: conversion across BaseTypes?
        return (fixedp<BaseType, Scale>::from_basetype(n.frac_part(true), 0)
                            / OtherScale)
                + fixedp<BaseType, Scale>::from_basetype(n.integral_part(), 0);
    }

    /**
     * Set the rounding behavior of multiplication/division on a per-class
     * basis.
     */
    static void set_rounding(bool r)
    {
        rounding = r;
    }

    BaseType constexpr to_scaled() const
    {
        return content;
    }

    /**
     * Give the fractional part of the fixedp.
     *
     * @param signed whether to provide a signed or absolute version. This is
     * safe with unsigned BaseTypes.
     */
    BaseType frac_part(bool calc_signed = false) const
    {
        // This needs to use a somewhat more complicated version of abs in
        // order to work with unsigned types.
        return (!calc_signed || content >= 0)
            ? _safe_abs<BaseType>(content) % static_cast<BaseType>(Scale)
            : -(_safe_abs<BaseType>(content) % static_cast<BaseType>(Scale));
    }

    /**
     * Give the integral part (de-scaled) of the fixedp. Preserves sign.
     */
    BaseType constexpr integral_part() const
    {
        return content / static_cast<BaseType>(Scale);
    }

    fixedp trunc() const
    {
        return from_scaled(content - frac_part(true));
    }

    bool rounds_nearby() const
    {
        // TODO: better handling of non-even Scale?
        return frac_part() >= Scale / 2;
    }

    fixedp round() const
    {
        // this is a bit ugly...
        return rounds_nearby() ? (content >= 0 ? ++trunc()
                                               : --trunc())
                               : trunc();
    }

    // provide some cmath-style functions
    friend inline fixedp round(const fixedp &n)
    {
        return n.round();
    }
    friend inline fixedp trunc(const fixedp &n)
    {
        return n.trunc();
    }
    friend inline fixedp ceil(const fixedp &n)
    {
        return n.frac_part() == 0 ? n :
                (n.content >= 0 ? ++n.trunc() : n.trunc());
    }
    friend inline fixedp floor(const fixedp &n)
    {
        return n.frac_part() == 0 ? n :
                (n.content >= 0 ? n.trunc() : --n.trunc());
    }
    friend inline fixedp abs(const fixedp &n)
    {
        static_assert(std::is_signed<BaseType>(),
            "Using abs() with fixedp requires a signed BaseType.");

        // Even without the static_assert, this won't work with an unsigned
        // BaseType because abs is ambiguous without an explicit cast.
        using std::abs;
        return from_scaled(abs(n.content));
    }
    friend inline fixedp max(const fixedp &l, const fixedp &r)
    {
        return (l.content > r.content) ? l : r;
    }
    friend inline fixedp min(const fixedp &l, const fixedp &r)
    {
        return (l.content < r.content) ? l : r;
    }

    // assignment -- uses copy & swap paradigm
    void swap(fixedp rhs)
    {
        using std::swap;
        swap(content, rhs.content);
    }

    fixedp& operator=(fixedp rhs)
    {
        swap(rhs);
        return *this;
    }

    // conversion, explicit only
    explicit constexpr operator int() const
    {
        // truncates, following standard behavior of cast to integral types
        return static_cast<int>(integral_part());
    }
    explicit constexpr operator bool() const
    {
        return static_cast<bool>(content);
    }
    explicit constexpr operator long() const
    {
        // truncates, following standard behavior of cast to integral types
        return static_cast<long>(integral_part());
    }
    explicit operator float() const
    {
        return static_cast<float>(integral_part()) +
            static_cast<float>(frac_part(true) / static_cast<float>(Scale));
    }
    explicit operator double() const
    {
        return static_cast<double>(integral_part()) +
            static_cast<double>(frac_part(true) / static_cast<double>(Scale));
    }

    // string and stream output

    // to use fixedp with printf, your best options are things like:
    // > printf("%g", (float) n);
    // > printf("%.2f", (float) n);
    // using printf with the casts to str is not really recommended.

    std::string str() const
    {
        // TODO: chop trailing zeros / adjust precision?
        // would prefer %g behavior here. stream works better...
        return std::to_string((float) *this);
    }
    explicit operator std::string() const
    {
        return str();
    }
    inline friend std::string to_string(const fixedp& n)
    {
        return n.str();
    }
    inline friend std::ostream &operator<< (std::ostream &os, fixedp const &m)
    {
        // use the cast to float so that float formatting options can
        // be applied.
        os << (float) m;
        return os;
    }

    // comparison with other fixedps at the same template parameters
    // to cast acrtoss different fixedps, you'll need to explicitly convert
    // to a common type

    // because the constructors from numeric types are implicit, these are also
    // used for comparison with regular numerics.

    inline friend bool operator==(const fixedp& lhs, const fixedp& rhs)
    {
        return lhs.content == rhs.content;
    }
    inline friend bool operator!=(const fixedp& lhs, const fixedp& rhs)
    {
        return !operator==(lhs, rhs);
    }

    inline friend bool operator< (const fixedp& lhs, const fixedp& rhs)
    {
        return lhs.content < rhs.content;
    }
    inline friend bool operator> (const fixedp& lhs, const fixedp& rhs)
    {
        return  operator< (rhs, lhs);
    }
    inline friend bool operator<=(const fixedp& lhs, const fixedp& rhs)
    {
        return !operator> (lhs, rhs);
    }
    inline friend bool operator>=(const fixedp& lhs, const fixedp& rhs)
    {
        return !operator< (lhs,rhs);
    }

    // arithmetic with other fixedps

    // because the constructors from numeric types are implicit, these are also
    // used for arithmetic with regular numerics.

    fixedp operator-() const
    {
        static_assert(std::is_signed<BaseType>(),
            "Using unary - with fixedp requires a signed BaseType.");

        // Even without the static_assert above, this will produce an error on
        // unsigned types, because the cast to a signed type implied by - (by
        // way of abs) is ambiguous.  see abs()
        return from_scaled(-content);
    }

    fixedp& operator++()
    {
        content += Scale;
        return *this;
    }
    fixedp operator++(int) // postfix
    {
        fixedp tmp(*this);
        operator++();
        return tmp;
    }
    fixedp& operator--()
    {
        content -= Scale;
        return *this;
    }
    fixedp operator--(int)
    {
        fixedp tmp(*this);
        operator--();
        return tmp;
    }

    fixedp& operator+=(const fixedp& rhs)
    {
        content += rhs.content;
        return *this;
    }
    inline friend fixedp operator+(fixedp lhs, const fixedp& rhs)
    {
        lhs += rhs; // lhs is a copy
        return lhs;
    }

    fixedp& operator-=(const fixedp& rhs)
    {
        content -= rhs.content;
        return *this;
    }
    inline friend fixedp operator-(fixedp lhs, const fixedp& rhs)
    {
        lhs -= rhs; // lhs is a copy
        return lhs;
    }

    fixedp& operator*=(const fixedp& rhs)
    {
        if (rhs.frac_part() == 0)
            content *= rhs.integral_part();
        else
        {
            // TODO: better multiplication algorithm?
            content *= rhs.content;
            if (rounding)
            {
                // exploit truncation to do rounding
                content = (content > 0) ? content + Scale / 2
                                        : content - Scale / 2;
            }
            // content now is scaled by Scale * Scale, so we need to divide
            // one out.
            content /= Scale;
        }
        return *this;
    }
    inline friend fixedp operator*(fixedp lhs, const fixedp& rhs)
    {
        lhs *= rhs; // lhs is a copy
        return lhs;
    }

    fixedp& operator/=(const fixedp& rhs)
    {
        // TODO: this is probably a bad way to do this
        content *= Scale;
        if (rounding)
        {
            // exploit truncation to do rounding
            content = (content >= 0) ? content + _safe_abs(rhs.content) / 2
                                     : content - _safe_abs(rhs.content) / 2;
        }
        content /= rhs.content;
        return *this;
    }
    inline friend fixedp operator/(fixedp lhs, const fixedp& rhs)
    {
        lhs /= rhs; // lhs is a copy
        return lhs;
    }

#ifdef ASSERT
// crawlcode
#   define FPASSERT ASSERT
#else
#   define FPASSERT assert
#endif

    static void test_cases()
    {
        // exercise constructors
        fixedp<int, 100> test_dbl((double) 10.629);
        int int_test;
        int_test = 292;
        fixedp<> test_int(int_test);
        fixedp<long long> test_long((long long) 2e10);
        FPASSERT(test_long == 2e10);

        // test some cv qualified fixedps
        const fixedp<> cv_test(500, 20);
        // no idea why you'd want a volatile fixedp in crawl, but you can
        // uncomment these to make sure it would compile.
        // volatile fixedp<> cv_test2(500, 20);
        // const volatile fixedp<> cv_test3(500, 20);
        const fixedp<> cv_test4 = cv_test * 2;
        FPASSERT(cv_test == 500.2);
        FPASSERT(cv_test * 2 == cv_test4);
        FPASSERT(cv_test.round() == 500);
        FPASSERT(abs(cv_test * -1) == 500.2);
        // shouldn't compile
        // cv_test *= 2;

        // test assignment operators
        fixedp<int, 100> test((float) 10.629);
        FPASSERT(test == 10.63);
        FPASSERT(test == 10.629);
        test += 10;
        FPASSERT(test == 20.63);
        test += 10.2;
        FPASSERT(test == 30.83);
        test *= 2;
        FPASSERT(test == 61.66);
        test *= 2.5;
        FPASSERT(test == 154.15);
        test /= 2.5;
        FPASSERT(test == 61.66);
        test -= 102;
        FPASSERT(test == -40.34);
        test *= 2;
        FPASSERT(test == -80.68);
        FPASSERT(-test == 80.68);
        test = 96.236;
        FPASSERT(test.to_scaled() == 9624);

        // exercise every comparison operator for both positive and negative
        FPASSERT(test == 96.24);
        FPASSERT(test >= 96.24);
        FPASSERT(test <= 96.24);
        FPASSERT(test >= 96);
        FPASSERT(test > 96);
        FPASSERT(test != 96);
        FPASSERT(test < 97);
        FPASSERT(test <= 97);
        test = -96.236;
        FPASSERT(test == -96.24);
        FPASSERT(test >= -96.24);
        FPASSERT(test <= -96.24);
        FPASSERT(test <= -96);
        FPASSERT(test < -96);
        FPASSERT(test != -96);
        FPASSERT(test != 96);
        FPASSERT(test > -97);
        FPASSERT(test >= -97);

        FPASSERT(fixedp<>(-10.23).frac_part() == fixedp<>(-10.23).frac_part());
        FPASSERT(fixedp<>(-10.23).frac_part()
                == static_cast<int>(fixedp<unsigned int>(10.23).frac_part()));

        // testing rounding and other math operators
        FPASSERT(abs(fixedp<>(-10)) == 10);
        FPASSERT(abs(fixedp<>(10)) == 10);

        FPASSERT(fixedp<>(2.6).round() == 3);
        FPASSERT(fixedp<>(2.6).round() != 2);
        FPASSERT(fixedp<>(2.4).round() == 2);
        FPASSERT(fixedp<>(2.4).round() != 3);
        FPASSERT(fixedp<>(-2.6).round() == -3);
        FPASSERT(fixedp<>(-2.6).round() != -2);
        FPASSERT(fixedp<>(-2.4).round() == -2);
        FPASSERT(fixedp<>(-2.4).round() != -3);

        FPASSERT(ceil(fixedp<>(2.6)) == 3);
        FPASSERT(ceil(fixedp<>(2.6)) != 2);
        FPASSERT(ceil(fixedp<>(2.4)) == 3);
        FPASSERT(ceil(fixedp<>(2.4)) != 2);
        FPASSERT(ceil(fixedp<>(-2.6)) == -2);
        FPASSERT(ceil(fixedp<>(-2.6)) != -3);
        FPASSERT(ceil(fixedp<>(-2.4)) == -2);
        FPASSERT(ceil(fixedp<>(-2.4)) != -3);

        FPASSERT(floor(fixedp<>(2.6)) == 2);
        FPASSERT(floor(fixedp<>(2.6)) != 3);
        FPASSERT(floor(fixedp<>(2.4)) == 2);
        FPASSERT(floor(fixedp<>(2.4)) != 3);
        FPASSERT(floor(fixedp<>(-2.6)) == -3);
        FPASSERT(floor(fixedp<>(-2.6)) != -2);
        FPASSERT(floor(fixedp<>(-2.4)) == -3);
        FPASSERT(floor(fixedp<>(-2.4)) != -2);

        FPASSERT(fixedp<>(2.6).trunc() == 2);
        FPASSERT(fixedp<>(2.6).trunc() != 3);
        FPASSERT(fixedp<>(2.4).trunc() == 2);
        FPASSERT(fixedp<>(2.4).trunc() != 3);
        FPASSERT(fixedp<>(-2.6).trunc() == -2);
        FPASSERT(fixedp<>(-2.6).trunc() != -3);
        FPASSERT(fixedp<>(-2.4).trunc() == -2);
        FPASSERT(fixedp<>(-2.4).trunc() != -3);

        FPASSERT(min(fixedp<>(5), fixedp<>(5.5)) == 5);
        FPASSERT(min(fixedp<>(-5), fixedp<>(5.5)) == -5);
        FPASSERT(min(fixedp<>(5), fixedp<>(-5.5)) == -5.5);
        FPASSERT(min(fixedp<>(-5), fixedp<>(-5.5)) == -5.5);
        FPASSERT(max(fixedp<>(5), fixedp<>(5.5)) == 5.5);
        FPASSERT(max(fixedp<>(-5), fixedp<>(5.5)) == 5.5);
        FPASSERT(max(fixedp<>(5), fixedp<>(-5.5)) == 5);
        FPASSERT(max(fixedp<>(-5), fixedp<>(-5.5)) == -5);
        // implicit cast examples for min/max
        FPASSERT(min(fixedp<>(5), 5.5) == 5);
        FPASSERT(max(fixedp<>(5), 5.5) == 5.5);

        // test casts
        FPASSERT(fixedp<>(10, 20) == fixedp<>(10.2));
        FPASSERT((bool) fixedp<>(1));
        FPASSERT((int) fixedp<>(1.2) == 1);
        FPASSERT((int) fixedp<>(-1.6) == -1);
        FPASSERT((long) fixedp<>(1.2) == 1);
        FPASSERT((long) fixedp<>(-1.6) == -1);
        FPASSERT((float) fixedp<>(1.2) > 1.1888);
        FPASSERT((float) fixedp<>(-1.6) < -1.1899);

        // test binary arithmetic
        FPASSERT(fixedp<>(15) + fixedp<>(12.2) == fixedp<>(27.2));
        FPASSERT(fixedp<>(15) - fixedp<>(12.2) == fixedp<>(2.8));
        FPASSERT(fixedp<>(15) * fixedp<>(12.2) == fixedp<>(183));
        FPASSERT(fixedp<>(183) / fixedp<>(12.2) == fixedp<>(15));

        // arithmetic with implicit casts to fixedp
        FPASSERT(fixedp<>(15) + 12.2 == fixedp<>(27.2));
        FPASSERT(fixedp<>(15) - 12.2 == fixedp<>(2.8));
        FPASSERT(fixedp<>(15) * 12.2 == fixedp<>(183));
        FPASSERT(fixedp<>(183) / 12.2 == fixedp<>(15));
        FPASSERT(15 + fixedp<>(12.2) == fixedp<>(27.2));
        FPASSERT(15 - fixedp<>(12.2) == fixedp<>(2.8));
        FPASSERT(15 * fixedp<>(12.2) == fixedp<>(183));
        FPASSERT(183 / fixedp<>(12.2) == fixedp<>(15));

        FPASSERT((fixedp<int, 10>(11) / 10 == 1.1));
        FPASSERT((fixedp<int, 10>(-11) / 10 == -1.1));
        FPASSERT((fixedp<int, 10>(11) / -10 == -1.1));
        FPASSERT((fixedp<int, 10>(-11) / -10 == 1.1));


        // test rounding and truncation behavior for /,*
        test = fixedp<>(116.66);
        auto test2 = fixedp<>(114.44);
        fixedp<>::set_rounding(false);
        FPASSERT((test * 100) / 100 == test);
        FPASSERT((test * 0.01) * 100 == 116); // loses 2 digits
        FPASSERT((test / 100) * 100 == 116);
        FPASSERT((test2 * 0.01) * 100 == 114);
        FPASSERT((test2 / 100) * 100 == 114);
        FPASSERT((test * 0.01) == (test / 100));
        FPASSERT((test / 0.01) == (test * 100));
        fixedp<>::set_rounding(true);
        FPASSERT((test * 100) / 100 == test);
        FPASSERT((test * 0.01) * 100 == 117); // loses 2 digits, rounds
        FPASSERT((test / 100) * 100 == 117);
        FPASSERT((test2 * 0.01) * 100 == 114);
        FPASSERT((test2 / 100) * 100 == 114);
        FPASSERT((test * 0.01) == (test / 100));
        FPASSERT((test / 0.01) == (test * 100));
        // full combination of rounding + signs
        FPASSERT((test * -0.01) * 100 == -117);
        FPASSERT((-test * 0.01) * 100 == -117);
        FPASSERT((-test * -0.01) * 100 == 117);
        FPASSERT((test / -100) * 100 == -117);
        FPASSERT((-test / 100) * 100 == -117);
        FPASSERT((-test / -100) * 100 == 117);

        // comparison with explicit casts to fixedp
        // didn't bother to do this thoroughly
        FPASSERT(15.0 + 12.2 == fixedp<>(27.2));
        FPASSERT(15.0 - 12.2 == fixedp<>(2.8));
        FPASSERT(15.0 * 12.2 == fixedp<>(183));
        FPASSERT(183.0 / 12.2 == fixedp<>(15));

        // test conversions between fixedps.
        // these need the double parens because `assert` gets confused with too
        // many commas.
        FPASSERT((fixedp<int, 100>::from_fixedp<10>(fixedp<int, 10>(10)) == 10));
        FPASSERT((fixedp<int, 100>::from_fixedp<1000>(
                                    fixedp<int, 1000>(10)) == 10));
        FPASSERT((fixedp<int, 100>::from_fixedp<10>(
                                    fixedp<int, 10>(11) / 10) == 1.1));
        FPASSERT((fixedp<int, 100>::from_fixedp<10>(
                                    fixedp<int, 10>(-11) / 10) == -1.1));
        FPASSERT((fixedp<int, 100>::from_fixedp<1000>(
                                    fixedp<int, 1000>(1111) / 1000) == 1.11));
        // test truncation behavior
        FPASSERT((fixedp<int, 100>::from_fixedp<1000>(
                                    fixedp<int, 1000>(1116) / 1000) == 1.12));

        FPASSERT((fixedp<int, 100>(fixedp<int, 10>(10)) == 10));
        FPASSERT((fixedp<int, 100>(fixedp<int, 1000>(10)) == 10));
        FPASSERT((fixedp<int, 100>(fixedp<int, 10>(11) / 10) == 1.1));
        FPASSERT((fixedp<int, 100>(fixedp<int, 10>(-11) / 10) == -1.1));
        FPASSERT((fixedp<int, 100>(fixedp<int, 1000>(1111) / 1000) == 1.11));
        FPASSERT((fixedp<int, 100>(fixedp<int, 1000>(1116) / 1000) == 1.12));


        // test a few max scales; these have only a fractional part
        fixedp<char, 127>(10);
        fixedp<unsigned char, 255>(10);

        // these should cause a static_assert failure if uncommented:
        // fixedp<char, 16000>(10);
        // fixedp<char, 128>(10);
        // fixedp<unsigned char, 256>(10);
        // float test_badneg = (float) -fixedp<unsigned int>(10,20);
        // float test_badabs = (float) abs(fixedp<unsigned int>(10,20));
        // fixedp<int, -100> test_badscale(100);
    }
#undef FPASSERT

};

// sources vary on whether you are allowed to and/or should do this. But we
// want a way of getting these values somehow, and unless/until some c++
// version prevents me from doing it, I'd rather use the standard API if at
// all possible.
namespace std
{
    template<typename BaseType, int Scale>
    class numeric_limits<fixedp<BaseType, Scale>>
    {
    public:
        static fixedp<BaseType, Scale> min()
        {
            return fixedp<BaseType, Scale>::from_scaled(numeric_limits<BaseType>::min());
        }

        static fixedp<BaseType, Scale> max()
        {
            return fixedp<BaseType, Scale>::from_scaled(numeric_limits<BaseType>::max());
        }
    };
}

// round by default
template <typename BaseType, int Scale> bool
                                    fixedp<BaseType, Scale>::rounding = true;