File size: 35,106 Bytes
669d6a1
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
"""
Unified cache system for AFML - eliminates all duplication.

This module replaces:
- robust_cache_keys.py (most functionality)
- cv_cache.py (CV-specific caching)
- Parts of cache_monitoring.py integration

One unified system with consistent behavior across all decorators.
"""

import hashlib
import inspect
import json
import time
from functools import wraps
from typing import Any, Callable, Dict, Optional, Tuple

import numpy as np
import pandas as pd
from loguru import logger
from scipy.stats._distn_infrastructure import (rv_continuous_frozen,
                                               rv_discrete_frozen)
from sklearn.base import BaseEstimator
from sklearn.pipeline import Pipeline

from ..util.pipelines import MyPipeline
from . import cache_stats, memory
from .cache_monitoring import get_cache_monitor

# =============================================================================
# Core: Function Versioning Support
# =============================================================================


def _get_function_source_hash(func: Callable) -> Optional[str]:
    """
    Get stable hash of function source code.

    Unwraps decorators to hash the original function, not wrapper code.
    This ensures nested @cacheable or other decorators don't affect the hash.

    Returns None if source cannot be obtained (built-ins, etc.)
    """
    # Unwrap to find original function
    # This handles @functools.wraps and similar patterns
    original_func = func
    while hasattr(original_func, "__wrapped__"):
        original_func = original_func.__wrapped__

    try:
        source = inspect.getsource(original_func)
        source_hash = hashlib.md5(source.encode()).hexdigest()[:12]

        # For nested functions with closures, include closure state in hash
        # This ensures different closure values create different cache keys
        if original_func.__closure__:
            closure_hash = _get_closure_hash(original_func)
            if closure_hash:
                return f"{source_hash}_{closure_hash}"

        return source_hash
    except (OSError, TypeError):
        # Can't get source (built-in, dynamically created, etc.)
        return None


def _get_closure_hash(func: Callable) -> Optional[str]:
    """
    Hash closure variables for nested functions.

    This ensures that nested functions with different closure state
    get different cache keys, even though their source code is identical.
    """
    if not func.__closure__:
        return None

    try:
        # Extract closure values
        closure_values = []
        for cell in func.__closure__:
            try:
                val = cell.cell_contents
                # Try to create a stable representation
                if isinstance(val, (int, float, str, bool, type(None))):
                    closure_values.append(f"{type(val).__name__}:{val}")
                elif isinstance(val, (list, tuple)):
                    closure_values.append(f"{type(val).__name__}:{hash(str(val))}")
                else:
                    # For complex objects, use type and id
                    closure_values.append(f"{type(val).__name__}:{id(val)}")
            except ValueError:
                # Cell is empty (rare)
                closure_values.append("empty")

        closure_str = "_".join(closure_values)
        return hashlib.md5(closure_str.encode()).hexdigest()[:8]
    except Exception as e:
        logger.debug(f"Failed to hash closure: {e}")
        return None


def _get_function_file_mtime(func: Callable) -> Optional[float]:
    """Get modification time of function's source file."""
    try:
        file_path = inspect.getfile(func)
        from pathlib import Path

        return Path(file_path).stat().st_mtime
    except (OSError, TypeError):
        return None


# =============================================================================
# Core: Unified Cache Key Generator (with versioning support)
# =============================================================================


class UnifiedCacheKeyGenerator:
    """
    Single cache key generator for all AFML use cases.

    Handles:
    - Pandas DataFrames (with temporal awareness)
    - NumPy arrays
    - Sklearn estimators
    - Scipy distributions (KEY FIX for clf_hyper_fit)
    - CV generators (KFold, PurgedKFold, etc.)
    - Time-series data with lookback periods
    - Function versioning (auto-invalidation on code changes) - DEFAULT ENABLED
    """

    @staticmethod
    def generate_key(
        func: Callable,
        args: tuple,
        kwargs: dict,
        time_aware: bool = False,
        auto_versioning: bool = True,
    ) -> str:
        """
        Generate unified cache key.

        Args:
            func: Function being cached
            args: Positional arguments
            kwargs: Keyword arguments
            time_aware: If True, include temporal bounds in key
            auto_versioning: If True, include function source hash in key (DEFAULT)

        Returns:
            MD5 hash representing unique call signature
        """
        key_parts = [func.__module__, func.__qualname__]

        # Add function version to key (NOW DEFAULT, with graceful fallback)
        if auto_versioning:
            func_hash = _get_function_source_hash(func)
            if func_hash:
                key_parts.append(f"v_{func_hash}")
                logger.trace(
                    f"Auto-versioning enabled for {func.__qualname__}: v_{func_hash[:8]}"
                )
            else:
                # Graceful fallback to mtime if source unavailable
                mtime = _get_function_file_mtime(func)
                if mtime:
                    key_parts.append(f"mtime_{int(mtime)}")
                    logger.debug(
                        f"Cannot hash source for {func.__qualname__} "
                        f"(built-in or dynamic), using file mtime for versioning"
                    )
                else:
                    # Last resort: no versioning (but don't crash)
                    logger.warning(
                        f"Auto-versioning unavailable for {func.__qualname__} "
                        f"(no source or file). Cache won't invalidate on changes. "
                        f"Consider using @cacheable(auto_versioning=False) explicitly."
                    )

        # Get function signature for proper parameter mapping
        sig = inspect.signature(func)

        try:
            bound = sig.bind(*args, **kwargs)
            bound.apply_defaults()

            # Extract time range if time-aware
            time_range = None
            if time_aware:
                time_range = UnifiedCacheKeyGenerator._extract_time_range(
                    bound.arguments
                )

            # Hash each parameter
            for param_name, param_value in bound.arguments.items():
                key_part = UnifiedCacheKeyGenerator._hash_parameter(
                    param_name, param_value
                )
                key_parts.append(key_part)

            # Add time range to key if present
            if time_range:
                start, end = time_range
                key_parts.append(f"time_{start}_{end}")

        except Exception as e:
            logger.debug(f"Parameter binding failed for {func.__name__}: {e}")
            # Fallback to positional hashing
            for i, arg in enumerate(args):
                key_parts.append(
                    UnifiedCacheKeyGenerator._hash_parameter(f"arg_{i}", arg)
                )
            for k, v in kwargs.items():
                key_parts.append(UnifiedCacheKeyGenerator._hash_parameter(k, v))

        combined = "_".join(key_parts)
        return hashlib.md5(combined.encode()).hexdigest()

    @staticmethod
    def _hash_parameter(name: str, value: Any) -> str:
        """Route parameter to appropriate hashing method."""

        # 1. Scipy distributions (fixes clf_hyper_fit caching)
        if isinstance(value, (rv_discrete_frozen, rv_continuous_frozen)):
            return UnifiedCacheKeyGenerator._hash_scipy_dist(name, value)

        # 2. Dictionaries (may contain scipy distributions)
        if isinstance(value, dict):
            return UnifiedCacheKeyGenerator._hash_dict(name, value)

        # 3. Sklearn estimators
        try:
            if isinstance(value, (BaseEstimator, Pipeline, MyPipeline)):
                return UnifiedCacheKeyGenerator._hash_estimator(name, value)
        except ImportError:
            pass

        # 4. CV generators (for cross-validation caching)
        if hasattr(value, "split") and hasattr(value, "n_splits"):
            return UnifiedCacheKeyGenerator._hash_cv_generator(name, value)

        # 5. Pandas DataFrames
        if isinstance(value, pd.DataFrame):
            return UnifiedCacheKeyGenerator._hash_dataframe(name, value)

        # 6. Pandas Series
        if isinstance(value, pd.Series):
            return UnifiedCacheKeyGenerator._hash_series(name, value)

        # 7. NumPy arrays
        if isinstance(value, np.ndarray):
            return UnifiedCacheKeyGenerator._hash_numpy_array(name, value)

        # 8. Sequences (lists, tuples)
        if isinstance(value, (list, tuple)):
            return UnifiedCacheKeyGenerator._hash_sequence(name, value)

        # 9. Primitives
        if isinstance(value, (int, float, str, bool, type(None))):
            return f"{name}_{type(value).__name__}_{hash(value)}"

        # 10. Fallback
        return UnifiedCacheKeyGenerator._hash_generic(name, value)

    @staticmethod
    def _hash_scipy_dist(name: str, dist) -> str:
        """Hash scipy distribution deterministically."""
        dist_type = type(dist).__name__
        args = dist.args if hasattr(dist, "args") else ()
        kwds = dist.kwds if hasattr(dist, "kwds") else {}

        # Serialize args and kwds for deterministic hashing
        args_serialized = UnifiedCacheKeyGenerator._serialize_for_hashing(args)
        kwds_serialized = UnifiedCacheKeyGenerator._serialize_for_hashing(kwds)

        params = {"type": dist_type, "args": args_serialized, "kwds": kwds_serialized}
        param_str = json.dumps(params, sort_keys=True, default=str)
        param_hash = hashlib.md5(param_str.encode()).hexdigest()[:8]

        return f"{name}_dist_{dist_type}_{param_hash}"

    @staticmethod
    def _hash_dict(name: str, d: dict) -> str:
        """Hash dictionary recursively using deterministic serialization."""
        if not d:
            return f"{name}_empty_dict"

        # Use the new serialization method
        serialized = UnifiedCacheKeyGenerator._serialize_for_hashing(d)
        serialized_str = json.dumps(serialized, sort_keys=True, default=str)
        return hashlib.md5(serialized_str.encode()).hexdigest()[:8]

    @staticmethod
    def _hash_estimator(name: str, estimator) -> str:
        """
        Hash sklearn estimator recursively to handle Pipelines.

        FIX: Use deterministic parameter extraction instead of recursive hashing
        to ensure consistent cache keys.
        """
        try:
            est_type = type(estimator).__name__

            # Special handling for Pipeline - serialize all parameters including nested steps
            if isinstance(estimator, (Pipeline, MyPipeline)):
                # Serialize the entire pipeline with all parameters
                params = estimator.get_params(deep=True)  # Use deep=True to get nested params

                # Extract step information for better debugging
                step_info = []
                if hasattr(estimator, "steps"):
                    for step_name, step_estimator in estimator.steps:
                        step_info.append(
                            {
                                "name": step_name,
                                "type": type(step_estimator).__name__,
                                "params": step_estimator.get_params(),
                            }
                        )

                # Create deterministic serialization
                serializable = {
                    "type": est_type,
                    "params": UnifiedCacheKeyGenerator._serialize_for_hashing(params),
                    "steps": UnifiedCacheKeyGenerator._serialize_for_hashing(step_info),
                }

                param_str = json.dumps(serializable, sort_keys=True, default=str)
                param_hash = hashlib.md5(param_str.encode()).hexdigest()[:12]

                return f"{name}_est_{param_hash}"

            elif isinstance(estimator, BaseEstimator):
                # For regular estimators, get ALL parameters (deep=True)
                params = estimator.get_params(deep=True)

                # Create deterministic serialization
                serializable = {
                    "type": est_type,
                    "params": UnifiedCacheKeyGenerator._serialize_for_hashing(params),
                }

                param_str = json.dumps(serializable, sort_keys=True, default=str)
                param_hash = hashlib.md5(param_str.encode()).hexdigest()[:12]

                return f"{name}_est_{param_hash}"

        except Exception as e:
            logger.warning(f"Estimator hashing failed for {name}: {e}")
            # Fallback: use type and hash of repr
            return f"{name}_est_{type(estimator).__name__}_{hash(repr(estimator))}"

    @staticmethod
    def _serialize_for_hashing(obj: Any) -> Any:
        """
        Serialize any object for consistent hashing.

        This ensures deterministic serialization for all parameter types.
        """
        # Handle None
        if obj is None:
            return None

        # Handle basic types
        if isinstance(obj, (int, float, str, bool)):
            return obj

        # Handle numpy types
        if isinstance(obj, np.integer):
            return int(obj)
        if isinstance(obj, np.floating):
            return float(obj)
        if isinstance(obj, np.ndarray):
            # For arrays, use shape and a sample for hashing
            if obj.size > 1000:
                sample = obj.flat[:: max(1, obj.size // 100)]
                rounded = np.round(sample, decimals=6)
                content = rounded.tolist()
            else:
                rounded = np.round(obj, decimals=6)
                content = rounded.tolist()
            return {
                "_type": "np_array",
                "shape": obj.shape,
                "dtype": str(obj.dtype),
                "content": content[:100],  # Limit size
            }

        # Handle scipy distributions
        if isinstance(obj, (rv_discrete_frozen, rv_continuous_frozen)):
            return {
                "_type": "scipy_dist",
                "dist": type(obj).__name__,
                "args": obj.args if hasattr(obj, "args") else (),
                "kwds": obj.kwds if hasattr(obj, "kwds") else {},
            }

        # Handle sklearn estimators
        try:
            if isinstance(obj, BaseEstimator):
                # For estimators in parameters, get their params but don't nest recursively
                return {
                    "_type": "estimator",
                    "class": type(obj).__name__,
                    "params": obj.get_params(deep=False),  # shallow only to avoid recursion
                }
        except Exception as e:
            logger.error(e)

        # Handle sequences
        if isinstance(obj, (list, tuple)):
            return [UnifiedCacheKeyGenerator._serialize_for_hashing(item) for item in obj]

        # Handle dictionaries
        if isinstance(obj, dict):
            return {
                k: UnifiedCacheKeyGenerator._serialize_for_hashing(v)
                for k, v in sorted(obj.items())  # Sort for deterministic ordering
            }

        # Handle pandas objects
        if isinstance(obj, pd.DataFrame):
            return {
                "_type": "dataframe",
                "shape": obj.shape,
                "columns": list(obj.columns),
                "index_type": type(obj.index).__name__,
                "dtypes": {col: str(dtype) for col, dtype in obj.dtypes.items()},
            }

        if isinstance(obj, pd.Series):
            return {
                "_type": "series",
                "shape": obj.shape,
                "dtype": str(obj.dtype),
                "index_type": type(obj.index).__name__,
            }

        # For everything else, use repr but with type prefix
        return {"_type": type(obj).__name__, "repr": repr(obj)}

    @staticmethod
    def _hash_cv_generator(name: str, cv_gen) -> str:
        """Hash cross-validation generator."""
        try:
            cv_type = type(cv_gen).__name__

            params = {}
            if hasattr(cv_gen, "n_splits"):
                params["n_splits"] = cv_gen.n_splits
            if hasattr(cv_gen, "pct_embargo"):
                params["pct_embargo"] = cv_gen.pct_embargo
            if hasattr(cv_gen, "t1") and isinstance(cv_gen.t1, pd.Series):
                t1 = cv_gen.t1
                params["t1_len"] = len(t1)
                params["t1_start"] = str(t1.index[0])
                params["t1_end"] = str(t1.index[-1])

            param_str = json.dumps(params, sort_keys=True, default=str)
            param_hash = hashlib.md5(param_str.encode()).hexdigest()[:8]

            return f"{name}_cv_{cv_type}_{param_hash}"
        except Exception:
            return f"{name}_cv_{type(cv_gen).__name__}_{id(cv_gen)}"

    @staticmethod
    def _hash_dataframe(name: str, df: pd.DataFrame) -> str:
        """Hash DataFrame including structure and content."""
        parts = [
            f"shape_{df.shape}",
            f"cols_{hashlib.md5(str(tuple(df.columns)).encode()).hexdigest()[:8]}",
        ]

        # Hash index
        if isinstance(df.index, pd.DatetimeIndex) and len(df) > 0:
            parts.append(f"idx_dt_{df.index[0]}_{df.index[-1]}_{len(df)}")
        else:
            idx_hash = hashlib.md5(str(tuple(df.index[:10])).encode()).hexdigest()[:8]
            parts.append(f"idx_{idx_hash}")

        # Sample content for large DataFrames
        if df.size > 10000:
            sample = df.iloc[:: max(1, len(df) // 100)]
            content_str = sample.to_json(orient="values", double_precision=6)
        else:
            content_str = df.to_json(orient="values", double_precision=6)

        content_hash = hashlib.md5(content_str.encode()).hexdigest()
        parts.append(f"data_{content_hash}")
        return f"{name}_df_{'_'.join(parts)}"

    @staticmethod
    def _hash_series(name: str, series: pd.Series) -> str:
        """Hash pandas Series."""
        parts = [f"len_{len(series)}", f"dtype_{series.dtype}"]

        if isinstance(series.index, pd.DatetimeIndex) and len(series) > 0:
            parts.append(f"idx_dt_{series.index[0]}_{series.index[-1]}")

        if len(series) > 1000:
            sample = series.iloc[:: max(1, len(series) // 100)]
            content_str = sample.to_json(orient="values", double_precision=6)
        else:
            content_str = series.to_json(orient="values", double_precision=6)

        content_hash = hashlib.md5(content_str.encode()).hexdigest()
        parts.append(f"data_{content_hash}")
        return f"{name}_ser_{'_'.join(parts)}"

    @staticmethod
    def _hash_numpy_array(name: str, arr: np.ndarray) -> str:
        """Hash numpy array."""
        if arr.size > 10000:
            sample = arr.flat[:: max(1, arr.size // 1000)]
            rounded = np.round(sample, decimals=6)  # Avoid float precision issues
        else:
            rounded = np.round(arr, decimals=6)  # Avoid float precision issues

        content_str = ",".join(str(x) for x in rounded)
        content_hash = hashlib.md5(content_str.encode()).hexdigest()
        return f"{name}_arr_{arr.shape}_{arr.dtype}_{content_hash}"

    @staticmethod
    def _hash_sequence(name: str, seq) -> str:
        """Hash list or tuple recursively."""
        if not seq:
            return f"{name}_empty_seq"

        elem_hashes = [
            UnifiedCacheKeyGenerator._hash_parameter(f"{name}_{i}", item)
            for i, item in enumerate(seq)
        ]
        combined = "_".join(elem_hashes)
        return hashlib.md5(combined.encode()).hexdigest()[:8]

    @staticmethod
    def _hash_generic(name: str, obj: Any) -> str:
        """Fallback for unknown types."""
        try:
            return f"{name}_{type(obj).__name__}_{hash(repr(obj))}"
        except Exception:
            return f"{name}_{type(obj).__name__}_{id(obj)}"

    @staticmethod
    def _extract_time_range(
        params: dict,
    ) -> Optional[Tuple[pd.Timestamp, pd.Timestamp]]:
        """Extract temporal range from parameters for time-aware caching."""
        # Check for explicit time parameters
        if "start_date" in params and "end_date" in params:
            return (
                pd.Timestamp(params["start_date"]),
                pd.Timestamp(params["end_date"]),
            )

        # Check for DataFrames with DatetimeIndex
        for param_value in params.values():
            if isinstance(param_value, pd.DataFrame):
                if (
                    isinstance(param_value.index, pd.DatetimeIndex)
                    and len(param_value) > 0
                ):
                    return (param_value.index[0], param_value.index[-1])
            elif isinstance(param_value, pd.Series):
                if (
                    isinstance(param_value.index, pd.DatetimeIndex)
                    and len(param_value) > 0
                ):
                    return (param_value.index[0], param_value.index[-1])

        return None


# =============================================================================
# Core: Unified Cache Monitor
# =============================================================================


class UnifiedCacheMonitor:
    """Single monitoring system for all cache operations."""

    def __init__(self):
        self.core_monitor = get_cache_monitor()
        self.cache_stats = cache_stats

    def track_cache_call(
        self,
        func_name: str,
        is_hit: bool,
        computation_time: Optional[float] = None,
        cache_key: Optional[str] = None,
    ):
        """Track cache operation (hit/miss + timing)."""
        # Update stats
        if is_hit:
            self.cache_stats.record_hit(func_name)
        else:
            self.cache_stats.record_miss(func_name)

        # Track access time
        self.core_monitor.track_access(func_name)

        # Track computation time for misses
        if computation_time is not None and not is_hit:
            self.core_monitor.track_computation_time(func_name, computation_time)

        # Debug logging
        status = "HIT" if is_hit else "MISS"
        log_msg = f"Cache {status}: {func_name}"
        if cache_key:
            log_msg += f" (key: {cache_key[:8]}...)"
        if computation_time:
            if computation_time < 60:
                log_msg += f" ({computation_time:.2f}s)"
            else:
                td = pd.Timedelta(seconds=computation_time).round("1s")
                log_msg += f" ({td})".replace("0 days ", "")
        logger.debug(log_msg)


# Global monitor instance
_unified_monitor: Optional[UnifiedCacheMonitor] = None


def get_unified_monitor() -> UnifiedCacheMonitor:
    """Get global unified monitor."""
    global _unified_monitor
    if _unified_monitor is None:
        _unified_monitor = UnifiedCacheMonitor()
    return _unified_monitor


# =============================================================================
# Core: Universal Cacheable Decorator
# =============================================================================


def cacheable(
    time_aware: bool = False,
    track_data_access: bool = False,
    dataset_name: Optional[str] = None,
    purpose: Optional[str] = None,
    auto_versioning: bool = True,
):
    """
    Universal caching decorator - replaces all previous decorators.

    This ONE decorator replaces:
    - robust_cacheable
    - time_aware_cacheable
    - data_tracking_cacheable
    - cv_cacheable
    - cv_cache_with_classifier_state
    - smart_cacheable (removed - now handled by auto_versioning)

    Args:
        time_aware: Include temporal bounds in cache key
        track_data_access: Log DataFrame access for contamination detection
        dataset_name: Name for data tracking
        purpose: train/test/validate/optimize
        auto_versioning: Include function source hash in key (DEFAULT: True)
                        Set to False ONLY if:
                        - Computation takes hours/days and you want to preserve cache
                        - You're absolutely certain the function won't change
                        - You understand the risk of stale cached results

    Usage:
        # Basic - auto_versioning is DEFAULT
        @cacheable()
        def my_function(df): ...
        # Cache invalidates automatically when function changes!

        # Opt-out only for expensive, stable functions
        @cacheable(auto_versioning=False)
        def expensive_stable_function(df): ...

        # Time-aware
        @cacheable(time_aware=True)
        def my_function(df): ...

        # Data tracking
        @cacheable(track_data_access=True, dataset_name="data", purpose="train")
        def my_function(df): ...

        # CV caching - just works automatically
        @cacheable()
        def ml_cross_val_score(clf, X, y, cv_gen): ...
    """

    def decorator(func: Callable) -> Callable:
        import pickle

        func_name = f"{func.__module__}.{func.__qualname__}"

        # Warn if function is already cached
        if hasattr(func, "_afml_cacheable") and func._afml_cacheable:
            logger.warning(
                f"Function {func_name} already has @cacheable decorator. "
                f"Nested @cacheable is redundant."
            )

        cached_func = memory.cache(func)
        monitor = get_unified_monitor()

        # Track seen cache keys for this session
        seen_signatures = set()

        @wraps(func)
        def wrapper(*args, **kwargs):
            nonlocal seen_signatures

            # Generate our custom cache key for tracking/monitoring
            cache_key = UnifiedCacheKeyGenerator.generate_key(
                func,
                args,
                kwargs,
                time_aware=time_aware,
                auto_versioning=auto_versioning,
            )

            is_hit = cached_func.check_call_in_cache(*args, **kwargs)
            computation_time = None

            # Track for session
            seen_signatures.add(cache_key)

            # Execute through joblib (it handles all persistence)
            try:
                if not is_hit:
                    start_time = time.time()
                    result = cached_func(*args, **kwargs)
                    computation_time = time.time() - start_time
                else:
                    result = cached_func(*args, **kwargs)

            except (EOFError, pickle.PickleError, OSError) as e:
                # Handle cache corruption - let joblib retry
                logger.warning(
                    f"Cache corruption for {func_name}: {type(e).__name__} - recomputing"
                )

                # Clear corrupted cache
                try:
                    cached_func.clear()
                except Exception:
                    pass

                # Execute directly
                start_time = time.time()
                result = func(*args, **kwargs)
                computation_time = time.time() - start_time

            # Track stats
            monitor.track_cache_call(
                func_name=func_name,
                is_hit=is_hit,
                computation_time=computation_time,
                cache_key=cache_key,
            )

            # Track data access if requested
            if track_data_access:
                _track_data_access(args, kwargs, dataset_name, purpose)

            return result

        # Expose cache management methods
        wrapper._afml_cacheable = True
        wrapper._auto_versioning = auto_versioning
        wrapper.cache_clear = cached_func.clear
        wrapper.cache_info = lambda: {
            "function_name": func_name,
            "auto_versioning": auto_versioning,
            "seen_signatures": len(seen_signatures),
        }

        return wrapper

    return decorator


def _track_data_access(args, kwargs, dataset_name, purpose):
    """Track DataFrame access for contamination detection."""
    try:
        from .data_access_tracker import get_data_tracker

        tracker = get_data_tracker()

        # Check all arguments
        for arg in args:
            if isinstance(arg, pd.DataFrame) and isinstance(
                arg.index, pd.DatetimeIndex
            ):
                if len(arg) > 0:
                    tracker.log_access(
                        dataset_name=dataset_name or "unknown",
                        start_date=arg.index[0],
                        end_date=arg.index[-1],
                        purpose=purpose or "unknown",
                        data_shape=arg.shape,
                    )

        for key, value in kwargs.items():
            if isinstance(value, pd.DataFrame) and isinstance(
                value.index, pd.DatetimeIndex
            ):
                if len(value) > 0:
                    tracker.log_access(
                        dataset_name=dataset_name or key,
                        start_date=value.index[0],
                        end_date=value.index[-1],
                        purpose=purpose or "unknown",
                        data_shape=value.shape,
                    )
    except Exception as e:
        logger.debug(f"Data tracking failed: {e}")


# =============================================================================
# Convenience Aliases (backward compatibility) - NOW WITH AUTO_VERSIONING
# =============================================================================

# Old names → new unified decorator (with auto_versioning enabled by default)
robust_cacheable = cacheable()  # auto_versioning=True by default
time_aware_cacheable = cacheable(time_aware=True)  # auto_versioning=True by default
cv_cacheable = cacheable()  # auto_versioning=True by default


def data_tracking_cacheable(dataset_name: str, purpose: str):
    """Backward compatible data tracking decorator."""
    return cacheable(
        track_data_access=True,
        dataset_name=dataset_name,
        purpose=purpose,
        # auto_versioning=True by default
    )


# =============================================================================
# Utility: Bulk disable auto_versioning
# =============================================================================


def disable_auto_versioning():
    """
    Factory function for bulk opt-out of auto_versioning.

    Use this when you have many expensive, stable functions and want
    to explicitly opt-out of auto_versioning for all of them.

    Usage:
        cacheable_stable = disable_auto_versioning()

        @cacheable_stable()
        def expensive_function_1(data): ...

        @cacheable_stable()
        def expensive_function_2(data): ...
    """

    def _cacheable_no_versioning(**kwargs):
        # Force auto_versioning to False
        kwargs["auto_versioning"] = False
        return cacheable(**kwargs)

    return _cacheable_no_versioning


# =============================================================================
# Special: clf_hyper_fit with scipy distribution support
# =============================================================================


def create_cacheable_param_grid(param_distributions: Dict) -> Dict:
    """Convert scipy distributions to cacheable representation."""
    cacheable_params = {}

    for key, value in param_distributions.items():
        if isinstance(value, (rv_discrete_frozen, rv_continuous_frozen)):
            dist_info = (
                type(value).__name__,
                value.args if hasattr(value, "args") else (),
                value.kwds if hasattr(value, "kwds") else {},
            )
            cacheable_params[key] = dist_info
        else:
            cacheable_params[key] = value

    return cacheable_params


def reconstruct_param_grid(cacheable_params: Dict) -> Dict:
    """Reconstruct scipy distributions from cacheable representation."""
    import scipy.stats as stats
    from scipy.stats import randint, uniform

    reconstructed = {}

    for key, value in cacheable_params.items():
        if isinstance(value, tuple) and len(value) == 3:
            dist_type, args, kwds = value

            if dist_type == "rv_discrete_frozen":
                reconstructed[key] = randint(*args, **kwds)
            elif dist_type == "rv_continuous_frozen":
                reconstructed[key] = uniform(*args, **kwds)
            else:
                try:
                    dist_class = getattr(stats, dist_type.replace("_frozen", ""))
                    reconstructed[key] = dist_class(*args, **kwds)
                except Exception:
                    logger.warning(f"Could not reconstruct: {dist_type}")
                    reconstructed[key] = value
        else:
            reconstructed[key] = value

    return reconstructed


# =============================================================================
# Convenience: Print cache report
# =============================================================================


def print_cache_report():
    """Print comprehensive cache report."""
    monitor = get_unified_monitor()

    print("\n" + "=" * 70)
    print("UNIFIED CACHE REPORT")
    print("=" * 70)

    # Get health report
    report = monitor.core_monitor.generate_health_report()

    print("\nOverall:")
    print(f"  Functions: {report.total_functions}")
    print(f"  Hit Rate: {report.overall_hit_rate:.1%}")
    print(f"  Total Calls: {report.total_calls:,}")
    print(f"  Cache Size: {report.total_cache_size_mb:.1f} MB")

    if report.top_performers:
        print("\nTop Performers:")
        for i, perf in enumerate(report.top_performers[:3], 1):
            name = perf.function_name.split(".")[-1]
            print(f"  {i}. {name}: {perf.hit_rate:.1%} ({perf.total_calls} calls)")

    print("\nNote: Auto-versioning is ENABLED by default.")
    print("Cache automatically invalidates when function code changes.")
    print("=" * 70 + "\n")


__all__ = [
    # Core components
    "UnifiedCacheKeyGenerator",
    "UnifiedCacheMonitor",
    "get_unified_monitor",
    # Main decorator (replaces all others)
    "cacheable",
    # Backward compatibility aliases
    "robust_cacheable",
    "time_aware_cacheable",
    "cv_cacheable",
    "data_tracking_cacheable",
    # Utilities
    "disable_auto_versioning",
    # clf_hyper_fit support
    "reconstruct_param_grid",
    "create_cacheable_param_grid",
    # Reports
    "print_cache_report",
]