Mercurial > code > home > repos > homeauto
annotate service/mqtt_to_rdf/inference.py @ 1694:73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
https://bigasterisk.com/post/inference/2021-09-27_11-11.png
author | drewp@bigasterisk.com |
---|---|
date | Mon, 27 Sep 2021 11:22:09 -0700 |
parents | 0455a1e18e4f |
children | 5c2565e63297 |
rev | line source |
---|---|
1587 | 1 """ |
2 copied from reasoning 2021-08-29. probably same api. should | |
3 be able to lib/ this out | |
4 """ | |
1588
0757fafbfdab
WIP inferencer - partial var and function support
drewp@bigasterisk.com
parents:
1587
diff
changeset
|
5 import itertools |
1587 | 6 import logging |
1601 | 7 import time |
1594 | 8 from collections import defaultdict |
1626 | 9 from dataclasses import dataclass |
1692
2883da14847c
debugging and cleanup, as i looked for a bug
drewp@bigasterisk.com
parents:
1677
diff
changeset
|
10 from typing import Dict, Iterator, List, Optional, Tuple, Union, cast |
1694
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
11 from pathlib import Path |
1588
0757fafbfdab
WIP inferencer - partial var and function support
drewp@bigasterisk.com
parents:
1587
diff
changeset
|
12 |
1631
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
13 from prometheus_client import Histogram, Summary |
1673 | 14 from rdflib import Graph, Namespace |
1651
20474ad4968e
WIP - functions are broken as i move most layers to work in Chunks not Triples
drewp@bigasterisk.com
parents:
1650
diff
changeset
|
15 from rdflib.graph import ConjunctiveGraph |
1640 | 16 from rdflib.term import Node, URIRef, Variable |
1587 | 17 |
1694
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
18 from candidate_binding import CandidateBinding |
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
19 from inference_types import (BindingUnknown, Inconsistent, RhsBnode, RuleUnboundBnode, Triple, WorkingSetBnode) |
1651
20474ad4968e
WIP - functions are broken as i move most layers to work in Chunks not Triples
drewp@bigasterisk.com
parents:
1650
diff
changeset
|
20 from lhs_evaluation import functionsFor |
1650
2061df259224
move graphDump (on its way out, since reprs are getting better)
drewp@bigasterisk.com
parents:
1649
diff
changeset
|
21 from rdf_debug import graphDump |
1664
1a7c1261302c
logic fix- some bindings were being returned 2+; some 0 times
drewp@bigasterisk.com
parents:
1661
diff
changeset
|
22 from stmt_chunk import AlignedRuleChunk, Chunk, ChunkedGraph, applyChunky |
1694
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
23 from structured_log import StructuredLog |
1605 | 24 |
1587 | 25 log = logging.getLogger('infer') |
1664
1a7c1261302c
logic fix- some bindings were being returned 2+; some 0 times
drewp@bigasterisk.com
parents:
1661
diff
changeset
|
26 odolog = logging.getLogger('infer.odo') # the "odometer" logic |
1a7c1261302c
logic fix- some bindings were being returned 2+; some 0 times
drewp@bigasterisk.com
parents:
1661
diff
changeset
|
27 ringlog = logging.getLogger('infer.ring') # for ChunkLooper |
1a7c1261302c
logic fix- some bindings were being returned 2+; some 0 times
drewp@bigasterisk.com
parents:
1661
diff
changeset
|
28 |
1593
b0df43d5494c
big rewrite- more classes, smaller methods, more typesafe, all current tests passing
drewp@bigasterisk.com
parents:
1592
diff
changeset
|
29 INDENT = ' ' |
1587 | 30 |
1626 | 31 INFER_CALLS = Summary('inference_infer_calls', 'calls') |
32 INFER_GRAPH_SIZE = Histogram('inference_graph_size', 'statements', buckets=[2**x for x in range(2, 20, 2)]) | |
1587 | 33 |
34 ROOM = Namespace("http://projects.bigasterisk.com/room/") | |
35 LOG = Namespace('http://www.w3.org/2000/10/swap/log#') | |
36 MATH = Namespace('http://www.w3.org/2000/10/swap/math#') | |
37 | |
38 | |
1651
20474ad4968e
WIP - functions are broken as i move most layers to work in Chunks not Triples
drewp@bigasterisk.com
parents:
1650
diff
changeset
|
39 class NoOptions(ValueError): |
20474ad4968e
WIP - functions are broken as i move most layers to work in Chunks not Triples
drewp@bigasterisk.com
parents:
1650
diff
changeset
|
40 """ChunkLooper has no possibilites to add to the binding; the whole rule must therefore not apply""" |
1631
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
41 |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
42 |
1694
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
43 def debug(logger, slog: Optional[StructuredLog], msg): |
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
44 logger.debug(msg) |
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
45 if slog: |
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
46 slog.say(msg) |
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
47 |
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
48 |
1651
20474ad4968e
WIP - functions are broken as i move most layers to work in Chunks not Triples
drewp@bigasterisk.com
parents:
1650
diff
changeset
|
49 _chunkLooperShortId = itertools.count() |
1632 | 50 |
51 | |
1631
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
52 @dataclass |
1651
20474ad4968e
WIP - functions are broken as i move most layers to work in Chunks not Triples
drewp@bigasterisk.com
parents:
1650
diff
changeset
|
53 class ChunkLooper: |
20474ad4968e
WIP - functions are broken as i move most layers to work in Chunks not Triples
drewp@bigasterisk.com
parents:
1650
diff
changeset
|
54 """given one LHS Chunk, iterate through the possible matches for it, |
1632 | 55 returning what bindings they would imply. Only distinct bindings are |
1651
20474ad4968e
WIP - functions are broken as i move most layers to work in Chunks not Triples
drewp@bigasterisk.com
parents:
1650
diff
changeset
|
56 returned. The bindings build on any `prev` ChunkLooper's results. |
1632 | 57 |
1664
1a7c1261302c
logic fix- some bindings were being returned 2+; some 0 times
drewp@bigasterisk.com
parents:
1661
diff
changeset
|
58 In the odometer metaphor used below, this is one of the rings. |
1a7c1261302c
logic fix- some bindings were being returned 2+; some 0 times
drewp@bigasterisk.com
parents:
1661
diff
changeset
|
59 |
1632 | 60 This iterator is restartable.""" |
1651
20474ad4968e
WIP - functions are broken as i move most layers to work in Chunks not Triples
drewp@bigasterisk.com
parents:
1650
diff
changeset
|
61 lhsChunk: Chunk |
20474ad4968e
WIP - functions are broken as i move most layers to work in Chunks not Triples
drewp@bigasterisk.com
parents:
1650
diff
changeset
|
62 prev: Optional['ChunkLooper'] |
20474ad4968e
WIP - functions are broken as i move most layers to work in Chunks not Triples
drewp@bigasterisk.com
parents:
1650
diff
changeset
|
63 workingSet: 'ChunkedGraph' |
1694
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
64 slog: Optional[StructuredLog] |
1631
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
65 |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
66 def __repr__(self): |
1651
20474ad4968e
WIP - functions are broken as i move most layers to work in Chunks not Triples
drewp@bigasterisk.com
parents:
1650
diff
changeset
|
67 return f'{self.__class__.__name__}{self._shortId}{"<pastEnd>" if self.pastEnd() else ""}' |
1631
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
68 |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
69 def __post_init__(self): |
1651
20474ad4968e
WIP - functions are broken as i move most layers to work in Chunks not Triples
drewp@bigasterisk.com
parents:
1650
diff
changeset
|
70 self._shortId = next(_chunkLooperShortId) |
1664
1a7c1261302c
logic fix- some bindings were being returned 2+; some 0 times
drewp@bigasterisk.com
parents:
1661
diff
changeset
|
71 self._alignedMatches = list(self.lhsChunk.ruleMatchesFrom(self.workingSet)) |
1694
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
72 del self.workingSet |
1631
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
73 |
1668
89e53cb8a01c
fix some harder tests. Mostly, _advanceTheStack needed to spin the odometer rings starting from the other side, to get all the right combos
drewp@bigasterisk.com
parents:
1667
diff
changeset
|
74 # only ours- do not store prev, since it could change without us |
89e53cb8a01c
fix some harder tests. Mostly, _advanceTheStack needed to spin the odometer rings starting from the other side, to get all the right combos
drewp@bigasterisk.com
parents:
1667
diff
changeset
|
75 self._current = CandidateBinding({}) |
1694
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
76 self.currentSourceChunk: Optional[Chunk] = None # for debugging only |
1631
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
77 self._pastEnd = False |
1667 | 78 self._seenBindings: List[CandidateBinding] = [] # combined bindings (up to our ring) that we've returned |
1648 | 79 |
1664
1a7c1261302c
logic fix- some bindings were being returned 2+; some 0 times
drewp@bigasterisk.com
parents:
1661
diff
changeset
|
80 if ringlog.isEnabledFor(logging.DEBUG): |
1667 | 81 ringlog.debug('') |
1694
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
82 msg = f'{INDENT*6} introducing {self!r}({self.lhsChunk}, {self._alignedMatches=})' |
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
83 msg = msg.replace('AlignedRuleChunk', f'\n{INDENT*12}AlignedRuleChunk') |
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
84 ringlog.debug(msg) |
1648 | 85 |
1631
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
86 self.restart() |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
87 |
1635
22d481f0a924
refactor: use CandidateBinding throughout, not loose dicts
drewp@bigasterisk.com
parents:
1634
diff
changeset
|
88 def _prevBindings(self) -> CandidateBinding: |
1631
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
89 if not self.prev or self.prev.pastEnd(): |
1635
22d481f0a924
refactor: use CandidateBinding throughout, not loose dicts
drewp@bigasterisk.com
parents:
1634
diff
changeset
|
90 return CandidateBinding({}) |
1631
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
91 |
1635
22d481f0a924
refactor: use CandidateBinding throughout, not loose dicts
drewp@bigasterisk.com
parents:
1634
diff
changeset
|
92 return self.prev.currentBinding() |
1631
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
93 |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
94 def advance(self): |
1668
89e53cb8a01c
fix some harder tests. Mostly, _advanceTheStack needed to spin the odometer rings starting from the other side, to get all the right combos
drewp@bigasterisk.com
parents:
1667
diff
changeset
|
95 """update _current to a new set of valid bindings we haven't seen (since |
89e53cb8a01c
fix some harder tests. Mostly, _advanceTheStack needed to spin the odometer rings starting from the other side, to get all the right combos
drewp@bigasterisk.com
parents:
1667
diff
changeset
|
96 last restart), or go into pastEnd mode. Note that _current is just our |
89e53cb8a01c
fix some harder tests. Mostly, _advanceTheStack needed to spin the odometer rings starting from the other side, to get all the right combos
drewp@bigasterisk.com
parents:
1667
diff
changeset
|
97 contribution, but returned valid bindings include all prev rings.""" |
1633
6107603ed455
fix farenheit rule case, fix some others that depend on rings order, but this breaks some performance because of itertools.perm
drewp@bigasterisk.com
parents:
1632
diff
changeset
|
98 if self._pastEnd: |
6107603ed455
fix farenheit rule case, fix some others that depend on rings order, but this breaks some performance because of itertools.perm
drewp@bigasterisk.com
parents:
1632
diff
changeset
|
99 raise NotImplementedError('need restart') |
1664
1a7c1261302c
logic fix- some bindings were being returned 2+; some 0 times
drewp@bigasterisk.com
parents:
1661
diff
changeset
|
100 ringlog.debug('') |
1694
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
101 debug(ringlog, self.slog, f'{INDENT*6} --> {self}.advance start:') |
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
102 |
1693
0455a1e18e4f
really rough fix to a broken test. hopefully this gets redone
drewp@bigasterisk.com
parents:
1692
diff
changeset
|
103 self._currentIsFromFunc = None |
1664
1a7c1261302c
logic fix- some bindings were being returned 2+; some 0 times
drewp@bigasterisk.com
parents:
1661
diff
changeset
|
104 augmentedWorkingSet: List[AlignedRuleChunk] = [] |
1633
6107603ed455
fix farenheit rule case, fix some others that depend on rings order, but this breaks some performance because of itertools.perm
drewp@bigasterisk.com
parents:
1632
diff
changeset
|
105 if self.prev is None: |
1664
1a7c1261302c
logic fix- some bindings were being returned 2+; some 0 times
drewp@bigasterisk.com
parents:
1661
diff
changeset
|
106 augmentedWorkingSet = self._alignedMatches |
1633
6107603ed455
fix farenheit rule case, fix some others that depend on rings order, but this breaks some performance because of itertools.perm
drewp@bigasterisk.com
parents:
1632
diff
changeset
|
107 else: |
1673 | 108 augmentedWorkingSet = list(applyChunky(self.prev.currentBinding(), self._alignedMatches)) |
1633
6107603ed455
fix farenheit rule case, fix some others that depend on rings order, but this breaks some performance because of itertools.perm
drewp@bigasterisk.com
parents:
1632
diff
changeset
|
109 |
1634
ba59cfc3c747
hack math:sum in there. Test suite is passing except some slow performers
drewp@bigasterisk.com
parents:
1633
diff
changeset
|
110 if self._advanceWithPlainMatches(augmentedWorkingSet): |
1694
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
111 debug(ringlog, self.slog, f'{INDENT*6} <-- {self}.advance finished with plain matches') |
1634
ba59cfc3c747
hack math:sum in there. Test suite is passing except some slow performers
drewp@bigasterisk.com
parents:
1633
diff
changeset
|
112 return |
ba59cfc3c747
hack math:sum in there. Test suite is passing except some slow performers
drewp@bigasterisk.com
parents:
1633
diff
changeset
|
113 |
1639 | 114 if self._advanceWithFunctions(): |
1694
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
115 debug(ringlog, self.slog, f'{INDENT*6} <-- {self}.advance finished with function matches') |
1634
ba59cfc3c747
hack math:sum in there. Test suite is passing except some slow performers
drewp@bigasterisk.com
parents:
1633
diff
changeset
|
116 return |
ba59cfc3c747
hack math:sum in there. Test suite is passing except some slow performers
drewp@bigasterisk.com
parents:
1633
diff
changeset
|
117 |
1694
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
118 debug(ringlog, self.slog, f'{INDENT*6} <-- {self}.advance had nothing and is now past end') |
1634
ba59cfc3c747
hack math:sum in there. Test suite is passing except some slow performers
drewp@bigasterisk.com
parents:
1633
diff
changeset
|
119 self._pastEnd = True |
ba59cfc3c747
hack math:sum in there. Test suite is passing except some slow performers
drewp@bigasterisk.com
parents:
1633
diff
changeset
|
120 |
1664
1a7c1261302c
logic fix- some bindings were being returned 2+; some 0 times
drewp@bigasterisk.com
parents:
1661
diff
changeset
|
121 def _advanceWithPlainMatches(self, augmentedWorkingSet: List[AlignedRuleChunk]) -> bool: |
1694
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
122 # if augmentedWorkingSet: |
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
123 # debug(ringlog, self.slog, f'{INDENT*7} {self} mines {len(augmentedWorkingSet)} matching augmented statements') |
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
124 # for s in augmentedWorkingSet: |
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
125 # debug(ringlog, self.slog, f'{INDENT*8} {s}') |
1633
6107603ed455
fix farenheit rule case, fix some others that depend on rings order, but this breaks some performance because of itertools.perm
drewp@bigasterisk.com
parents:
1632
diff
changeset
|
126 |
1664
1a7c1261302c
logic fix- some bindings were being returned 2+; some 0 times
drewp@bigasterisk.com
parents:
1661
diff
changeset
|
127 for aligned in augmentedWorkingSet: |
1631
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
128 try: |
1668
89e53cb8a01c
fix some harder tests. Mostly, _advanceTheStack needed to spin the odometer rings starting from the other side, to get all the right combos
drewp@bigasterisk.com
parents:
1667
diff
changeset
|
129 newBinding = aligned.newBindingIfMatched(self._prevBindings()) |
89e53cb8a01c
fix some harder tests. Mostly, _advanceTheStack needed to spin the odometer rings starting from the other side, to get all the right combos
drewp@bigasterisk.com
parents:
1667
diff
changeset
|
130 except Inconsistent as exc: |
1694
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
131 debug(ringlog, self.slog, |
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
132 f'{INDENT*7} ChunkLooper{self._shortId} - {aligned} would be inconsistent with prev bindings ({exc})') |
1631
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
133 continue |
1632 | 134 |
1694
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
135 if self._testAndKeepNewBinding(newBinding, aligned.workingSetChunk): |
1634
ba59cfc3c747
hack math:sum in there. Test suite is passing except some slow performers
drewp@bigasterisk.com
parents:
1633
diff
changeset
|
136 return True |
ba59cfc3c747
hack math:sum in there. Test suite is passing except some slow performers
drewp@bigasterisk.com
parents:
1633
diff
changeset
|
137 return False |
1631
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
138 |
1639 | 139 def _advanceWithFunctions(self) -> bool: |
1651
20474ad4968e
WIP - functions are broken as i move most layers to work in Chunks not Triples
drewp@bigasterisk.com
parents:
1650
diff
changeset
|
140 pred: Node = self.lhsChunk.predicate |
1640 | 141 if not isinstance(pred, URIRef): |
142 raise NotImplementedError | |
1632 | 143 |
1637 | 144 for functionType in functionsFor(pred): |
1661 | 145 fn = functionType(self.lhsChunk) |
1694
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
146 # debug(ringlog, self.slog, f'{INDENT*7} ChunkLooper{self._shortId} advanceWithFunctions, {functionType=}') |
1652 | 147 |
1634
ba59cfc3c747
hack math:sum in there. Test suite is passing except some slow performers
drewp@bigasterisk.com
parents:
1633
diff
changeset
|
148 try: |
1692
2883da14847c
debugging and cleanup, as i looked for a bug
drewp@bigasterisk.com
parents:
1677
diff
changeset
|
149 log.debug(f'fn.bind {self._prevBindings()} ...') |
2883da14847c
debugging and cleanup, as i looked for a bug
drewp@bigasterisk.com
parents:
1677
diff
changeset
|
150 #fullBinding = self._prevBindings().copy() |
1664
1a7c1261302c
logic fix- some bindings were being returned 2+; some 0 times
drewp@bigasterisk.com
parents:
1661
diff
changeset
|
151 newBinding = fn.bind(self._prevBindings()) |
1692
2883da14847c
debugging and cleanup, as i looked for a bug
drewp@bigasterisk.com
parents:
1677
diff
changeset
|
152 log.debug(f'...makes {newBinding=}') |
1637 | 153 except BindingUnknown: |
1634
ba59cfc3c747
hack math:sum in there. Test suite is passing except some slow performers
drewp@bigasterisk.com
parents:
1633
diff
changeset
|
154 pass |
ba59cfc3c747
hack math:sum in there. Test suite is passing except some slow performers
drewp@bigasterisk.com
parents:
1633
diff
changeset
|
155 else: |
1664
1a7c1261302c
logic fix- some bindings were being returned 2+; some 0 times
drewp@bigasterisk.com
parents:
1661
diff
changeset
|
156 if newBinding is not None: |
1693
0455a1e18e4f
really rough fix to a broken test. hopefully this gets redone
drewp@bigasterisk.com
parents:
1692
diff
changeset
|
157 self._currentIsFromFunc = fn |
1694
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
158 if self._testAndKeepNewBinding(newBinding, self.lhsChunk): |
1637 | 159 return True |
1634
ba59cfc3c747
hack math:sum in there. Test suite is passing except some slow performers
drewp@bigasterisk.com
parents:
1633
diff
changeset
|
160 |
ba59cfc3c747
hack math:sum in there. Test suite is passing except some slow performers
drewp@bigasterisk.com
parents:
1633
diff
changeset
|
161 return False |
ba59cfc3c747
hack math:sum in there. Test suite is passing except some slow performers
drewp@bigasterisk.com
parents:
1633
diff
changeset
|
162 |
1694
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
163 def _testAndKeepNewBinding(self, newBinding: CandidateBinding, sourceChunk: Chunk): |
1668
89e53cb8a01c
fix some harder tests. Mostly, _advanceTheStack needed to spin the odometer rings starting from the other side, to get all the right combos
drewp@bigasterisk.com
parents:
1667
diff
changeset
|
164 fullBinding: CandidateBinding = self._prevBindings().copy() |
89e53cb8a01c
fix some harder tests. Mostly, _advanceTheStack needed to spin the odometer rings starting from the other side, to get all the right combos
drewp@bigasterisk.com
parents:
1667
diff
changeset
|
165 fullBinding.addNewBindings(newBinding) |
89e53cb8a01c
fix some harder tests. Mostly, _advanceTheStack needed to spin the odometer rings starting from the other side, to get all the right combos
drewp@bigasterisk.com
parents:
1667
diff
changeset
|
166 isNew = fullBinding not in self._seenBindings |
89e53cb8a01c
fix some harder tests. Mostly, _advanceTheStack needed to spin the odometer rings starting from the other side, to get all the right combos
drewp@bigasterisk.com
parents:
1667
diff
changeset
|
167 ringlog.debug(f'{INDENT*7} {self} considering {newBinding=} to make {fullBinding}. {isNew=}') |
1694
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
168 # if self.slog: |
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
169 # self.slog.looperConsider(self, newBinding, fullBinding, isNew) |
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
170 |
1668
89e53cb8a01c
fix some harder tests. Mostly, _advanceTheStack needed to spin the odometer rings starting from the other side, to get all the right combos
drewp@bigasterisk.com
parents:
1667
diff
changeset
|
171 if isNew: |
89e53cb8a01c
fix some harder tests. Mostly, _advanceTheStack needed to spin the odometer rings starting from the other side, to get all the right combos
drewp@bigasterisk.com
parents:
1667
diff
changeset
|
172 self._seenBindings.append(fullBinding.copy()) |
89e53cb8a01c
fix some harder tests. Mostly, _advanceTheStack needed to spin the odometer rings starting from the other side, to get all the right combos
drewp@bigasterisk.com
parents:
1667
diff
changeset
|
173 self._current = newBinding |
1694
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
174 self.currentSourceChunk = sourceChunk |
1668
89e53cb8a01c
fix some harder tests. Mostly, _advanceTheStack needed to spin the odometer rings starting from the other side, to get all the right combos
drewp@bigasterisk.com
parents:
1667
diff
changeset
|
175 return True |
89e53cb8a01c
fix some harder tests. Mostly, _advanceTheStack needed to spin the odometer rings starting from the other side, to get all the right combos
drewp@bigasterisk.com
parents:
1667
diff
changeset
|
176 return False |
89e53cb8a01c
fix some harder tests. Mostly, _advanceTheStack needed to spin the odometer rings starting from the other side, to get all the right combos
drewp@bigasterisk.com
parents:
1667
diff
changeset
|
177 |
1694
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
178 def localBinding(self) -> CandidateBinding: |
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
179 if self.pastEnd(): |
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
180 raise NotImplementedError() |
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
181 return self._current |
1631
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
182 |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
183 def currentBinding(self) -> CandidateBinding: |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
184 if self.pastEnd(): |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
185 raise NotImplementedError() |
1664
1a7c1261302c
logic fix- some bindings were being returned 2+; some 0 times
drewp@bigasterisk.com
parents:
1661
diff
changeset
|
186 together = self._prevBindings().copy() |
1a7c1261302c
logic fix- some bindings were being returned 2+; some 0 times
drewp@bigasterisk.com
parents:
1661
diff
changeset
|
187 together.addNewBindings(self._current) |
1a7c1261302c
logic fix- some bindings were being returned 2+; some 0 times
drewp@bigasterisk.com
parents:
1661
diff
changeset
|
188 return together |
1631
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
189 |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
190 def pastEnd(self) -> bool: |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
191 return self._pastEnd |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
192 |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
193 def restart(self): |
1694
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
194 try: |
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
195 self._pastEnd = False |
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
196 self._seenBindings = [] |
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
197 self.advance() |
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
198 if self.pastEnd(): |
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
199 raise NoOptions() |
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
200 finally: |
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
201 debug(ringlog, self.slog, f'{INDENT*7} ChunkLooper{self._shortId} restarts: pastEnd={self.pastEnd()}') |
1631
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
202 |
1693
0455a1e18e4f
really rough fix to a broken test. hopefully this gets redone
drewp@bigasterisk.com
parents:
1692
diff
changeset
|
203 def prevMayHaveChanged(self): |
1694
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
204 pass |
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
205 # self._advanceWithFunctions() |
1693
0455a1e18e4f
really rough fix to a broken test. hopefully this gets redone
drewp@bigasterisk.com
parents:
1692
diff
changeset
|
206 # This is a total patch for a test failure. This should be generalized |
0455a1e18e4f
really rough fix to a broken test. hopefully this gets redone
drewp@bigasterisk.com
parents:
1692
diff
changeset
|
207 # to a Looper that can keep itself correct when prev changes. |
1694
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
208 # if self.pastEnd(): |
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
209 # self.restart() |
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
210 # else: |
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
211 # self.advance() |
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
212 # if self._currentIsFromFunc: |
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
213 # self._advanceWithFunctions() |
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
214 # if self.pastEnd(): |
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
215 # self.restart() |
1693
0455a1e18e4f
really rough fix to a broken test. hopefully this gets redone
drewp@bigasterisk.com
parents:
1692
diff
changeset
|
216 |
1631
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
217 |
1594 | 218 @dataclass |
1593
b0df43d5494c
big rewrite- more classes, smaller methods, more typesafe, all current tests passing
drewp@bigasterisk.com
parents:
1592
diff
changeset
|
219 class Lhs: |
1651
20474ad4968e
WIP - functions are broken as i move most layers to work in Chunks not Triples
drewp@bigasterisk.com
parents:
1650
diff
changeset
|
220 graph: ChunkedGraph # our full LHS graph, as input. See below for the statements partitioned into groups. |
1594 | 221 |
222 def __post_init__(self): | |
1648 | 223 |
1651
20474ad4968e
WIP - functions are broken as i move most layers to work in Chunks not Triples
drewp@bigasterisk.com
parents:
1650
diff
changeset
|
224 self.myPreds = self.graph.allPredicatesExceptFunctions() |
1602 | 225 |
1609
34f2817320cc
new tests for a smaller part of the inner loop
drewp@bigasterisk.com
parents:
1608
diff
changeset
|
226 def __repr__(self): |
1651
20474ad4968e
WIP - functions are broken as i move most layers to work in Chunks not Triples
drewp@bigasterisk.com
parents:
1650
diff
changeset
|
227 return f"Lhs({self.graph!r})" |
1694
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
228 def findCandidateBindings(self, knownTrue: ChunkedGraph, stats, slog: Optional[StructuredLog], |
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
229 ruleStatementsIterationLimit) -> Iterator['BoundLhs']: |
1664
1a7c1261302c
logic fix- some bindings were being returned 2+; some 0 times
drewp@bigasterisk.com
parents:
1661
diff
changeset
|
230 """distinct bindings that fit the LHS of a rule, using statements from |
1a7c1261302c
logic fix- some bindings were being returned 2+; some 0 times
drewp@bigasterisk.com
parents:
1661
diff
changeset
|
231 workingSet and functions from LHS""" |
1651
20474ad4968e
WIP - functions are broken as i move most layers to work in Chunks not Triples
drewp@bigasterisk.com
parents:
1650
diff
changeset
|
232 if not self.graph: |
1633
6107603ed455
fix farenheit rule case, fix some others that depend on rings order, but this breaks some performance because of itertools.perm
drewp@bigasterisk.com
parents:
1632
diff
changeset
|
233 # special case- no LHS! |
6107603ed455
fix farenheit rule case, fix some others that depend on rings order, but this breaks some performance because of itertools.perm
drewp@bigasterisk.com
parents:
1632
diff
changeset
|
234 yield BoundLhs(self, CandidateBinding({})) |
6107603ed455
fix farenheit rule case, fix some others that depend on rings order, but this breaks some performance because of itertools.perm
drewp@bigasterisk.com
parents:
1632
diff
changeset
|
235 return |
6107603ed455
fix farenheit rule case, fix some others that depend on rings order, but this breaks some performance because of itertools.perm
drewp@bigasterisk.com
parents:
1632
diff
changeset
|
236 |
1640 | 237 if self._checkPredicateCounts(knownTrue): |
238 stats['_checkPredicateCountsCulls'] += 1 | |
239 return | |
240 | |
1651
20474ad4968e
WIP - functions are broken as i move most layers to work in Chunks not Triples
drewp@bigasterisk.com
parents:
1650
diff
changeset
|
241 if not all(ch in knownTrue for ch in self.graph.staticChunks): |
1648 | 242 stats['staticStmtCulls'] += 1 |
243 return | |
1664
1a7c1261302c
logic fix- some bindings were being returned 2+; some 0 times
drewp@bigasterisk.com
parents:
1661
diff
changeset
|
244 # After this point we don't need to consider self.graph.staticChunks. |
1648 | 245 |
1656
d8e5b2232474
bugfix for simple (no arg lists) functions not evaluating
drewp@bigasterisk.com
parents:
1653
diff
changeset
|
246 if not self.graph.patternChunks and not self.graph.chunksUsedByFuncs: |
1648 | 247 # static only |
248 yield BoundLhs(self, CandidateBinding({})) | |
249 return | |
250 | |
1664
1a7c1261302c
logic fix- some bindings were being returned 2+; some 0 times
drewp@bigasterisk.com
parents:
1661
diff
changeset
|
251 log.debug('') |
1631
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
252 try: |
1694
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
253 chunkStack = self._assembleRings(knownTrue, stats, slog) |
1631
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
254 except NoOptions: |
1664
1a7c1261302c
logic fix- some bindings were being returned 2+; some 0 times
drewp@bigasterisk.com
parents:
1661
diff
changeset
|
255 ringlog.debug(f'{INDENT*5} start up with no options; 0 bindings') |
1593
b0df43d5494c
big rewrite- more classes, smaller methods, more typesafe, all current tests passing
drewp@bigasterisk.com
parents:
1592
diff
changeset
|
256 return |
1692
2883da14847c
debugging and cleanup, as i looked for a bug
drewp@bigasterisk.com
parents:
1677
diff
changeset
|
257 log.debug('') |
2883da14847c
debugging and cleanup, as i looked for a bug
drewp@bigasterisk.com
parents:
1677
diff
changeset
|
258 log.debug('') |
1664
1a7c1261302c
logic fix- some bindings were being returned 2+; some 0 times
drewp@bigasterisk.com
parents:
1661
diff
changeset
|
259 self._debugChunkStack('time to spin: initial odometer is', chunkStack) |
1694
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
260 |
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
261 if slog: |
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
262 slog.say('time to spin') |
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
263 slog.odometer(chunkStack) |
1651
20474ad4968e
WIP - functions are broken as i move most layers to work in Chunks not Triples
drewp@bigasterisk.com
parents:
1650
diff
changeset
|
264 self._assertAllRingsAreValid(chunkStack) |
1631
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
265 |
1651
20474ad4968e
WIP - functions are broken as i move most layers to work in Chunks not Triples
drewp@bigasterisk.com
parents:
1650
diff
changeset
|
266 lastRing = chunkStack[-1] |
1631
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
267 iterCount = 0 |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
268 while True: |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
269 iterCount += 1 |
1648 | 270 if iterCount > ruleStatementsIterationLimit: |
271 raise ValueError('rule too complex') | |
1631
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
272 |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
273 log.debug(f'{INDENT*4} vv findCandBindings iteration {iterCount}') |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
274 |
1633
6107603ed455
fix farenheit rule case, fix some others that depend on rings order, but this breaks some performance because of itertools.perm
drewp@bigasterisk.com
parents:
1632
diff
changeset
|
275 yield BoundLhs(self, lastRing.currentBinding()) |
1631
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
276 |
1667 | 277 # self._debugChunkStack('odometer', chunkStack) |
1607
b21885181e35
more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents:
1605
diff
changeset
|
278 |
1664
1a7c1261302c
logic fix- some bindings were being returned 2+; some 0 times
drewp@bigasterisk.com
parents:
1661
diff
changeset
|
279 done = self._advanceTheStack(chunkStack) |
1631
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
280 |
1651
20474ad4968e
WIP - functions are broken as i move most layers to work in Chunks not Triples
drewp@bigasterisk.com
parents:
1650
diff
changeset
|
281 self._debugChunkStack(f'odometer after ({done=})', chunkStack) |
1694
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
282 if slog: |
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
283 slog.odometer(chunkStack) |
1631
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
284 |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
285 log.debug(f'{INDENT*4} ^^ findCandBindings iteration done') |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
286 if done: |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
287 break |
1607
b21885181e35
more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents:
1605
diff
changeset
|
288 |
1651
20474ad4968e
WIP - functions are broken as i move most layers to work in Chunks not Triples
drewp@bigasterisk.com
parents:
1650
diff
changeset
|
289 def _debugChunkStack(self, label: str, chunkStack: List[ChunkLooper]): |
1664
1a7c1261302c
logic fix- some bindings were being returned 2+; some 0 times
drewp@bigasterisk.com
parents:
1661
diff
changeset
|
290 odolog.debug(f'{INDENT*4} {label}:') |
1652 | 291 for i, l in enumerate(chunkStack): |
1694
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
292 odolog.debug(f'{INDENT*5} [{i}] {l} curbind={l.localBinding() if not l.pastEnd() else "<end>"}') |
1632 | 293 |
1640 | 294 def _checkPredicateCounts(self, knownTrue): |
295 """raise NoOptions quickly in some cases""" | |
1648 | 296 |
1651
20474ad4968e
WIP - functions are broken as i move most layers to work in Chunks not Triples
drewp@bigasterisk.com
parents:
1650
diff
changeset
|
297 if self.graph.noPredicatesAppear(self.myPreds): |
1652 | 298 log.debug(f'{INDENT*3} checkPredicateCounts does cull because not all {self.myPreds=} are in knownTrue') |
1640 | 299 return True |
1652 | 300 log.debug(f'{INDENT*3} checkPredicateCounts does not cull because all {self.myPreds=} are in knownTrue') |
1640 | 301 return False |
302 | |
1694
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
303 def _assembleRings(self, knownTrue: ChunkedGraph, stats, slog) -> List[ChunkLooper]: |
1651
20474ad4968e
WIP - functions are broken as i move most layers to work in Chunks not Triples
drewp@bigasterisk.com
parents:
1650
diff
changeset
|
304 """make ChunkLooper for each stmt in our LHS graph, but do it in a way that they all |
1652 | 305 start out valid (or else raise NoOptions). static chunks have already been confirmed.""" |
1633
6107603ed455
fix farenheit rule case, fix some others that depend on rings order, but this breaks some performance because of itertools.perm
drewp@bigasterisk.com
parents:
1632
diff
changeset
|
306 |
1664
1a7c1261302c
logic fix- some bindings were being returned 2+; some 0 times
drewp@bigasterisk.com
parents:
1661
diff
changeset
|
307 log.debug(f'{INDENT*4} stats={dict(stats)}') |
1a7c1261302c
logic fix- some bindings were being returned 2+; some 0 times
drewp@bigasterisk.com
parents:
1661
diff
changeset
|
308 odolog.debug(f'{INDENT*3} build new ChunkLooper stack') |
1a7c1261302c
logic fix- some bindings were being returned 2+; some 0 times
drewp@bigasterisk.com
parents:
1661
diff
changeset
|
309 chunks = list(self.graph.patternChunks.union(self.graph.chunksUsedByFuncs)) |
1a7c1261302c
logic fix- some bindings were being returned 2+; some 0 times
drewp@bigasterisk.com
parents:
1661
diff
changeset
|
310 chunks.sort(key=None) |
1a7c1261302c
logic fix- some bindings were being returned 2+; some 0 times
drewp@bigasterisk.com
parents:
1661
diff
changeset
|
311 odolog.info(f' {INDENT*3} taking permutations of {len(chunks)=}') |
1692
2883da14847c
debugging and cleanup, as i looked for a bug
drewp@bigasterisk.com
parents:
1677
diff
changeset
|
312 |
2883da14847c
debugging and cleanup, as i looked for a bug
drewp@bigasterisk.com
parents:
1677
diff
changeset
|
313 permsTried = 0 |
2883da14847c
debugging and cleanup, as i looked for a bug
drewp@bigasterisk.com
parents:
1677
diff
changeset
|
314 |
2883da14847c
debugging and cleanup, as i looked for a bug
drewp@bigasterisk.com
parents:
1677
diff
changeset
|
315 for perm in self._partitionedGraphPermutations(): |
1664
1a7c1261302c
logic fix- some bindings were being returned 2+; some 0 times
drewp@bigasterisk.com
parents:
1661
diff
changeset
|
316 looperRings: List[ChunkLooper] = [] |
1651
20474ad4968e
WIP - functions are broken as i move most layers to work in Chunks not Triples
drewp@bigasterisk.com
parents:
1650
diff
changeset
|
317 prev: Optional[ChunkLooper] = None |
1664
1a7c1261302c
logic fix- some bindings were being returned 2+; some 0 times
drewp@bigasterisk.com
parents:
1661
diff
changeset
|
318 if odolog.isEnabledFor(logging.DEBUG): |
1692
2883da14847c
debugging and cleanup, as i looked for a bug
drewp@bigasterisk.com
parents:
1677
diff
changeset
|
319 odolog.debug( |
2883da14847c
debugging and cleanup, as i looked for a bug
drewp@bigasterisk.com
parents:
1677
diff
changeset
|
320 f'{INDENT*4} [perm {permsTried}] try rule chunks in this order: {" THEN ".join(repr(p) for p in perm)}') |
1633
6107603ed455
fix farenheit rule case, fix some others that depend on rings order, but this breaks some performance because of itertools.perm
drewp@bigasterisk.com
parents:
1632
diff
changeset
|
321 |
1664
1a7c1261302c
logic fix- some bindings were being returned 2+; some 0 times
drewp@bigasterisk.com
parents:
1661
diff
changeset
|
322 for ruleChunk in perm: |
1633
6107603ed455
fix farenheit rule case, fix some others that depend on rings order, but this breaks some performance because of itertools.perm
drewp@bigasterisk.com
parents:
1632
diff
changeset
|
323 try: |
1661 | 324 # These are getting rebuilt a lot which takes time. It would |
325 # be nice if they could accept a changing `prev` order | |
326 # (which might already be ok). | |
1694
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
327 looper = ChunkLooper(ruleChunk, prev, knownTrue, slog) |
1633
6107603ed455
fix farenheit rule case, fix some others that depend on rings order, but this breaks some performance because of itertools.perm
drewp@bigasterisk.com
parents:
1632
diff
changeset
|
328 except NoOptions: |
1664
1a7c1261302c
logic fix- some bindings were being returned 2+; some 0 times
drewp@bigasterisk.com
parents:
1661
diff
changeset
|
329 odolog.debug(f'{INDENT*5} permutation didnt work, try another') |
1633
6107603ed455
fix farenheit rule case, fix some others that depend on rings order, but this breaks some performance because of itertools.perm
drewp@bigasterisk.com
parents:
1632
diff
changeset
|
330 break |
1664
1a7c1261302c
logic fix- some bindings were being returned 2+; some 0 times
drewp@bigasterisk.com
parents:
1661
diff
changeset
|
331 looperRings.append(looper) |
1a7c1261302c
logic fix- some bindings were being returned 2+; some 0 times
drewp@bigasterisk.com
parents:
1661
diff
changeset
|
332 prev = looperRings[-1] |
1633
6107603ed455
fix farenheit rule case, fix some others that depend on rings order, but this breaks some performance because of itertools.perm
drewp@bigasterisk.com
parents:
1632
diff
changeset
|
333 else: |
1664
1a7c1261302c
logic fix- some bindings were being returned 2+; some 0 times
drewp@bigasterisk.com
parents:
1661
diff
changeset
|
334 # bug: At this point we've only shown that these are valid |
1a7c1261302c
logic fix- some bindings were being returned 2+; some 0 times
drewp@bigasterisk.com
parents:
1661
diff
changeset
|
335 # starting rings. The rules might be tricky enough that this |
1a7c1261302c
logic fix- some bindings were being returned 2+; some 0 times
drewp@bigasterisk.com
parents:
1661
diff
changeset
|
336 # permutation won't get us to the solution. |
1a7c1261302c
logic fix- some bindings were being returned 2+; some 0 times
drewp@bigasterisk.com
parents:
1661
diff
changeset
|
337 return looperRings |
1692
2883da14847c
debugging and cleanup, as i looked for a bug
drewp@bigasterisk.com
parents:
1677
diff
changeset
|
338 if permsTried > 50000: |
1652 | 339 raise NotImplementedError(f'trying too many permutations {len(chunks)=}') |
1692
2883da14847c
debugging and cleanup, as i looked for a bug
drewp@bigasterisk.com
parents:
1677
diff
changeset
|
340 permsTried += 1 |
1648 | 341 |
1664
1a7c1261302c
logic fix- some bindings were being returned 2+; some 0 times
drewp@bigasterisk.com
parents:
1661
diff
changeset
|
342 odolog.debug(f'{INDENT*5} no perms worked- rule cannot match anything') |
1633
6107603ed455
fix farenheit rule case, fix some others that depend on rings order, but this breaks some performance because of itertools.perm
drewp@bigasterisk.com
parents:
1632
diff
changeset
|
343 raise NoOptions() |
6107603ed455
fix farenheit rule case, fix some others that depend on rings order, but this breaks some performance because of itertools.perm
drewp@bigasterisk.com
parents:
1632
diff
changeset
|
344 |
1692
2883da14847c
debugging and cleanup, as i looked for a bug
drewp@bigasterisk.com
parents:
1677
diff
changeset
|
345 def _unpartitionedGraphPermutations(self) -> Iterator[Tuple[Chunk, ...]]: |
2883da14847c
debugging and cleanup, as i looked for a bug
drewp@bigasterisk.com
parents:
1677
diff
changeset
|
346 for perm in itertools.permutations(sorted(list(self.graph.patternChunks.union(self.graph.chunksUsedByFuncs)))): |
2883da14847c
debugging and cleanup, as i looked for a bug
drewp@bigasterisk.com
parents:
1677
diff
changeset
|
347 yield perm |
2883da14847c
debugging and cleanup, as i looked for a bug
drewp@bigasterisk.com
parents:
1677
diff
changeset
|
348 |
2883da14847c
debugging and cleanup, as i looked for a bug
drewp@bigasterisk.com
parents:
1677
diff
changeset
|
349 def _partitionedGraphPermutations(self) -> Iterator[Tuple[Chunk, ...]]: |
2883da14847c
debugging and cleanup, as i looked for a bug
drewp@bigasterisk.com
parents:
1677
diff
changeset
|
350 """always puts function chunks after pattern chunks |
2883da14847c
debugging and cleanup, as i looked for a bug
drewp@bigasterisk.com
parents:
1677
diff
changeset
|
351 |
2883da14847c
debugging and cleanup, as i looked for a bug
drewp@bigasterisk.com
parents:
1677
diff
changeset
|
352 (and, if we cared, static chunks could go before that. Currently they're |
2883da14847c
debugging and cleanup, as i looked for a bug
drewp@bigasterisk.com
parents:
1677
diff
changeset
|
353 culled out elsewhere, but that's done as a special case) |
2883da14847c
debugging and cleanup, as i looked for a bug
drewp@bigasterisk.com
parents:
1677
diff
changeset
|
354 """ |
2883da14847c
debugging and cleanup, as i looked for a bug
drewp@bigasterisk.com
parents:
1677
diff
changeset
|
355 tupleOfNoChunks: Tuple[Chunk, ...] = () |
2883da14847c
debugging and cleanup, as i looked for a bug
drewp@bigasterisk.com
parents:
1677
diff
changeset
|
356 pats = sorted(self.graph.patternChunks) |
2883da14847c
debugging and cleanup, as i looked for a bug
drewp@bigasterisk.com
parents:
1677
diff
changeset
|
357 funcs = sorted(self.graph.chunksUsedByFuncs) |
2883da14847c
debugging and cleanup, as i looked for a bug
drewp@bigasterisk.com
parents:
1677
diff
changeset
|
358 for patternPart in itertools.permutations(pats) if pats else [tupleOfNoChunks]: |
2883da14847c
debugging and cleanup, as i looked for a bug
drewp@bigasterisk.com
parents:
1677
diff
changeset
|
359 for funcPart in itertools.permutations(funcs) if funcs else [tupleOfNoChunks]: |
2883da14847c
debugging and cleanup, as i looked for a bug
drewp@bigasterisk.com
parents:
1677
diff
changeset
|
360 perm = patternPart + funcPart |
2883da14847c
debugging and cleanup, as i looked for a bug
drewp@bigasterisk.com
parents:
1677
diff
changeset
|
361 yield perm |
2883da14847c
debugging and cleanup, as i looked for a bug
drewp@bigasterisk.com
parents:
1677
diff
changeset
|
362 |
1664
1a7c1261302c
logic fix- some bindings were being returned 2+; some 0 times
drewp@bigasterisk.com
parents:
1661
diff
changeset
|
363 def _advanceTheStack(self, looperRings: List[ChunkLooper]) -> bool: |
1693
0455a1e18e4f
really rough fix to a broken test. hopefully this gets redone
drewp@bigasterisk.com
parents:
1692
diff
changeset
|
364 |
0455a1e18e4f
really rough fix to a broken test. hopefully this gets redone
drewp@bigasterisk.com
parents:
1692
diff
changeset
|
365 def freshenRight(i): |
0455a1e18e4f
really rough fix to a broken test. hopefully this gets redone
drewp@bigasterisk.com
parents:
1692
diff
changeset
|
366 for ring in looperRings[i + 1:]: |
0455a1e18e4f
really rough fix to a broken test. hopefully this gets redone
drewp@bigasterisk.com
parents:
1692
diff
changeset
|
367 ring.prevMayHaveChanged() |
0455a1e18e4f
really rough fix to a broken test. hopefully this gets redone
drewp@bigasterisk.com
parents:
1692
diff
changeset
|
368 |
1668
89e53cb8a01c
fix some harder tests. Mostly, _advanceTheStack needed to spin the odometer rings starting from the other side, to get all the right combos
drewp@bigasterisk.com
parents:
1667
diff
changeset
|
369 carry = True # last elem always must advance |
89e53cb8a01c
fix some harder tests. Mostly, _advanceTheStack needed to spin the odometer rings starting from the other side, to get all the right combos
drewp@bigasterisk.com
parents:
1667
diff
changeset
|
370 for i, ring in reversed(list(enumerate(looperRings))): |
1631
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
371 # unlike normal odometer, advancing any earlier ring could invalidate later ones |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
372 if carry: |
1692
2883da14847c
debugging and cleanup, as i looked for a bug
drewp@bigasterisk.com
parents:
1677
diff
changeset
|
373 odolog.debug(f'{INDENT*4} advanceAll [ring={i}] {ring} carry/advance') |
1631
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
374 ring.advance() |
1693
0455a1e18e4f
really rough fix to a broken test. hopefully this gets redone
drewp@bigasterisk.com
parents:
1692
diff
changeset
|
375 freshenRight(i) |
1631
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
376 carry = False |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
377 if ring.pastEnd(): |
1668
89e53cb8a01c
fix some harder tests. Mostly, _advanceTheStack needed to spin the odometer rings starting from the other side, to get all the right combos
drewp@bigasterisk.com
parents:
1667
diff
changeset
|
378 if ring is looperRings[0]: |
1664
1a7c1261302c
logic fix- some bindings were being returned 2+; some 0 times
drewp@bigasterisk.com
parents:
1661
diff
changeset
|
379 allRingsDone = [r.pastEnd() for r in looperRings] |
1692
2883da14847c
debugging and cleanup, as i looked for a bug
drewp@bigasterisk.com
parents:
1677
diff
changeset
|
380 odolog.debug(f'{INDENT*5} advanceAll [ring={i}] {ring} says we done {allRingsDone=}') |
1631
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
381 return True |
1692
2883da14847c
debugging and cleanup, as i looked for a bug
drewp@bigasterisk.com
parents:
1677
diff
changeset
|
382 odolog.debug(f'{INDENT*5} advanceAll [ring={i}] {ring} restart') |
1631
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
383 ring.restart() |
1693
0455a1e18e4f
really rough fix to a broken test. hopefully this gets redone
drewp@bigasterisk.com
parents:
1692
diff
changeset
|
384 freshenRight(i) |
1631
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
385 carry = True |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
386 return False |
1607
b21885181e35
more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents:
1605
diff
changeset
|
387 |
1664
1a7c1261302c
logic fix- some bindings were being returned 2+; some 0 times
drewp@bigasterisk.com
parents:
1661
diff
changeset
|
388 def _assertAllRingsAreValid(self, looperRings): |
1a7c1261302c
logic fix- some bindings were being returned 2+; some 0 times
drewp@bigasterisk.com
parents:
1661
diff
changeset
|
389 if any(ring.pastEnd() for ring in looperRings): # this is an unexpected debug assertion |
1a7c1261302c
logic fix- some bindings were being returned 2+; some 0 times
drewp@bigasterisk.com
parents:
1661
diff
changeset
|
390 odolog.warning(f'{INDENT*4} some rings started at pastEnd {looperRings}') |
1633
6107603ed455
fix farenheit rule case, fix some others that depend on rings order, but this breaks some performance because of itertools.perm
drewp@bigasterisk.com
parents:
1632
diff
changeset
|
391 raise NoOptions() |
6107603ed455
fix farenheit rule case, fix some others that depend on rings order, but this breaks some performance because of itertools.perm
drewp@bigasterisk.com
parents:
1632
diff
changeset
|
392 |
1592
d7b66234064b
pure reordering of funcs to make the next diffs smaller
drewp@bigasterisk.com
parents:
1591
diff
changeset
|
393 |
1622
38bd8ef9ef67
add CandidateTermMatches, unused so far
drewp@bigasterisk.com
parents:
1621
diff
changeset
|
394 @dataclass |
1607
b21885181e35
more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents:
1605
diff
changeset
|
395 class BoundLhs: |
b21885181e35
more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents:
1605
diff
changeset
|
396 lhs: Lhs |
1610 | 397 binding: CandidateBinding |
1607
b21885181e35
more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents:
1605
diff
changeset
|
398 |
b21885181e35
more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents:
1605
diff
changeset
|
399 |
b21885181e35
more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents:
1605
diff
changeset
|
400 @dataclass |
b21885181e35
more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents:
1605
diff
changeset
|
401 class Rule: |
b21885181e35
more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents:
1605
diff
changeset
|
402 lhsGraph: Graph |
b21885181e35
more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents:
1605
diff
changeset
|
403 rhsGraph: Graph |
1632 | 404 |
1607
b21885181e35
more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents:
1605
diff
changeset
|
405 def __post_init__(self): |
1673 | 406 self.lhs = Lhs(ChunkedGraph(self.lhsGraph, RuleUnboundBnode, functionsFor)) |
407 | |
408 self.maps = {} | |
409 | |
410 self.rhsGraphConvert: List[Triple] = [] | |
411 for s, p, o in self.rhsGraph: | |
412 from rdflib import BNode | |
413 if isinstance(s, BNode): | |
414 s = RhsBnode(s) | |
415 if isinstance(p, BNode): | |
416 p = RhsBnode(p) | |
417 if isinstance(o, BNode): | |
418 o = RhsBnode(o) | |
419 self.rhsGraphConvert.append((s, p, o)) | |
1607
b21885181e35
more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents:
1605
diff
changeset
|
420 |
1694
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
421 def applyRule(self, workingSet: Graph, implied: Graph, stats: Dict, slog: Optional[StructuredLog], |
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
422 ruleStatementsIterationLimit): |
1651
20474ad4968e
WIP - functions are broken as i move most layers to work in Chunks not Triples
drewp@bigasterisk.com
parents:
1650
diff
changeset
|
423 # this does not change for the current applyRule call. The rule will be |
20474ad4968e
WIP - functions are broken as i move most layers to work in Chunks not Triples
drewp@bigasterisk.com
parents:
1650
diff
changeset
|
424 # tried again in an outer loop, in case it can produce more. |
1673 | 425 workingSetChunked = ChunkedGraph(workingSet, WorkingSetBnode, functionsFor) |
1651
20474ad4968e
WIP - functions are broken as i move most layers to work in Chunks not Triples
drewp@bigasterisk.com
parents:
1650
diff
changeset
|
426 |
1694
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
427 for bound in self.lhs.findCandidateBindings(workingSetChunked, stats, slog, ruleStatementsIterationLimit): |
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
428 if slog: |
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
429 slog.foundBinding(bound) |
1631
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
430 log.debug(f'{INDENT*5} +rule has a working binding: {bound}') |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
431 |
1673 | 432 newStmts = self.generateImpliedFromRhs(bound.binding) |
1612
272f78d4671a
mark skipped tests. move applyRule into Rule. minor cleanups.
drewp@bigasterisk.com
parents:
1611
diff
changeset
|
433 |
1673 | 434 for newStmt in newStmts: |
1631
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
435 # log.debug(f'{INDENT*6} adding {newStmt=}') |
1612
272f78d4671a
mark skipped tests. move applyRule into Rule. minor cleanups.
drewp@bigasterisk.com
parents:
1611
diff
changeset
|
436 workingSet.add(newStmt) |
272f78d4671a
mark skipped tests. move applyRule into Rule. minor cleanups.
drewp@bigasterisk.com
parents:
1611
diff
changeset
|
437 implied.add(newStmt) |
272f78d4671a
mark skipped tests. move applyRule into Rule. minor cleanups.
drewp@bigasterisk.com
parents:
1611
diff
changeset
|
438 |
1673 | 439 def generateImpliedFromRhs(self, binding: CandidateBinding) -> List[Triple]: |
440 | |
441 out: List[Triple] = [] | |
442 | |
443 # Each time the RHS is used (in a rule firing), its own BNodes (which | |
444 # are subtype RhsBnode) need to be turned into distinct ones. Note that | |
445 # bnodes that come from the working set should not be remapped. | |
446 rhsBnodeMap: Dict[RhsBnode, WorkingSetBnode] = {} | |
447 | |
448 # but, the iteration loop could come back with the same bindings again | |
449 key = binding.key() | |
450 rhsBnodeMap = self.maps.setdefault(key, {}) | |
451 | |
452 for stmt in binding.apply(self.rhsGraphConvert): | |
453 | |
454 outStmt: List[Node] = [] | |
455 | |
456 for t in stmt: | |
457 if isinstance(t, RhsBnode): | |
458 if t not in rhsBnodeMap: | |
459 rhsBnodeMap[t] = WorkingSetBnode() | |
460 t = rhsBnodeMap[t] | |
461 | |
462 outStmt.append(t) | |
463 | |
464 log.debug(f'{INDENT*6} rhs stmt {stmt} became {outStmt}') | |
465 out.append((outStmt[0], outStmt[1], outStmt[2])) | |
466 | |
467 return out | |
468 | |
1607
b21885181e35
more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents:
1605
diff
changeset
|
469 |
1648 | 470 @dataclass |
1587 | 471 class Inference: |
1673 | 472 rulesIterationLimit = 4 |
1664
1a7c1261302c
logic fix- some bindings were being returned 2+; some 0 times
drewp@bigasterisk.com
parents:
1661
diff
changeset
|
473 ruleStatementsIterationLimit = 5000 |
1587 | 474 |
475 def __init__(self) -> None: | |
1648 | 476 self.rules: List[Rule] = [] |
477 self._nonRuleStmts: List[Triple] = [] | |
1587 | 478 |
479 def setRules(self, g: ConjunctiveGraph): | |
1648 | 480 self.rules = [] |
481 self._nonRuleStmts = [] | |
1599
abbf0eb0e640
fix a bug with a slightly moer complicated set of rules
drewp@bigasterisk.com
parents:
1598
diff
changeset
|
482 for stmt in g: |
abbf0eb0e640
fix a bug with a slightly moer complicated set of rules
drewp@bigasterisk.com
parents:
1598
diff
changeset
|
483 if stmt[1] == LOG['implies']: |
1607
b21885181e35
more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents:
1605
diff
changeset
|
484 self.rules.append(Rule(stmt[0], stmt[2])) |
1649
bb5d2b5370ac
add nonRuleStatments to Inference api. there's already a test in an eariler commit
drewp@bigasterisk.com
parents:
1648
diff
changeset
|
485 else: |
bb5d2b5370ac
add nonRuleStatments to Inference api. there's already a test in an eariler commit
drewp@bigasterisk.com
parents:
1648
diff
changeset
|
486 self._nonRuleStmts.append(stmt) |
bb5d2b5370ac
add nonRuleStatments to Inference api. there's already a test in an eariler commit
drewp@bigasterisk.com
parents:
1648
diff
changeset
|
487 |
bb5d2b5370ac
add nonRuleStatments to Inference api. there's already a test in an eariler commit
drewp@bigasterisk.com
parents:
1648
diff
changeset
|
488 def nonRuleStatements(self) -> List[Triple]: |
bb5d2b5370ac
add nonRuleStatments to Inference api. there's already a test in an eariler commit
drewp@bigasterisk.com
parents:
1648
diff
changeset
|
489 return self._nonRuleStmts |
1587 | 490 |
1601 | 491 @INFER_CALLS.time() |
1694
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
492 def infer(self, graph: Graph, htmlLog: Optional[Path] = None): |
1587 | 493 """ |
494 returns new graph of inferred statements. | |
495 """ | |
1626 | 496 n = graph.__len__() |
497 INFER_GRAPH_SIZE.observe(n) | |
498 log.info(f'{INDENT*0} Begin inference of graph len={n} with rules len={len(self.rules)}:') | |
1601 | 499 startTime = time.time() |
1607
b21885181e35
more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents:
1605
diff
changeset
|
500 stats: Dict[str, Union[int, float]] = defaultdict(lambda: 0) |
1649
bb5d2b5370ac
add nonRuleStatments to Inference api. there's already a test in an eariler commit
drewp@bigasterisk.com
parents:
1648
diff
changeset
|
501 |
1589
5c1055be3c36
WIP more debugging, working towards bnode-matching support
drewp@bigasterisk.com
parents:
1588
diff
changeset
|
502 # everything that is true: the input graph, plus every rule conclusion we can make |
1593
b0df43d5494c
big rewrite- more classes, smaller methods, more typesafe, all current tests passing
drewp@bigasterisk.com
parents:
1592
diff
changeset
|
503 workingSet = Graph() |
1649
bb5d2b5370ac
add nonRuleStatments to Inference api. there's already a test in an eariler commit
drewp@bigasterisk.com
parents:
1648
diff
changeset
|
504 workingSet += self._nonRuleStmts |
1593
b0df43d5494c
big rewrite- more classes, smaller methods, more typesafe, all current tests passing
drewp@bigasterisk.com
parents:
1592
diff
changeset
|
505 workingSet += graph |
1587 | 506 |
1594 | 507 # just the statements that came from RHS's of rules that fired. |
1587 | 508 implied = ConjunctiveGraph() |
509 | |
1694
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
510 slog = StructuredLog(htmlLog) if htmlLog else None |
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
511 |
1648 | 512 rulesIterations = 0 |
1587 | 513 delta = 1 |
1607
b21885181e35
more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents:
1605
diff
changeset
|
514 stats['initWorkingSet'] = cast(int, workingSet.__len__()) |
1694
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
515 if slog: |
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
516 slog.workingSet = workingSet |
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
517 |
1674
4a15b4cd4600
don't jsut quietly succeed if iteration limit is reached
drewp@bigasterisk.com
parents:
1673
diff
changeset
|
518 while delta > 0: |
1620 | 519 log.debug('') |
1648 | 520 log.info(f'{INDENT*1}*iteration {rulesIterations}') |
1694
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
521 if slog: |
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
522 slog.startIteration(rulesIterations) |
1648 | 523 |
1587 | 524 delta = -len(implied) |
1694
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
525 self._iterateAllRules(workingSet, implied, stats, slog) |
1587 | 526 delta += len(implied) |
1648 | 527 rulesIterations += 1 |
1597 | 528 log.info(f'{INDENT*2} this inference iteration added {delta} more implied stmts') |
1674
4a15b4cd4600
don't jsut quietly succeed if iteration limit is reached
drewp@bigasterisk.com
parents:
1673
diff
changeset
|
529 if rulesIterations >= self.rulesIterationLimit: |
4a15b4cd4600
don't jsut quietly succeed if iteration limit is reached
drewp@bigasterisk.com
parents:
1673
diff
changeset
|
530 raise ValueError(f"rule too complex after {rulesIterations=}") |
1648 | 531 stats['iterations'] = rulesIterations |
1607
b21885181e35
more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents:
1605
diff
changeset
|
532 stats['timeSpent'] = round(time.time() - startTime, 3) |
b21885181e35
more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents:
1605
diff
changeset
|
533 stats['impliedStmts'] = len(implied) |
1648 | 534 log.info(f'{INDENT*0} Inference done {dict(stats)}.') |
535 log.debug('Implied:') | |
536 log.debug(graphDump(implied)) | |
1694
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
537 |
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
538 if slog: |
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
539 slog.render() |
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
540 log.info(f'wrote {htmlLog}') |
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
541 |
1587 | 542 return implied |
543 | |
1694
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
544 def _iterateAllRules(self, workingSet: Graph, implied: Graph, stats, slog: Optional[StructuredLog]): |
1612
272f78d4671a
mark skipped tests. move applyRule into Rule. minor cleanups.
drewp@bigasterisk.com
parents:
1611
diff
changeset
|
545 for i, rule in enumerate(self.rules): |
272f78d4671a
mark skipped tests. move applyRule into Rule. minor cleanups.
drewp@bigasterisk.com
parents:
1611
diff
changeset
|
546 self._logRuleApplicationHeader(workingSet, i, rule) |
1694
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
547 if slog: |
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
548 slog.rule(workingSet, i, rule) |
73abfd4cf5d0
new html log and other refactoring as i work on the advanceTheStack problems
drewp@bigasterisk.com
parents:
1693
diff
changeset
|
549 rule.applyRule(workingSet, implied, stats, slog, self.ruleStatementsIterationLimit) |
1587 | 550 |
1607
b21885181e35
more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents:
1605
diff
changeset
|
551 def _logRuleApplicationHeader(self, workingSet, i, r: Rule): |
1594 | 552 if not log.isEnabledFor(logging.DEBUG): |
553 return | |
554 | |
555 log.debug('') | |
556 log.debug(f'{INDENT*2} workingSet:') | |
1648 | 557 # for j, stmt in enumerate(sorted(workingSet)): |
558 # log.debug(f'{INDENT*3} ({j}) {stmt}') | |
559 log.debug(f'{INDENT*3} {graphDump(workingSet, oneLine=False)}') | |
1594 | 560 |
561 log.debug('') | |
562 log.debug(f'{INDENT*2}-applying rule {i}') | |
1632 | 563 log.debug(f'{INDENT*3} rule def lhs:') |
1677 | 564 for stmt in sorted(r.lhs.graph.allChunks()): |
1632 | 565 log.debug(f'{INDENT*4} {stmt}') |
1607
b21885181e35
more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents:
1605
diff
changeset
|
566 log.debug(f'{INDENT*3} rule def rhs: {graphDump(r.rhsGraph)}') |