annotate service/mqtt_to_rdf/inference.py @ 1626:7b3656867185

metrics on input graph sizes
author drewp@bigasterisk.com
date Sat, 11 Sep 2021 23:21:11 -0700
parents cf901d219007
children ea559a846714
Ignore whitespace changes - Everywhere: Within whitespace: At end of lines:
rev   line source
1587
9a3a18c494f9 WIP new inferencer. no vars yet.
drewp@bigasterisk.com
parents:
diff changeset
1 """
9a3a18c494f9 WIP new inferencer. no vars yet.
drewp@bigasterisk.com
parents:
diff changeset
2 copied from reasoning 2021-08-29. probably same api. should
9a3a18c494f9 WIP new inferencer. no vars yet.
drewp@bigasterisk.com
parents:
diff changeset
3 be able to lib/ this out
9a3a18c494f9 WIP new inferencer. no vars yet.
drewp@bigasterisk.com
parents:
diff changeset
4 """
1588
0757fafbfdab WIP inferencer - partial var and function support
drewp@bigasterisk.com
parents: 1587
diff changeset
5 import itertools
1587
9a3a18c494f9 WIP new inferencer. no vars yet.
drewp@bigasterisk.com
parents:
diff changeset
6 import logging
1601
30463df12d89 infer() dumps stats
drewp@bigasterisk.com
parents: 1600
diff changeset
7 import time
1594
e58bcfa66093 cleanups and a few fixed cases
drewp@bigasterisk.com
parents: 1593
diff changeset
8 from collections import defaultdict
1626
7b3656867185 metrics on input graph sizes
drewp@bigasterisk.com
parents: 1623
diff changeset
9 from dataclasses import dataclass
7b3656867185 metrics on input graph sizes
drewp@bigasterisk.com
parents: 1623
diff changeset
10 from typing import Dict, Iterator, List, Set, Tuple, Union, cast
1588
0757fafbfdab WIP inferencer - partial var and function support
drewp@bigasterisk.com
parents: 1587
diff changeset
11
1626
7b3656867185 metrics on input graph sizes
drewp@bigasterisk.com
parents: 1623
diff changeset
12 from prometheus_client import Summary, Histogram
1607
b21885181e35 more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents: 1605
diff changeset
13 from rdflib import BNode, Graph, Namespace, URIRef
1589
5c1055be3c36 WIP more debugging, working towards bnode-matching support
drewp@bigasterisk.com
parents: 1588
diff changeset
14 from rdflib.graph import ConjunctiveGraph, ReadOnlyGraphAggregate
1587
9a3a18c494f9 WIP new inferencer. no vars yet.
drewp@bigasterisk.com
parents:
diff changeset
15 from rdflib.term import Node, Variable
9a3a18c494f9 WIP new inferencer. no vars yet.
drewp@bigasterisk.com
parents:
diff changeset
16
1607
b21885181e35 more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents: 1605
diff changeset
17 from candidate_binding import CandidateBinding
b21885181e35 more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents: 1605
diff changeset
18 from inference_types import (BindableTerm, EvaluationFailed, ReadOnlyWorkingSet, Triple)
b21885181e35 more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents: 1605
diff changeset
19 from lhs_evaluation import Evaluation
1605
449746d1598f WIP move evaluation to new file
drewp@bigasterisk.com
parents: 1603
diff changeset
20
1587
9a3a18c494f9 WIP new inferencer. no vars yet.
drewp@bigasterisk.com
parents:
diff changeset
21 log = logging.getLogger('infer')
1593
b0df43d5494c big rewrite- more classes, smaller methods, more typesafe, all current tests passing
drewp@bigasterisk.com
parents: 1592
diff changeset
22 INDENT = ' '
1587
9a3a18c494f9 WIP new inferencer. no vars yet.
drewp@bigasterisk.com
parents:
diff changeset
23
1626
7b3656867185 metrics on input graph sizes
drewp@bigasterisk.com
parents: 1623
diff changeset
24 INFER_CALLS = Summary('inference_infer_calls', 'calls')
7b3656867185 metrics on input graph sizes
drewp@bigasterisk.com
parents: 1623
diff changeset
25 INFER_GRAPH_SIZE = Histogram('inference_graph_size', 'statements', buckets=[2**x for x in range(2, 20, 2)])
1587
9a3a18c494f9 WIP new inferencer. no vars yet.
drewp@bigasterisk.com
parents:
diff changeset
26
9a3a18c494f9 WIP new inferencer. no vars yet.
drewp@bigasterisk.com
parents:
diff changeset
27 ROOM = Namespace("http://projects.bigasterisk.com/room/")
9a3a18c494f9 WIP new inferencer. no vars yet.
drewp@bigasterisk.com
parents:
diff changeset
28 LOG = Namespace('http://www.w3.org/2000/10/swap/log#')
9a3a18c494f9 WIP new inferencer. no vars yet.
drewp@bigasterisk.com
parents:
diff changeset
29 MATH = Namespace('http://www.w3.org/2000/10/swap/math#')
9a3a18c494f9 WIP new inferencer. no vars yet.
drewp@bigasterisk.com
parents:
diff changeset
30
9a3a18c494f9 WIP new inferencer. no vars yet.
drewp@bigasterisk.com
parents:
diff changeset
31
1594
e58bcfa66093 cleanups and a few fixed cases
drewp@bigasterisk.com
parents: 1593
diff changeset
32 @dataclass
1593
b0df43d5494c big rewrite- more classes, smaller methods, more typesafe, all current tests passing
drewp@bigasterisk.com
parents: 1592
diff changeset
33 class Lhs:
1594
e58bcfa66093 cleanups and a few fixed cases
drewp@bigasterisk.com
parents: 1593
diff changeset
34 graph: Graph
e58bcfa66093 cleanups and a few fixed cases
drewp@bigasterisk.com
parents: 1593
diff changeset
35
e58bcfa66093 cleanups and a few fixed cases
drewp@bigasterisk.com
parents: 1593
diff changeset
36 def __post_init__(self):
1608
f928eb06a4f6 cleaning up inner loop
drewp@bigasterisk.com
parents: 1607
diff changeset
37 # do precomputation in here that's not specific to the workingSet
f928eb06a4f6 cleaning up inner loop
drewp@bigasterisk.com
parents: 1607
diff changeset
38 self.staticRuleStmts = Graph()
1611
a794a150a89b a little inner-loop performace. same number of iterations, but less time on logging and some stmt filtering
drewp@bigasterisk.com
parents: 1610
diff changeset
39 self.nonStaticRuleStmts = Graph()
a794a150a89b a little inner-loop performace. same number of iterations, but less time on logging and some stmt filtering
drewp@bigasterisk.com
parents: 1610
diff changeset
40
1608
f928eb06a4f6 cleaning up inner loop
drewp@bigasterisk.com
parents: 1607
diff changeset
41 self.lhsBindables: Set[BindableTerm] = set()
f928eb06a4f6 cleaning up inner loop
drewp@bigasterisk.com
parents: 1607
diff changeset
42 self.lhsBnodes: Set[BNode] = set()
1594
e58bcfa66093 cleanups and a few fixed cases
drewp@bigasterisk.com
parents: 1593
diff changeset
43 for ruleStmt in self.graph:
e58bcfa66093 cleanups and a few fixed cases
drewp@bigasterisk.com
parents: 1593
diff changeset
44 varsAndBnodesInStmt = [term for term in ruleStmt if isinstance(term, (Variable, BNode))]
e58bcfa66093 cleanups and a few fixed cases
drewp@bigasterisk.com
parents: 1593
diff changeset
45 self.lhsBindables.update(varsAndBnodesInStmt)
e58bcfa66093 cleanups and a few fixed cases
drewp@bigasterisk.com
parents: 1593
diff changeset
46 self.lhsBnodes.update(x for x in varsAndBnodesInStmt if isinstance(x, BNode))
e58bcfa66093 cleanups and a few fixed cases
drewp@bigasterisk.com
parents: 1593
diff changeset
47 if not varsAndBnodesInStmt:
e58bcfa66093 cleanups and a few fixed cases
drewp@bigasterisk.com
parents: 1593
diff changeset
48 self.staticRuleStmts.add(ruleStmt)
1611
a794a150a89b a little inner-loop performace. same number of iterations, but less time on logging and some stmt filtering
drewp@bigasterisk.com
parents: 1610
diff changeset
49 else:
a794a150a89b a little inner-loop performace. same number of iterations, but less time on logging and some stmt filtering
drewp@bigasterisk.com
parents: 1610
diff changeset
50 self.nonStaticRuleStmts.add(ruleStmt)
1593
b0df43d5494c big rewrite- more classes, smaller methods, more typesafe, all current tests passing
drewp@bigasterisk.com
parents: 1592
diff changeset
51
1615
bcfa368e5498 change a Graph.__sub__ to Set.difference in verify() for a big speedup
drewp@bigasterisk.com
parents: 1614
diff changeset
52 self.nonStaticRuleStmtsSet = set(self.nonStaticRuleStmts)
bcfa368e5498 change a Graph.__sub__ to Set.difference in verify() for a big speedup
drewp@bigasterisk.com
parents: 1614
diff changeset
53
1602
e3c44ac6d3c5 do findEvals once at setRules time
drewp@bigasterisk.com
parents: 1601
diff changeset
54 self.evaluations = list(Evaluation.findEvals(self.graph))
e3c44ac6d3c5 do findEvals once at setRules time
drewp@bigasterisk.com
parents: 1601
diff changeset
55
1609
34f2817320cc new tests for a smaller part of the inner loop
drewp@bigasterisk.com
parents: 1608
diff changeset
56 def __repr__(self):
34f2817320cc new tests for a smaller part of the inner loop
drewp@bigasterisk.com
parents: 1608
diff changeset
57 return f"Lhs({graphDump(self.graph)})"
34f2817320cc new tests for a smaller part of the inner loop
drewp@bigasterisk.com
parents: 1608
diff changeset
58
1621
da235054caa0 rename workingSet, sometimes
drewp@bigasterisk.com
parents: 1620
diff changeset
59 def findCandidateBindings(self, knownTrue: ReadOnlyWorkingSet, stats) -> Iterator['BoundLhs']:
1593
b0df43d5494c big rewrite- more classes, smaller methods, more typesafe, all current tests passing
drewp@bigasterisk.com
parents: 1592
diff changeset
60 """bindings that fit the LHS of a rule, using statements from workingSet and functions
b0df43d5494c big rewrite- more classes, smaller methods, more typesafe, all current tests passing
drewp@bigasterisk.com
parents: 1592
diff changeset
61 from LHS"""
1597
387a9cb66517 logging adjustments
drewp@bigasterisk.com
parents: 1596
diff changeset
62 log.debug(f'{INDENT*3} nodesToBind: {self.lhsBindables}')
1607
b21885181e35 more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents: 1605
diff changeset
63 stats['findCandidateBindingsCalls'] += 1
1593
b0df43d5494c big rewrite- more classes, smaller methods, more typesafe, all current tests passing
drewp@bigasterisk.com
parents: 1592
diff changeset
64
1621
da235054caa0 rename workingSet, sometimes
drewp@bigasterisk.com
parents: 1620
diff changeset
65 if not self._allStaticStatementsMatch(knownTrue):
1607
b21885181e35 more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents: 1605
diff changeset
66 stats['findCandidateBindingEarlyExits'] += 1
1593
b0df43d5494c big rewrite- more classes, smaller methods, more typesafe, all current tests passing
drewp@bigasterisk.com
parents: 1592
diff changeset
67 return
b0df43d5494c big rewrite- more classes, smaller methods, more typesafe, all current tests passing
drewp@bigasterisk.com
parents: 1592
diff changeset
68
1621
da235054caa0 rename workingSet, sometimes
drewp@bigasterisk.com
parents: 1620
diff changeset
69 for binding in self._possibleBindings(knownTrue, stats):
1607
b21885181e35 more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents: 1605
diff changeset
70 log.debug('')
b21885181e35 more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents: 1605
diff changeset
71 log.debug(f'{INDENT*4}*trying {binding.binding}')
b21885181e35 more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents: 1605
diff changeset
72
1621
da235054caa0 rename workingSet, sometimes
drewp@bigasterisk.com
parents: 1620
diff changeset
73 if not binding.verify(knownTrue):
1607
b21885181e35 more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents: 1605
diff changeset
74 log.debug(f'{INDENT*4} this binding did not verify')
b21885181e35 more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents: 1605
diff changeset
75 stats['permCountFailingVerify'] += 1
b21885181e35 more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents: 1605
diff changeset
76 continue
b21885181e35 more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents: 1605
diff changeset
77
b21885181e35 more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents: 1605
diff changeset
78 stats['permCountSucceeding'] += 1
b21885181e35 more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents: 1605
diff changeset
79 yield binding
b21885181e35 more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents: 1605
diff changeset
80
1621
da235054caa0 rename workingSet, sometimes
drewp@bigasterisk.com
parents: 1620
diff changeset
81 def _allStaticStatementsMatch(self, knownTrue: ReadOnlyWorkingSet) -> bool:
1613
03ed8c9abd5b forget GRAPH_ID optimization in this case
drewp@bigasterisk.com
parents: 1612
diff changeset
82 # bug: see TestSelfFulfillingRule.test3 for a case where this rule's
03ed8c9abd5b forget GRAPH_ID optimization in this case
drewp@bigasterisk.com
parents: 1612
diff changeset
83 # static stmt is matched by a non-static stmt in the rule itself
1608
f928eb06a4f6 cleaning up inner loop
drewp@bigasterisk.com
parents: 1607
diff changeset
84 for ruleStmt in self.staticRuleStmts:
1621
da235054caa0 rename workingSet, sometimes
drewp@bigasterisk.com
parents: 1620
diff changeset
85 if ruleStmt not in knownTrue:
1608
f928eb06a4f6 cleaning up inner loop
drewp@bigasterisk.com
parents: 1607
diff changeset
86 log.debug(f'{INDENT*3} {ruleStmt} not in working set- skip rule')
f928eb06a4f6 cleaning up inner loop
drewp@bigasterisk.com
parents: 1607
diff changeset
87 return False
f928eb06a4f6 cleaning up inner loop
drewp@bigasterisk.com
parents: 1607
diff changeset
88 return True
f928eb06a4f6 cleaning up inner loop
drewp@bigasterisk.com
parents: 1607
diff changeset
89
1607
b21885181e35 more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents: 1605
diff changeset
90 def _possibleBindings(self, workingSet, stats) -> Iterator['BoundLhs']:
b21885181e35 more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents: 1605
diff changeset
91 """this yields at least the working bindings, and possibly others"""
1600
89a50242cb5e cleanup internal names, imports
drewp@bigasterisk.com
parents: 1599
diff changeset
92 candidateTermMatches: Dict[BindableTerm, Set[Node]] = self._allCandidateTermMatches(workingSet)
1614
97b2c3cfdb83 refactor: inline an odometer algorithm in place of itertools.product
drewp@bigasterisk.com
parents: 1613
diff changeset
93 for bindRow in self._product(candidateTermMatches):
97b2c3cfdb83 refactor: inline an odometer algorithm in place of itertools.product
drewp@bigasterisk.com
parents: 1613
diff changeset
94 try:
97b2c3cfdb83 refactor: inline an odometer algorithm in place of itertools.product
drewp@bigasterisk.com
parents: 1613
diff changeset
95 yield BoundLhs(self, bindRow)
97b2c3cfdb83 refactor: inline an odometer algorithm in place of itertools.product
drewp@bigasterisk.com
parents: 1613
diff changeset
96 except EvaluationFailed:
97b2c3cfdb83 refactor: inline an odometer algorithm in place of itertools.product
drewp@bigasterisk.com
parents: 1613
diff changeset
97 stats['permCountFailingEval'] += 1
1593
b0df43d5494c big rewrite- more classes, smaller methods, more typesafe, all current tests passing
drewp@bigasterisk.com
parents: 1592
diff changeset
98
1614
97b2c3cfdb83 refactor: inline an odometer algorithm in place of itertools.product
drewp@bigasterisk.com
parents: 1613
diff changeset
99 def _product(self, candidateTermMatches: Dict[BindableTerm, Set[Node]]) -> Iterator[CandidateBinding]:
1600
89a50242cb5e cleanup internal names, imports
drewp@bigasterisk.com
parents: 1599
diff changeset
100 orderedVars, orderedValueSets = _organize(candidateTermMatches)
89a50242cb5e cleanup internal names, imports
drewp@bigasterisk.com
parents: 1599
diff changeset
101 self._logCandidates(orderedVars, orderedValueSets)
1597
387a9cb66517 logging adjustments
drewp@bigasterisk.com
parents: 1596
diff changeset
102 log.debug(f'{INDENT*3} trying all permutations:')
1614
97b2c3cfdb83 refactor: inline an odometer algorithm in place of itertools.product
drewp@bigasterisk.com
parents: 1613
diff changeset
103 if not orderedValueSets:
97b2c3cfdb83 refactor: inline an odometer algorithm in place of itertools.product
drewp@bigasterisk.com
parents: 1613
diff changeset
104 yield CandidateBinding({})
97b2c3cfdb83 refactor: inline an odometer algorithm in place of itertools.product
drewp@bigasterisk.com
parents: 1613
diff changeset
105 return
1620
92f8deb59735 log layout
drewp@bigasterisk.com
parents: 1616
diff changeset
106
1614
97b2c3cfdb83 refactor: inline an odometer algorithm in place of itertools.product
drewp@bigasterisk.com
parents: 1613
diff changeset
107 if not orderedValueSets or not all(orderedValueSets):
97b2c3cfdb83 refactor: inline an odometer algorithm in place of itertools.product
drewp@bigasterisk.com
parents: 1613
diff changeset
108 # some var or bnode has no options at all
97b2c3cfdb83 refactor: inline an odometer algorithm in place of itertools.product
drewp@bigasterisk.com
parents: 1613
diff changeset
109 return
97b2c3cfdb83 refactor: inline an odometer algorithm in place of itertools.product
drewp@bigasterisk.com
parents: 1613
diff changeset
110 rings: List[Iterator[Node]] = [itertools.cycle(valSet) for valSet in orderedValueSets]
97b2c3cfdb83 refactor: inline an odometer algorithm in place of itertools.product
drewp@bigasterisk.com
parents: 1613
diff changeset
111 currentSet: List[Node] = [next(ring) for ring in rings]
97b2c3cfdb83 refactor: inline an odometer algorithm in place of itertools.product
drewp@bigasterisk.com
parents: 1613
diff changeset
112 starts = [valSet[-1] for valSet in orderedValueSets]
97b2c3cfdb83 refactor: inline an odometer algorithm in place of itertools.product
drewp@bigasterisk.com
parents: 1613
diff changeset
113 while True:
97b2c3cfdb83 refactor: inline an odometer algorithm in place of itertools.product
drewp@bigasterisk.com
parents: 1613
diff changeset
114 for col, curr in enumerate(currentSet):
97b2c3cfdb83 refactor: inline an odometer algorithm in place of itertools.product
drewp@bigasterisk.com
parents: 1613
diff changeset
115 currentSet[col] = next(rings[col])
1620
92f8deb59735 log layout
drewp@bigasterisk.com
parents: 1616
diff changeset
116 log.debug(f'{INDENT*4} currentSet: {repr(currentSet)}')
1614
97b2c3cfdb83 refactor: inline an odometer algorithm in place of itertools.product
drewp@bigasterisk.com
parents: 1613
diff changeset
117 yield CandidateBinding(dict(zip(orderedVars, currentSet)))
97b2c3cfdb83 refactor: inline an odometer algorithm in place of itertools.product
drewp@bigasterisk.com
parents: 1613
diff changeset
118 if curr is not starts[col]:
97b2c3cfdb83 refactor: inline an odometer algorithm in place of itertools.product
drewp@bigasterisk.com
parents: 1613
diff changeset
119 break
97b2c3cfdb83 refactor: inline an odometer algorithm in place of itertools.product
drewp@bigasterisk.com
parents: 1613
diff changeset
120 if col == len(orderedValueSets) - 1:
97b2c3cfdb83 refactor: inline an odometer algorithm in place of itertools.product
drewp@bigasterisk.com
parents: 1613
diff changeset
121 return
1592
d7b66234064b pure reordering of funcs to make the next diffs smaller
drewp@bigasterisk.com
parents: 1591
diff changeset
122
1600
89a50242cb5e cleanup internal names, imports
drewp@bigasterisk.com
parents: 1599
diff changeset
123 def _allCandidateTermMatches(self, workingSet: ReadOnlyWorkingSet) -> Dict[BindableTerm, Set[Node]]:
1593
b0df43d5494c big rewrite- more classes, smaller methods, more typesafe, all current tests passing
drewp@bigasterisk.com
parents: 1592
diff changeset
124 """the total set of terms each variable could possibly match"""
1592
d7b66234064b pure reordering of funcs to make the next diffs smaller
drewp@bigasterisk.com
parents: 1591
diff changeset
125
1593
b0df43d5494c big rewrite- more classes, smaller methods, more typesafe, all current tests passing
drewp@bigasterisk.com
parents: 1592
diff changeset
126 candidateTermMatches: Dict[BindableTerm, Set[Node]] = defaultdict(set)
1594
e58bcfa66093 cleanups and a few fixed cases
drewp@bigasterisk.com
parents: 1593
diff changeset
127 for lhsStmt in self.graph:
1597
387a9cb66517 logging adjustments
drewp@bigasterisk.com
parents: 1596
diff changeset
128 log.debug(f'{INDENT*4} possibles for this lhs stmt: {lhsStmt}')
1608
f928eb06a4f6 cleaning up inner loop
drewp@bigasterisk.com
parents: 1607
diff changeset
129 for i, trueStmt in enumerate(workingSet):
1597
387a9cb66517 logging adjustments
drewp@bigasterisk.com
parents: 1596
diff changeset
130 # log.debug(f'{INDENT*5} consider this true stmt ({i}): {trueStmt}')
1592
d7b66234064b pure reordering of funcs to make the next diffs smaller
drewp@bigasterisk.com
parents: 1591
diff changeset
131
1594
e58bcfa66093 cleanups and a few fixed cases
drewp@bigasterisk.com
parents: 1593
diff changeset
132 for v, vals in self._bindingsFromStatement(lhsStmt, trueStmt):
e58bcfa66093 cleanups and a few fixed cases
drewp@bigasterisk.com
parents: 1593
diff changeset
133 candidateTermMatches[v].update(vals)
e58bcfa66093 cleanups and a few fixed cases
drewp@bigasterisk.com
parents: 1593
diff changeset
134
1593
b0df43d5494c big rewrite- more classes, smaller methods, more typesafe, all current tests passing
drewp@bigasterisk.com
parents: 1592
diff changeset
135 return candidateTermMatches
b0df43d5494c big rewrite- more classes, smaller methods, more typesafe, all current tests passing
drewp@bigasterisk.com
parents: 1592
diff changeset
136
1594
e58bcfa66093 cleanups and a few fixed cases
drewp@bigasterisk.com
parents: 1593
diff changeset
137 def _bindingsFromStatement(self, stmt1: Triple, stmt2: Triple) -> Iterator[Tuple[Variable, Set[Node]]]:
e58bcfa66093 cleanups and a few fixed cases
drewp@bigasterisk.com
parents: 1593
diff changeset
138 """if these stmts match otherwise, what BNode or Variable mappings do we learn?
e58bcfa66093 cleanups and a few fixed cases
drewp@bigasterisk.com
parents: 1593
diff changeset
139
e58bcfa66093 cleanups and a few fixed cases
drewp@bigasterisk.com
parents: 1593
diff changeset
140 e.g. stmt1=(?x B ?y) and stmt2=(A B C), then we yield (?x, {A}) and (?y, {C})
e58bcfa66093 cleanups and a few fixed cases
drewp@bigasterisk.com
parents: 1593
diff changeset
141 or stmt1=(_:x B C) and stmt2=(A B C), then we yield (_:x, {A})
e58bcfa66093 cleanups and a few fixed cases
drewp@bigasterisk.com
parents: 1593
diff changeset
142 or stmt1=(?x B C) and stmt2=(A B D), then we yield nothing
e58bcfa66093 cleanups and a few fixed cases
drewp@bigasterisk.com
parents: 1593
diff changeset
143 """
e58bcfa66093 cleanups and a few fixed cases
drewp@bigasterisk.com
parents: 1593
diff changeset
144 bindingsFromStatement = {}
e58bcfa66093 cleanups and a few fixed cases
drewp@bigasterisk.com
parents: 1593
diff changeset
145 for term1, term2 in zip(stmt1, stmt2):
e58bcfa66093 cleanups and a few fixed cases
drewp@bigasterisk.com
parents: 1593
diff changeset
146 if isinstance(term1, (BNode, Variable)):
e58bcfa66093 cleanups and a few fixed cases
drewp@bigasterisk.com
parents: 1593
diff changeset
147 bindingsFromStatement.setdefault(term1, set()).add(term2)
e58bcfa66093 cleanups and a few fixed cases
drewp@bigasterisk.com
parents: 1593
diff changeset
148 elif term1 != term2:
e58bcfa66093 cleanups and a few fixed cases
drewp@bigasterisk.com
parents: 1593
diff changeset
149 break
e58bcfa66093 cleanups and a few fixed cases
drewp@bigasterisk.com
parents: 1593
diff changeset
150 else:
e58bcfa66093 cleanups and a few fixed cases
drewp@bigasterisk.com
parents: 1593
diff changeset
151 for v, vals in bindingsFromStatement.items():
1597
387a9cb66517 logging adjustments
drewp@bigasterisk.com
parents: 1596
diff changeset
152 log.debug(f'{INDENT*5} {v=} {vals=}')
1594
e58bcfa66093 cleanups and a few fixed cases
drewp@bigasterisk.com
parents: 1593
diff changeset
153 yield v, vals
e58bcfa66093 cleanups and a few fixed cases
drewp@bigasterisk.com
parents: 1593
diff changeset
154
1600
89a50242cb5e cleanup internal names, imports
drewp@bigasterisk.com
parents: 1599
diff changeset
155 def _logCandidates(self, orderedVars, orderedValueSets):
1593
b0df43d5494c big rewrite- more classes, smaller methods, more typesafe, all current tests passing
drewp@bigasterisk.com
parents: 1592
diff changeset
156 if not log.isEnabledFor(logging.DEBUG):
b0df43d5494c big rewrite- more classes, smaller methods, more typesafe, all current tests passing
drewp@bigasterisk.com
parents: 1592
diff changeset
157 return
1597
387a9cb66517 logging adjustments
drewp@bigasterisk.com
parents: 1596
diff changeset
158 log.debug(f'{INDENT*3} resulting candidate terms:')
1593
b0df43d5494c big rewrite- more classes, smaller methods, more typesafe, all current tests passing
drewp@bigasterisk.com
parents: 1592
diff changeset
159 for v, vals in zip(orderedVars, orderedValueSets):
1597
387a9cb66517 logging adjustments
drewp@bigasterisk.com
parents: 1596
diff changeset
160 log.debug(f'{INDENT*4} {v!r} could be:')
1593
b0df43d5494c big rewrite- more classes, smaller methods, more typesafe, all current tests passing
drewp@bigasterisk.com
parents: 1592
diff changeset
161 for val in vals:
1597
387a9cb66517 logging adjustments
drewp@bigasterisk.com
parents: 1596
diff changeset
162 log.debug(f'{INDENT*5}{val!r}')
1592
d7b66234064b pure reordering of funcs to make the next diffs smaller
drewp@bigasterisk.com
parents: 1591
diff changeset
163
d7b66234064b pure reordering of funcs to make the next diffs smaller
drewp@bigasterisk.com
parents: 1591
diff changeset
164
1607
b21885181e35 more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents: 1605
diff changeset
165 @dataclass
1622
38bd8ef9ef67 add CandidateTermMatches, unused so far
drewp@bigasterisk.com
parents: 1621
diff changeset
166 class CandidateTermMatches:
38bd8ef9ef67 add CandidateTermMatches, unused so far
drewp@bigasterisk.com
parents: 1621
diff changeset
167 """lazily find the possible matches for this term"""
38bd8ef9ef67 add CandidateTermMatches, unused so far
drewp@bigasterisk.com
parents: 1621
diff changeset
168 term: BindableTerm
38bd8ef9ef67 add CandidateTermMatches, unused so far
drewp@bigasterisk.com
parents: 1621
diff changeset
169 lhs: Lhs
38bd8ef9ef67 add CandidateTermMatches, unused so far
drewp@bigasterisk.com
parents: 1621
diff changeset
170 workingSet: Graph
38bd8ef9ef67 add CandidateTermMatches, unused so far
drewp@bigasterisk.com
parents: 1621
diff changeset
171 boundSoFar: CandidateBinding
38bd8ef9ef67 add CandidateTermMatches, unused so far
drewp@bigasterisk.com
parents: 1621
diff changeset
172
38bd8ef9ef67 add CandidateTermMatches, unused so far
drewp@bigasterisk.com
parents: 1621
diff changeset
173 def __post_init__(self):
38bd8ef9ef67 add CandidateTermMatches, unused so far
drewp@bigasterisk.com
parents: 1621
diff changeset
174 self.results: List[Node] = [] # we have to be able to repeat the results
38bd8ef9ef67 add CandidateTermMatches, unused so far
drewp@bigasterisk.com
parents: 1621
diff changeset
175
38bd8ef9ef67 add CandidateTermMatches, unused so far
drewp@bigasterisk.com
parents: 1621
diff changeset
176 res: Set[Node] = set()
38bd8ef9ef67 add CandidateTermMatches, unused so far
drewp@bigasterisk.com
parents: 1621
diff changeset
177 for trueStmt in self.workingSet: # all bound
38bd8ef9ef67 add CandidateTermMatches, unused so far
drewp@bigasterisk.com
parents: 1621
diff changeset
178 lStmts = list(self.lhsStmtsContainingTerm())
38bd8ef9ef67 add CandidateTermMatches, unused so far
drewp@bigasterisk.com
parents: 1621
diff changeset
179 log.debug(f'{INDENT*4} {trueStmt=} {len(lStmts)}')
38bd8ef9ef67 add CandidateTermMatches, unused so far
drewp@bigasterisk.com
parents: 1621
diff changeset
180 for pat in self.boundSoFar.apply(lStmts, returnBoundStatementsOnly=False):
38bd8ef9ef67 add CandidateTermMatches, unused so far
drewp@bigasterisk.com
parents: 1621
diff changeset
181 log.debug(f'{INDENT*4} {pat=}')
38bd8ef9ef67 add CandidateTermMatches, unused so far
drewp@bigasterisk.com
parents: 1621
diff changeset
182 implied = self._stmtImplies(pat, trueStmt)
38bd8ef9ef67 add CandidateTermMatches, unused so far
drewp@bigasterisk.com
parents: 1621
diff changeset
183 if implied is not None:
38bd8ef9ef67 add CandidateTermMatches, unused so far
drewp@bigasterisk.com
parents: 1621
diff changeset
184 res.add(implied)
38bd8ef9ef67 add CandidateTermMatches, unused so far
drewp@bigasterisk.com
parents: 1621
diff changeset
185 self.results = list(res)
38bd8ef9ef67 add CandidateTermMatches, unused so far
drewp@bigasterisk.com
parents: 1621
diff changeset
186 # self.results.sort()
38bd8ef9ef67 add CandidateTermMatches, unused so far
drewp@bigasterisk.com
parents: 1621
diff changeset
187
38bd8ef9ef67 add CandidateTermMatches, unused so far
drewp@bigasterisk.com
parents: 1621
diff changeset
188 log.debug(f'{INDENT*3} CandTermMatches: {self.term} {graphDump(self.lhs.graph)} {self.boundSoFar=} ===> {self.results=}')
38bd8ef9ef67 add CandidateTermMatches, unused so far
drewp@bigasterisk.com
parents: 1621
diff changeset
189
38bd8ef9ef67 add CandidateTermMatches, unused so far
drewp@bigasterisk.com
parents: 1621
diff changeset
190 def _stmtImplies(self, pat: Triple, trueStmt: Triple) -> Optional[Node]:
38bd8ef9ef67 add CandidateTermMatches, unused so far
drewp@bigasterisk.com
parents: 1621
diff changeset
191 """what value, if any, do we learn for our term from this LHS pattern statement and this known-true stmt"""
38bd8ef9ef67 add CandidateTermMatches, unused so far
drewp@bigasterisk.com
parents: 1621
diff changeset
192 r = None
38bd8ef9ef67 add CandidateTermMatches, unused so far
drewp@bigasterisk.com
parents: 1621
diff changeset
193 for p, t in zip(pat, trueStmt):
38bd8ef9ef67 add CandidateTermMatches, unused so far
drewp@bigasterisk.com
parents: 1621
diff changeset
194 if isinstance(p, (Variable, BNode)):
38bd8ef9ef67 add CandidateTermMatches, unused so far
drewp@bigasterisk.com
parents: 1621
diff changeset
195 if p != self.term:
38bd8ef9ef67 add CandidateTermMatches, unused so far
drewp@bigasterisk.com
parents: 1621
diff changeset
196 # stmt is unbound in more than just our term
38bd8ef9ef67 add CandidateTermMatches, unused so far
drewp@bigasterisk.com
parents: 1621
diff changeset
197 continue # unsure what to do - err on the side of too many bindings, since they get rechecked later
38bd8ef9ef67 add CandidateTermMatches, unused so far
drewp@bigasterisk.com
parents: 1621
diff changeset
198 if r is None:
38bd8ef9ef67 add CandidateTermMatches, unused so far
drewp@bigasterisk.com
parents: 1621
diff changeset
199 r = t
38bd8ef9ef67 add CandidateTermMatches, unused so far
drewp@bigasterisk.com
parents: 1621
diff changeset
200 log.debug(f'{INDENT*4} implied term value {p=} {t=}')
38bd8ef9ef67 add CandidateTermMatches, unused so far
drewp@bigasterisk.com
parents: 1621
diff changeset
201 elif r != t:
38bd8ef9ef67 add CandidateTermMatches, unused so far
drewp@bigasterisk.com
parents: 1621
diff changeset
202 # (?x c ?x) matched with (a b c) doesn't work
38bd8ef9ef67 add CandidateTermMatches, unused so far
drewp@bigasterisk.com
parents: 1621
diff changeset
203 return None
38bd8ef9ef67 add CandidateTermMatches, unused so far
drewp@bigasterisk.com
parents: 1621
diff changeset
204 return r
38bd8ef9ef67 add CandidateTermMatches, unused so far
drewp@bigasterisk.com
parents: 1621
diff changeset
205
38bd8ef9ef67 add CandidateTermMatches, unused so far
drewp@bigasterisk.com
parents: 1621
diff changeset
206 def lhsStmtsContainingTerm(self):
38bd8ef9ef67 add CandidateTermMatches, unused so far
drewp@bigasterisk.com
parents: 1621
diff changeset
207 # lhs could precompute this
38bd8ef9ef67 add CandidateTermMatches, unused so far
drewp@bigasterisk.com
parents: 1621
diff changeset
208 for lhsStmt in self.lhs.graph:
38bd8ef9ef67 add CandidateTermMatches, unused so far
drewp@bigasterisk.com
parents: 1621
diff changeset
209 if self.term in lhsStmt:
38bd8ef9ef67 add CandidateTermMatches, unused so far
drewp@bigasterisk.com
parents: 1621
diff changeset
210 yield lhsStmt
38bd8ef9ef67 add CandidateTermMatches, unused so far
drewp@bigasterisk.com
parents: 1621
diff changeset
211
38bd8ef9ef67 add CandidateTermMatches, unused so far
drewp@bigasterisk.com
parents: 1621
diff changeset
212 def __iter__(self):
38bd8ef9ef67 add CandidateTermMatches, unused so far
drewp@bigasterisk.com
parents: 1621
diff changeset
213 return iter(self.results)
38bd8ef9ef67 add CandidateTermMatches, unused so far
drewp@bigasterisk.com
parents: 1621
diff changeset
214
38bd8ef9ef67 add CandidateTermMatches, unused so far
drewp@bigasterisk.com
parents: 1621
diff changeset
215
38bd8ef9ef67 add CandidateTermMatches, unused so far
drewp@bigasterisk.com
parents: 1621
diff changeset
216 @dataclass
1607
b21885181e35 more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents: 1605
diff changeset
217 class BoundLhs:
b21885181e35 more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents: 1605
diff changeset
218 lhs: Lhs
1610
6fc48ef4c696 mysteriously lost an important line
drewp@bigasterisk.com
parents: 1609
diff changeset
219 binding: CandidateBinding
1607
b21885181e35 more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents: 1605
diff changeset
220
b21885181e35 more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents: 1605
diff changeset
221 def __post_init__(self):
1613
03ed8c9abd5b forget GRAPH_ID optimization in this case
drewp@bigasterisk.com
parents: 1612
diff changeset
222 self.usedByFuncs = Graph()
1607
b21885181e35 more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents: 1605
diff changeset
223 self._applyFunctions()
b21885181e35 more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents: 1605
diff changeset
224
1616
3a6ed545357f optimization: stream stmts instead of building a Graph
drewp@bigasterisk.com
parents: 1615
diff changeset
225 def lhsStmtsWithoutEvals(self):
3a6ed545357f optimization: stream stmts instead of building a Graph
drewp@bigasterisk.com
parents: 1615
diff changeset
226 for stmt in self.lhs.graph:
3a6ed545357f optimization: stream stmts instead of building a Graph
drewp@bigasterisk.com
parents: 1615
diff changeset
227 if stmt in self.usedByFuncs:
3a6ed545357f optimization: stream stmts instead of building a Graph
drewp@bigasterisk.com
parents: 1615
diff changeset
228 continue
3a6ed545357f optimization: stream stmts instead of building a Graph
drewp@bigasterisk.com
parents: 1615
diff changeset
229 yield stmt
1607
b21885181e35 more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents: 1605
diff changeset
230
b21885181e35 more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents: 1605
diff changeset
231 def _applyFunctions(self):
b21885181e35 more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents: 1605
diff changeset
232 """may grow the binding with some results"""
b21885181e35 more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents: 1605
diff changeset
233 while True:
b21885181e35 more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents: 1605
diff changeset
234 delta = self._applyFunctionsIteration()
b21885181e35 more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents: 1605
diff changeset
235 if delta == 0:
b21885181e35 more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents: 1605
diff changeset
236 break
b21885181e35 more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents: 1605
diff changeset
237
b21885181e35 more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents: 1605
diff changeset
238 def _applyFunctionsIteration(self):
b21885181e35 more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents: 1605
diff changeset
239 before = len(self.binding.binding)
b21885181e35 more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents: 1605
diff changeset
240 delta = 0
b21885181e35 more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents: 1605
diff changeset
241 for ev in self.lhs.evaluations:
b21885181e35 more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents: 1605
diff changeset
242 newBindings, usedGraph = ev.resultBindings(self.binding)
b21885181e35 more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents: 1605
diff changeset
243 self.usedByFuncs += usedGraph
b21885181e35 more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents: 1605
diff changeset
244 self.binding.addNewBindings(newBindings)
1608
f928eb06a4f6 cleaning up inner loop
drewp@bigasterisk.com
parents: 1607
diff changeset
245 delta = len(self.binding.binding) - before
f928eb06a4f6 cleaning up inner loop
drewp@bigasterisk.com
parents: 1607
diff changeset
246 log.debug(f'{INDENT*4} eval rules made {delta} new bindings')
1607
b21885181e35 more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents: 1605
diff changeset
247 return delta
b21885181e35 more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents: 1605
diff changeset
248
b21885181e35 more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents: 1605
diff changeset
249 def verify(self, workingSet: ReadOnlyWorkingSet) -> bool:
b21885181e35 more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents: 1605
diff changeset
250 """Can this bound lhs be true all at once in workingSet?"""
1615
bcfa368e5498 change a Graph.__sub__ to Set.difference in verify() for a big speedup
drewp@bigasterisk.com
parents: 1614
diff changeset
251 rem = cast(Set[Triple], self.lhs.nonStaticRuleStmtsSet.difference(self.usedByFuncs))
bcfa368e5498 change a Graph.__sub__ to Set.difference in verify() for a big speedup
drewp@bigasterisk.com
parents: 1614
diff changeset
252 boundLhs = self.binding.apply(rem)
1607
b21885181e35 more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents: 1605
diff changeset
253
1611
a794a150a89b a little inner-loop performace. same number of iterations, but less time on logging and some stmt filtering
drewp@bigasterisk.com
parents: 1610
diff changeset
254 if log.isEnabledFor(logging.DEBUG):
a794a150a89b a little inner-loop performace. same number of iterations, but less time on logging and some stmt filtering
drewp@bigasterisk.com
parents: 1610
diff changeset
255 boundLhs = list(boundLhs)
a794a150a89b a little inner-loop performace. same number of iterations, but less time on logging and some stmt filtering
drewp@bigasterisk.com
parents: 1610
diff changeset
256 self._logVerifyBanner(boundLhs, workingSet)
1607
b21885181e35 more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents: 1605
diff changeset
257
b21885181e35 more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents: 1605
diff changeset
258 for stmt in boundLhs:
1611
a794a150a89b a little inner-loop performace. same number of iterations, but less time on logging and some stmt filtering
drewp@bigasterisk.com
parents: 1610
diff changeset
259 log.debug(f'{INDENT*4} check for %s', stmt)
1607
b21885181e35 more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents: 1605
diff changeset
260
1611
a794a150a89b a little inner-loop performace. same number of iterations, but less time on logging and some stmt filtering
drewp@bigasterisk.com
parents: 1610
diff changeset
261 if stmt not in workingSet:
1607
b21885181e35 more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents: 1605
diff changeset
262 log.debug(f'{INDENT*5} stmt not known to be true')
b21885181e35 more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents: 1605
diff changeset
263 return False
b21885181e35 more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents: 1605
diff changeset
264 return True
b21885181e35 more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents: 1605
diff changeset
265
1611
a794a150a89b a little inner-loop performace. same number of iterations, but less time on logging and some stmt filtering
drewp@bigasterisk.com
parents: 1610
diff changeset
266 def _logVerifyBanner(self, boundLhs, workingSet: ReadOnlyWorkingSet):
1607
b21885181e35 more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents: 1605
diff changeset
267 log.debug(f'{INDENT*4}/ verify all bindings against this boundLhs:')
b21885181e35 more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents: 1605
diff changeset
268 for stmt in sorted(boundLhs):
b21885181e35 more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents: 1605
diff changeset
269 log.debug(f'{INDENT*4}|{INDENT} {stmt}')
b21885181e35 more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents: 1605
diff changeset
270
b21885181e35 more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents: 1605
diff changeset
271 # log.debug(f'{INDENT*4}| and against this workingSet:')
b21885181e35 more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents: 1605
diff changeset
272 # for stmt in sorted(workingSet):
b21885181e35 more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents: 1605
diff changeset
273 # log.debug(f'{INDENT*4}|{INDENT} {stmt}')
b21885181e35 more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents: 1605
diff changeset
274
b21885181e35 more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents: 1605
diff changeset
275 log.debug(f'{INDENT*4}\\')
b21885181e35 more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents: 1605
diff changeset
276
b21885181e35 more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents: 1605
diff changeset
277
b21885181e35 more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents: 1605
diff changeset
278 @dataclass
b21885181e35 more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents: 1605
diff changeset
279 class Rule:
b21885181e35 more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents: 1605
diff changeset
280 lhsGraph: Graph
b21885181e35 more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents: 1605
diff changeset
281 rhsGraph: Graph
b21885181e35 more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents: 1605
diff changeset
282
b21885181e35 more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents: 1605
diff changeset
283 def __post_init__(self):
b21885181e35 more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents: 1605
diff changeset
284 self.lhs = Lhs(self.lhsGraph)
b21885181e35 more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents: 1605
diff changeset
285
1612
272f78d4671a mark skipped tests. move applyRule into Rule. minor cleanups.
drewp@bigasterisk.com
parents: 1611
diff changeset
286 def applyRule(self, workingSet: Graph, implied: Graph, stats: Dict):
272f78d4671a mark skipped tests. move applyRule into Rule. minor cleanups.
drewp@bigasterisk.com
parents: 1611
diff changeset
287 for bound in self.lhs.findCandidateBindings(ReadOnlyGraphAggregate([workingSet]), stats):
272f78d4671a mark skipped tests. move applyRule into Rule. minor cleanups.
drewp@bigasterisk.com
parents: 1611
diff changeset
288 log.debug(f'{INDENT*3} rule has a working binding:')
272f78d4671a mark skipped tests. move applyRule into Rule. minor cleanups.
drewp@bigasterisk.com
parents: 1611
diff changeset
289
1616
3a6ed545357f optimization: stream stmts instead of building a Graph
drewp@bigasterisk.com
parents: 1615
diff changeset
290 for lhsBoundStmt in bound.binding.apply(bound.lhsStmtsWithoutEvals()):
1620
92f8deb59735 log layout
drewp@bigasterisk.com
parents: 1616
diff changeset
291 log.debug(f'{INDENT*4} adding {lhsBoundStmt=}')
1612
272f78d4671a mark skipped tests. move applyRule into Rule. minor cleanups.
drewp@bigasterisk.com
parents: 1611
diff changeset
292 workingSet.add(lhsBoundStmt)
272f78d4671a mark skipped tests. move applyRule into Rule. minor cleanups.
drewp@bigasterisk.com
parents: 1611
diff changeset
293 for newStmt in bound.binding.apply(self.rhsGraph):
1620
92f8deb59735 log layout
drewp@bigasterisk.com
parents: 1616
diff changeset
294 log.debug(f'{INDENT*4} adding {newStmt=}')
1612
272f78d4671a mark skipped tests. move applyRule into Rule. minor cleanups.
drewp@bigasterisk.com
parents: 1611
diff changeset
295 workingSet.add(newStmt)
272f78d4671a mark skipped tests. move applyRule into Rule. minor cleanups.
drewp@bigasterisk.com
parents: 1611
diff changeset
296 implied.add(newStmt)
272f78d4671a mark skipped tests. move applyRule into Rule. minor cleanups.
drewp@bigasterisk.com
parents: 1611
diff changeset
297
1607
b21885181e35 more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents: 1605
diff changeset
298
1587
9a3a18c494f9 WIP new inferencer. no vars yet.
drewp@bigasterisk.com
parents:
diff changeset
299 class Inference:
9a3a18c494f9 WIP new inferencer. no vars yet.
drewp@bigasterisk.com
parents:
diff changeset
300
9a3a18c494f9 WIP new inferencer. no vars yet.
drewp@bigasterisk.com
parents:
diff changeset
301 def __init__(self) -> None:
1607
b21885181e35 more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents: 1605
diff changeset
302 self.rules = []
1587
9a3a18c494f9 WIP new inferencer. no vars yet.
drewp@bigasterisk.com
parents:
diff changeset
303
9a3a18c494f9 WIP new inferencer. no vars yet.
drewp@bigasterisk.com
parents:
diff changeset
304 def setRules(self, g: ConjunctiveGraph):
1607
b21885181e35 more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents: 1605
diff changeset
305 self.rules: List[Rule] = []
1599
abbf0eb0e640 fix a bug with a slightly moer complicated set of rules
drewp@bigasterisk.com
parents: 1598
diff changeset
306 for stmt in g:
abbf0eb0e640 fix a bug with a slightly moer complicated set of rules
drewp@bigasterisk.com
parents: 1598
diff changeset
307 if stmt[1] == LOG['implies']:
1607
b21885181e35 more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents: 1605
diff changeset
308 self.rules.append(Rule(stmt[0], stmt[2]))
1612
272f78d4671a mark skipped tests. move applyRule into Rule. minor cleanups.
drewp@bigasterisk.com
parents: 1611
diff changeset
309 # other stmts should go to a default working set?
1587
9a3a18c494f9 WIP new inferencer. no vars yet.
drewp@bigasterisk.com
parents:
diff changeset
310
1601
30463df12d89 infer() dumps stats
drewp@bigasterisk.com
parents: 1600
diff changeset
311 @INFER_CALLS.time()
1587
9a3a18c494f9 WIP new inferencer. no vars yet.
drewp@bigasterisk.com
parents:
diff changeset
312 def infer(self, graph: Graph):
9a3a18c494f9 WIP new inferencer. no vars yet.
drewp@bigasterisk.com
parents:
diff changeset
313 """
9a3a18c494f9 WIP new inferencer. no vars yet.
drewp@bigasterisk.com
parents:
diff changeset
314 returns new graph of inferred statements.
9a3a18c494f9 WIP new inferencer. no vars yet.
drewp@bigasterisk.com
parents:
diff changeset
315 """
1626
7b3656867185 metrics on input graph sizes
drewp@bigasterisk.com
parents: 1623
diff changeset
316 n = graph.__len__()
7b3656867185 metrics on input graph sizes
drewp@bigasterisk.com
parents: 1623
diff changeset
317 INFER_GRAPH_SIZE.observe(n)
7b3656867185 metrics on input graph sizes
drewp@bigasterisk.com
parents: 1623
diff changeset
318 log.info(f'{INDENT*0} Begin inference of graph len={n} with rules len={len(self.rules)}:')
1601
30463df12d89 infer() dumps stats
drewp@bigasterisk.com
parents: 1600
diff changeset
319 startTime = time.time()
1607
b21885181e35 more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents: 1605
diff changeset
320 stats: Dict[str, Union[int, float]] = defaultdict(lambda: 0)
1589
5c1055be3c36 WIP more debugging, working towards bnode-matching support
drewp@bigasterisk.com
parents: 1588
diff changeset
321 # 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
322 workingSet = Graph()
b0df43d5494c big rewrite- more classes, smaller methods, more typesafe, all current tests passing
drewp@bigasterisk.com
parents: 1592
diff changeset
323 workingSet += graph
1587
9a3a18c494f9 WIP new inferencer. no vars yet.
drewp@bigasterisk.com
parents:
diff changeset
324
1594
e58bcfa66093 cleanups and a few fixed cases
drewp@bigasterisk.com
parents: 1593
diff changeset
325 # just the statements that came from RHS's of rules that fired.
1587
9a3a18c494f9 WIP new inferencer. no vars yet.
drewp@bigasterisk.com
parents:
diff changeset
326 implied = ConjunctiveGraph()
9a3a18c494f9 WIP new inferencer. no vars yet.
drewp@bigasterisk.com
parents:
diff changeset
327
9a3a18c494f9 WIP new inferencer. no vars yet.
drewp@bigasterisk.com
parents:
diff changeset
328 bailout_iterations = 100
9a3a18c494f9 WIP new inferencer. no vars yet.
drewp@bigasterisk.com
parents:
diff changeset
329 delta = 1
1607
b21885181e35 more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents: 1605
diff changeset
330 stats['initWorkingSet'] = cast(int, workingSet.__len__())
1587
9a3a18c494f9 WIP new inferencer. no vars yet.
drewp@bigasterisk.com
parents:
diff changeset
331 while delta > 0 and bailout_iterations > 0:
1620
92f8deb59735 log layout
drewp@bigasterisk.com
parents: 1616
diff changeset
332 log.debug('')
1597
387a9cb66517 logging adjustments
drewp@bigasterisk.com
parents: 1596
diff changeset
333 log.info(f'{INDENT*1}*iteration ({bailout_iterations} left)')
1587
9a3a18c494f9 WIP new inferencer. no vars yet.
drewp@bigasterisk.com
parents:
diff changeset
334 bailout_iterations -= 1
9a3a18c494f9 WIP new inferencer. no vars yet.
drewp@bigasterisk.com
parents:
diff changeset
335 delta = -len(implied)
1607
b21885181e35 more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents: 1605
diff changeset
336 self._iterateAllRules(workingSet, implied, stats)
1587
9a3a18c494f9 WIP new inferencer. no vars yet.
drewp@bigasterisk.com
parents:
diff changeset
337 delta += len(implied)
1607
b21885181e35 more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents: 1605
diff changeset
338 stats['iterations'] += 1
1597
387a9cb66517 logging adjustments
drewp@bigasterisk.com
parents: 1596
diff changeset
339 log.info(f'{INDENT*2} this inference iteration added {delta} more implied stmts')
1607
b21885181e35 more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents: 1605
diff changeset
340 stats['timeSpent'] = round(time.time() - startTime, 3)
b21885181e35 more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents: 1605
diff changeset
341 stats['impliedStmts'] = len(implied)
b21885181e35 more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents: 1605
diff changeset
342 log.info(f'{INDENT*0} Inference done {dict(stats)}. Implied:')
1587
9a3a18c494f9 WIP new inferencer. no vars yet.
drewp@bigasterisk.com
parents:
diff changeset
343 for st in implied:
1597
387a9cb66517 logging adjustments
drewp@bigasterisk.com
parents: 1596
diff changeset
344 log.info(f'{INDENT*1} {st}')
1587
9a3a18c494f9 WIP new inferencer. no vars yet.
drewp@bigasterisk.com
parents:
diff changeset
345 return implied
9a3a18c494f9 WIP new inferencer. no vars yet.
drewp@bigasterisk.com
parents:
diff changeset
346
1607
b21885181e35 more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents: 1605
diff changeset
347 def _iterateAllRules(self, workingSet: Graph, implied: Graph, stats):
1612
272f78d4671a mark skipped tests. move applyRule into Rule. minor cleanups.
drewp@bigasterisk.com
parents: 1611
diff changeset
348 for i, rule in enumerate(self.rules):
272f78d4671a mark skipped tests. move applyRule into Rule. minor cleanups.
drewp@bigasterisk.com
parents: 1611
diff changeset
349 self._logRuleApplicationHeader(workingSet, i, rule)
272f78d4671a mark skipped tests. move applyRule into Rule. minor cleanups.
drewp@bigasterisk.com
parents: 1611
diff changeset
350 rule.applyRule(workingSet, implied, stats)
1587
9a3a18c494f9 WIP new inferencer. no vars yet.
drewp@bigasterisk.com
parents:
diff changeset
351
1607
b21885181e35 more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents: 1605
diff changeset
352 def _logRuleApplicationHeader(self, workingSet, i, r: Rule):
1594
e58bcfa66093 cleanups and a few fixed cases
drewp@bigasterisk.com
parents: 1593
diff changeset
353 if not log.isEnabledFor(logging.DEBUG):
e58bcfa66093 cleanups and a few fixed cases
drewp@bigasterisk.com
parents: 1593
diff changeset
354 return
e58bcfa66093 cleanups and a few fixed cases
drewp@bigasterisk.com
parents: 1593
diff changeset
355
e58bcfa66093 cleanups and a few fixed cases
drewp@bigasterisk.com
parents: 1593
diff changeset
356 log.debug('')
e58bcfa66093 cleanups and a few fixed cases
drewp@bigasterisk.com
parents: 1593
diff changeset
357 log.debug(f'{INDENT*2} workingSet:')
1597
387a9cb66517 logging adjustments
drewp@bigasterisk.com
parents: 1596
diff changeset
358 for j, stmt in enumerate(sorted(workingSet)):
387a9cb66517 logging adjustments
drewp@bigasterisk.com
parents: 1596
diff changeset
359 log.debug(f'{INDENT*3} ({j}) {stmt}')
1594
e58bcfa66093 cleanups and a few fixed cases
drewp@bigasterisk.com
parents: 1593
diff changeset
360
e58bcfa66093 cleanups and a few fixed cases
drewp@bigasterisk.com
parents: 1593
diff changeset
361 log.debug('')
e58bcfa66093 cleanups and a few fixed cases
drewp@bigasterisk.com
parents: 1593
diff changeset
362 log.debug(f'{INDENT*2}-applying rule {i}')
1607
b21885181e35 more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents: 1605
diff changeset
363 log.debug(f'{INDENT*3} rule def lhs: {graphDump(r.lhsGraph)}')
b21885181e35 more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents: 1605
diff changeset
364 log.debug(f'{INDENT*3} rule def rhs: {graphDump(r.rhsGraph)}')
1594
e58bcfa66093 cleanups and a few fixed cases
drewp@bigasterisk.com
parents: 1593
diff changeset
365
1588
0757fafbfdab WIP inferencer - partial var and function support
drewp@bigasterisk.com
parents: 1587
diff changeset
366
1590
327202020892 WIP inference- getting into more degenerate test cases
drewp@bigasterisk.com
parents: 1589
diff changeset
367 def graphDump(g: Union[Graph, List[Triple]]):
327202020892 WIP inference- getting into more degenerate test cases
drewp@bigasterisk.com
parents: 1589
diff changeset
368 if not isinstance(g, Graph):
1607
b21885181e35 more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents: 1605
diff changeset
369 log.warning(f"it's a {type(g)}")
1590
327202020892 WIP inference- getting into more degenerate test cases
drewp@bigasterisk.com
parents: 1589
diff changeset
370 g2 = Graph()
1594
e58bcfa66093 cleanups and a few fixed cases
drewp@bigasterisk.com
parents: 1593
diff changeset
371 g2 += g
1590
327202020892 WIP inference- getting into more degenerate test cases
drewp@bigasterisk.com
parents: 1589
diff changeset
372 g = g2
1589
5c1055be3c36 WIP more debugging, working towards bnode-matching support
drewp@bigasterisk.com
parents: 1588
diff changeset
373 g.bind('', ROOM)
5c1055be3c36 WIP more debugging, working towards bnode-matching support
drewp@bigasterisk.com
parents: 1588
diff changeset
374 g.bind('ex', Namespace('http://example.com/'))
5c1055be3c36 WIP more debugging, working towards bnode-matching support
drewp@bigasterisk.com
parents: 1588
diff changeset
375 lines = cast(bytes, g.serialize(format='n3')).decode('utf8').splitlines()
1612
272f78d4671a mark skipped tests. move applyRule into Rule. minor cleanups.
drewp@bigasterisk.com
parents: 1611
diff changeset
376 lines = [line.strip() for line in lines if not line.startswith('@prefix')]
1589
5c1055be3c36 WIP more debugging, working towards bnode-matching support
drewp@bigasterisk.com
parents: 1588
diff changeset
377 return ' '.join(lines)
5c1055be3c36 WIP more debugging, working towards bnode-matching support
drewp@bigasterisk.com
parents: 1588
diff changeset
378
5c1055be3c36 WIP more debugging, working towards bnode-matching support
drewp@bigasterisk.com
parents: 1588
diff changeset
379
1600
89a50242cb5e cleanup internal names, imports
drewp@bigasterisk.com
parents: 1599
diff changeset
380 def _organize(candidateTermMatches: Dict[BindableTerm, Set[Node]]) -> Tuple[List[BindableTerm], List[List[Node]]]:
1592
d7b66234064b pure reordering of funcs to make the next diffs smaller
drewp@bigasterisk.com
parents: 1591
diff changeset
381 items = list(candidateTermMatches.items())
d7b66234064b pure reordering of funcs to make the next diffs smaller
drewp@bigasterisk.com
parents: 1591
diff changeset
382 items.sort()
d7b66234064b pure reordering of funcs to make the next diffs smaller
drewp@bigasterisk.com
parents: 1591
diff changeset
383 orderedVars: List[BindableTerm] = []
d7b66234064b pure reordering of funcs to make the next diffs smaller
drewp@bigasterisk.com
parents: 1591
diff changeset
384 orderedValueSets: List[List[Node]] = []
d7b66234064b pure reordering of funcs to make the next diffs smaller
drewp@bigasterisk.com
parents: 1591
diff changeset
385 for v, vals in items:
d7b66234064b pure reordering of funcs to make the next diffs smaller
drewp@bigasterisk.com
parents: 1591
diff changeset
386 orderedVars.append(v)
d7b66234064b pure reordering of funcs to make the next diffs smaller
drewp@bigasterisk.com
parents: 1591
diff changeset
387 orderedValues: List[Node] = list(vals)
d7b66234064b pure reordering of funcs to make the next diffs smaller
drewp@bigasterisk.com
parents: 1591
diff changeset
388 orderedValues.sort(key=str)
d7b66234064b pure reordering of funcs to make the next diffs smaller
drewp@bigasterisk.com
parents: 1591
diff changeset
389 orderedValueSets.append(orderedValues)
1588
0757fafbfdab WIP inferencer - partial var and function support
drewp@bigasterisk.com
parents: 1587
diff changeset
390
1592
d7b66234064b pure reordering of funcs to make the next diffs smaller
drewp@bigasterisk.com
parents: 1591
diff changeset
391 return orderedVars, orderedValueSets