Mercurial > code > home > repos > homeauto
annotate service/mqtt_to_rdf/inference.py @ 1637:ec3f98d0c1d8
refactor rules eval
author | drewp@bigasterisk.com |
---|---|
date | Mon, 13 Sep 2021 01:36:06 -0700 |
parents | 3252bdc284bc |
children | 0ba1625037ae |
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 |
1634
ba59cfc3c747
hack math:sum in there. Test suite is passing except some slow performers
drewp@bigasterisk.com
parents:
1633
diff
changeset
|
10 from typing import (Dict, Iterator, List, Optional, Sequence, Set, Tuple, Union, cast) |
1588
0757fafbfdab
WIP inferencer - partial var and function support
drewp@bigasterisk.com
parents:
1587
diff
changeset
|
11 |
1631
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
12 from prometheus_client import Histogram, Summary |
1634
ba59cfc3c747
hack math:sum in there. Test suite is passing except some slow performers
drewp@bigasterisk.com
parents:
1633
diff
changeset
|
13 from rdflib import RDF, BNode, Graph, Namespace |
1589
5c1055be3c36
WIP more debugging, working towards bnode-matching support
drewp@bigasterisk.com
parents:
1588
diff
changeset
|
14 from rdflib.graph import ConjunctiveGraph, ReadOnlyGraphAggregate |
1637 | 15 from rdflib.term import Node, Variable |
1587 | 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 |
1637 | 18 from inference_types import BindingUnknown, ReadOnlyWorkingSet, Triple |
19 from lhs_evaluation import functionsFor, lhsStmtsUsedByFuncs | |
1605 | 20 |
1587 | 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 | 23 |
1626 | 24 INFER_CALLS = Summary('inference_infer_calls', 'calls') |
25 INFER_GRAPH_SIZE = Histogram('inference_graph_size', 'statements', buckets=[2**x for x in range(2, 20, 2)]) | |
1587 | 26 |
27 ROOM = Namespace("http://projects.bigasterisk.com/room/") | |
28 LOG = Namespace('http://www.w3.org/2000/10/swap/log#') | |
29 MATH = Namespace('http://www.w3.org/2000/10/swap/math#') | |
30 | |
31 | |
1631
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
32 def stmtTemplate(stmt: Triple) -> Tuple[Optional[Node], Optional[Node], Optional[Node]]: |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
33 return ( |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
34 None if isinstance(stmt[0], (Variable, BNode)) else stmt[0], |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
35 None if isinstance(stmt[1], (Variable, BNode)) else stmt[1], |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
36 None if isinstance(stmt[2], (Variable, BNode)) else stmt[2], |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
37 ) |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
38 |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
39 |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
40 class NoOptions(ValueError): |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
41 """stmtlooper has no possibilites to add to the binding; the whole rule must therefore not apply""" |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
42 |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
43 |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
44 class Inconsistent(ValueError): |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
45 """adding this stmt would be inconsistent with an existing binding""" |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
46 |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
47 |
1632 | 48 _stmtLooperShortId = itertools.count() |
49 | |
50 | |
1631
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
51 @dataclass |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
52 class StmtLooper: |
1632 | 53 """given one LHS stmt, iterate through the possible matches for it, |
54 returning what bindings they would imply. Only distinct bindings are | |
55 returned. The bindings build on any `prev` StmtLooper's results. | |
56 | |
57 This iterator is restartable.""" | |
1631
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
58 lhsStmt: Triple |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
59 prev: Optional['StmtLooper'] |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
60 workingSet: ReadOnlyWorkingSet |
1634
ba59cfc3c747
hack math:sum in there. Test suite is passing except some slow performers
drewp@bigasterisk.com
parents:
1633
diff
changeset
|
61 parent: 'Lhs' # just for lhs.graph, really |
1631
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
62 |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
63 def __repr__(self): |
1632 | 64 return f'StmtLooper{self._shortId}({graphDump([self.lhsStmt])} {"<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
|
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 __post_init__(self): |
1632 | 67 self._shortId = next(_stmtLooperShortId) |
1631
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
68 self._myWorkingSetMatches = self._myMatches(self.workingSet) |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
69 |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
70 self._current = CandidateBinding({}) |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
71 self._pastEnd = False |
1635
22d481f0a924
refactor: use CandidateBinding throughout, not loose dicts
drewp@bigasterisk.com
parents:
1634
diff
changeset
|
72 self._seenBindings: List[CandidateBinding] = [] |
1631
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
73 self.restart() |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
74 |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
75 def _myMatches(self, g: Graph) -> List[Triple]: |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
76 template = stmtTemplate(self.lhsStmt) |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
77 |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
78 stmts = sorted(cast(Iterator[Triple], list(g.triples(template)))) |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
79 # plus new lhs possibilties... |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
80 # log.debug(f'{INDENT*6} {self} find {len(stmts)=} in {len(self.workingSet)=}') |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
81 |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
82 return stmts |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
83 |
1635
22d481f0a924
refactor: use CandidateBinding throughout, not loose dicts
drewp@bigasterisk.com
parents:
1634
diff
changeset
|
84 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
|
85 if not self.prev or self.prev.pastEnd(): |
1635
22d481f0a924
refactor: use CandidateBinding throughout, not loose dicts
drewp@bigasterisk.com
parents:
1634
diff
changeset
|
86 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
|
87 |
1635
22d481f0a924
refactor: use CandidateBinding throughout, not loose dicts
drewp@bigasterisk.com
parents:
1634
diff
changeset
|
88 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
|
89 |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
90 def advance(self): |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
91 """update to a new set of bindings we haven't seen (since last restart), or go into pastEnd mode""" |
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
|
92 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
|
93 raise NotImplementedError('need restart') |
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
|
94 log.debug('') |
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
|
95 augmentedWorkingSet: Sequence[Triple] = [] |
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
|
96 if self.prev is None: |
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
|
97 augmentedWorkingSet = self._myWorkingSetMatches |
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 else: |
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 augmentedWorkingSet = list(self.prev.currentBinding().apply(self._myWorkingSetMatches, |
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
|
100 returnBoundStatementsOnly=False)) |
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
|
101 |
1634
ba59cfc3c747
hack math:sum in there. Test suite is passing except some slow performers
drewp@bigasterisk.com
parents:
1633
diff
changeset
|
102 log.debug(f'{INDENT*6} {self}.advance has {augmentedWorkingSet=}') |
ba59cfc3c747
hack math:sum in there. Test suite is passing except some slow performers
drewp@bigasterisk.com
parents:
1633
diff
changeset
|
103 |
ba59cfc3c747
hack math:sum in there. Test suite is passing except some slow performers
drewp@bigasterisk.com
parents:
1633
diff
changeset
|
104 if self._advanceWithPlainMatches(augmentedWorkingSet): |
ba59cfc3c747
hack math:sum in there. Test suite is passing except some slow performers
drewp@bigasterisk.com
parents:
1633
diff
changeset
|
105 return |
ba59cfc3c747
hack math:sum in there. Test suite is passing except some slow performers
drewp@bigasterisk.com
parents:
1633
diff
changeset
|
106 |
ba59cfc3c747
hack math:sum in there. Test suite is passing except some slow performers
drewp@bigasterisk.com
parents:
1633
diff
changeset
|
107 curBind = self.prev.currentBinding() if self.prev else CandidateBinding({}) |
ba59cfc3c747
hack math:sum in there. Test suite is passing except some slow performers
drewp@bigasterisk.com
parents:
1633
diff
changeset
|
108 [lhsStmtBound] = curBind.apply([self.lhsStmt], returnBoundStatementsOnly=False) |
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 fullWorkingSet = self.workingSet + self.parent.graph |
ba59cfc3c747
hack math:sum in there. Test suite is passing except some slow performers
drewp@bigasterisk.com
parents:
1633
diff
changeset
|
111 boundFullWorkingSet = list(curBind.apply(fullWorkingSet, returnBoundStatementsOnly=False)) |
ba59cfc3c747
hack math:sum in there. Test suite is passing except some slow performers
drewp@bigasterisk.com
parents:
1633
diff
changeset
|
112 log.debug(f'{fullWorkingSet.__len__()=} {len(boundFullWorkingSet)=}') |
ba59cfc3c747
hack math:sum in there. Test suite is passing except some slow performers
drewp@bigasterisk.com
parents:
1633
diff
changeset
|
113 |
ba59cfc3c747
hack math:sum in there. Test suite is passing except some slow performers
drewp@bigasterisk.com
parents:
1633
diff
changeset
|
114 if self._advanceWithFunctions(augmentedWorkingSet, boundFullWorkingSet, lhsStmtBound): |
ba59cfc3c747
hack math:sum in there. Test suite is passing except some slow performers
drewp@bigasterisk.com
parents:
1633
diff
changeset
|
115 return |
ba59cfc3c747
hack math:sum in there. Test suite is passing except some slow performers
drewp@bigasterisk.com
parents:
1633
diff
changeset
|
116 |
ba59cfc3c747
hack math:sum in there. Test suite is passing except some slow performers
drewp@bigasterisk.com
parents:
1633
diff
changeset
|
117 log.debug(f'{INDENT*6} {self} is past end') |
ba59cfc3c747
hack math:sum in there. Test suite is passing except some slow performers
drewp@bigasterisk.com
parents:
1633
diff
changeset
|
118 self._pastEnd = True |
ba59cfc3c747
hack math:sum in there. Test suite is passing except some slow performers
drewp@bigasterisk.com
parents:
1633
diff
changeset
|
119 |
ba59cfc3c747
hack math:sum in there. Test suite is passing except some slow performers
drewp@bigasterisk.com
parents:
1633
diff
changeset
|
120 def _advanceWithPlainMatches(self, augmentedWorkingSet: Sequence[Triple]) -> bool: |
ba59cfc3c747
hack math:sum in there. Test suite is passing except some slow performers
drewp@bigasterisk.com
parents:
1633
diff
changeset
|
121 log.debug(f'{INDENT*7} {self} mines {len(augmentedWorkingSet)} matching augmented statements') |
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
|
122 for s in augmentedWorkingSet: |
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
|
123 log.debug(f'{INDENT*7} {s}') |
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
|
124 |
1637 | 125 for stmt 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
|
126 try: |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
127 outBinding = self._totalBindingIfThisStmtWereTrue(stmt) |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
128 except Inconsistent: |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
129 log.debug(f'{INDENT*7} {self} - {stmt} would be inconsistent with prev bindings') |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
130 continue |
1632 | 131 |
1634
ba59cfc3c747
hack math:sum in there. Test suite is passing except some slow performers
drewp@bigasterisk.com
parents:
1633
diff
changeset
|
132 log.debug(f'{INDENT*7} {outBinding=} {self._seenBindings=}') |
1635
22d481f0a924
refactor: use CandidateBinding throughout, not loose dicts
drewp@bigasterisk.com
parents:
1634
diff
changeset
|
133 if outBinding not in self._seenBindings: |
22d481f0a924
refactor: use CandidateBinding throughout, not loose dicts
drewp@bigasterisk.com
parents:
1634
diff
changeset
|
134 self._seenBindings.append(outBinding.copy()) |
1631
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
135 self._current = outBinding |
1632 | 136 log.debug(f'{INDENT*7} new binding from {self} -> {outBinding}') |
1634
ba59cfc3c747
hack math:sum in there. Test suite is passing except some slow performers
drewp@bigasterisk.com
parents:
1633
diff
changeset
|
137 return True |
ba59cfc3c747
hack math:sum in there. Test suite is passing except some slow performers
drewp@bigasterisk.com
parents:
1633
diff
changeset
|
138 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
|
139 |
1634
ba59cfc3c747
hack math:sum in there. Test suite is passing except some slow performers
drewp@bigasterisk.com
parents:
1633
diff
changeset
|
140 def _advanceWithFunctions(self, augmentedWorkingSet: Sequence[Triple], boundFullWorkingSet, lhsStmtBound) -> bool: |
1637 | 141 pred: Node = self.lhsStmt[1] |
1632 | 142 |
1637 | 143 for functionType in functionsFor(pred): |
144 fn = functionType(self.lhsStmt, self.parent.graph) | |
1634
ba59cfc3c747
hack math:sum in there. Test suite is passing except some slow performers
drewp@bigasterisk.com
parents:
1633
diff
changeset
|
145 try: |
1637 | 146 out = fn.bind(self._prevBindings()) |
147 except BindingUnknown: | |
1634
ba59cfc3c747
hack math:sum in there. Test suite is passing except some slow performers
drewp@bigasterisk.com
parents:
1633
diff
changeset
|
148 pass |
ba59cfc3c747
hack math:sum in there. Test suite is passing except some slow performers
drewp@bigasterisk.com
parents:
1633
diff
changeset
|
149 else: |
1637 | 150 if out is not None: |
151 binding: CandidateBinding = self._prevBindings().copy() | |
152 binding.addNewBindings(out) | |
153 if binding not in self._seenBindings: | |
154 self._seenBindings.append(binding) | |
155 self._current = binding | |
156 log.debug(f'{INDENT*7} new binding from {self} -> {binding}') | |
157 return True | |
1634
ba59cfc3c747
hack math:sum in there. Test suite is passing except some slow performers
drewp@bigasterisk.com
parents:
1633
diff
changeset
|
158 |
ba59cfc3c747
hack math:sum in there. Test suite is passing except some slow performers
drewp@bigasterisk.com
parents:
1633
diff
changeset
|
159 return False |
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 def _boundOperands(self, operands) -> List[Node]: |
1635
22d481f0a924
refactor: use CandidateBinding throughout, not loose dicts
drewp@bigasterisk.com
parents:
1634
diff
changeset
|
162 pb: CandidateBinding = self._prevBindings() |
1634
ba59cfc3c747
hack math:sum in there. Test suite is passing except some slow performers
drewp@bigasterisk.com
parents:
1633
diff
changeset
|
163 |
ba59cfc3c747
hack math:sum in there. Test suite is passing except some slow performers
drewp@bigasterisk.com
parents:
1633
diff
changeset
|
164 boundOperands: List[Node] = [] |
ba59cfc3c747
hack math:sum in there. Test suite is passing except some slow performers
drewp@bigasterisk.com
parents:
1633
diff
changeset
|
165 for op in operands: |
ba59cfc3c747
hack math:sum in there. Test suite is passing except some slow performers
drewp@bigasterisk.com
parents:
1633
diff
changeset
|
166 if isinstance(op, (Variable, BNode)): |
1635
22d481f0a924
refactor: use CandidateBinding throughout, not loose dicts
drewp@bigasterisk.com
parents:
1634
diff
changeset
|
167 boundOperands.append(pb.applyTerm(op)) |
1634
ba59cfc3c747
hack math:sum in there. Test suite is passing except some slow performers
drewp@bigasterisk.com
parents:
1633
diff
changeset
|
168 else: |
ba59cfc3c747
hack math:sum in there. Test suite is passing except some slow performers
drewp@bigasterisk.com
parents:
1633
diff
changeset
|
169 boundOperands.append(op) |
ba59cfc3c747
hack math:sum in there. Test suite is passing except some slow performers
drewp@bigasterisk.com
parents:
1633
diff
changeset
|
170 return boundOperands |
1631
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
171 |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
172 def _totalBindingIfThisStmtWereTrue(self, newStmt: Triple) -> CandidateBinding: |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
173 outBinding = self._prevBindings().copy() |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
174 for rt, ct in zip(self.lhsStmt, newStmt): |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
175 if isinstance(rt, (Variable, BNode)): |
1635
22d481f0a924
refactor: use CandidateBinding throughout, not loose dicts
drewp@bigasterisk.com
parents:
1634
diff
changeset
|
176 if outBinding.contains(rt) and outBinding.applyTerm(rt) != ct: |
1634
ba59cfc3c747
hack math:sum in there. Test suite is passing except some slow performers
drewp@bigasterisk.com
parents:
1633
diff
changeset
|
177 raise Inconsistent(f'{rt=} {ct=} {outBinding=}') |
1635
22d481f0a924
refactor: use CandidateBinding throughout, not loose dicts
drewp@bigasterisk.com
parents:
1634
diff
changeset
|
178 outBinding.addNewBindings(CandidateBinding({rt: ct})) |
22d481f0a924
refactor: use CandidateBinding throughout, not loose dicts
drewp@bigasterisk.com
parents:
1634
diff
changeset
|
179 return outBinding |
1631
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
180 |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
181 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
|
182 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
|
183 raise NotImplementedError() |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
184 return self._current |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
185 |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
186 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
|
187 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
|
188 |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
189 def restart(self): |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
190 self._pastEnd = False |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
191 self._seenBindings = [] |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
192 self.advance() |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
193 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
|
194 raise NoOptions() |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
195 |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
196 |
1594 | 197 @dataclass |
1593
b0df43d5494c
big rewrite- more classes, smaller methods, more typesafe, all current tests passing
drewp@bigasterisk.com
parents:
1592
diff
changeset
|
198 class Lhs: |
1594 | 199 graph: Graph |
200 | |
201 def __post_init__(self): | |
1636 | 202 pass |
1602 | 203 |
1609
34f2817320cc
new tests for a smaller part of the inner loop
drewp@bigasterisk.com
parents:
1608
diff
changeset
|
204 def __repr__(self): |
34f2817320cc
new tests for a smaller part of the inner loop
drewp@bigasterisk.com
parents:
1608
diff
changeset
|
205 return f"Lhs({graphDump(self.graph)})" |
34f2817320cc
new tests for a smaller part of the inner loop
drewp@bigasterisk.com
parents:
1608
diff
changeset
|
206 |
1621 | 207 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
|
208 """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
|
209 from LHS""" |
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
|
210 if self.graph.__len__() == 0: |
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
|
211 # 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
|
212 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
|
213 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
|
214 |
1631
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
215 log.debug(f'{INDENT*4} build new StmtLooper stack') |
1593
b0df43d5494c
big rewrite- more classes, smaller methods, more typesafe, all current tests passing
drewp@bigasterisk.com
parents:
1592
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 try: |
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
|
218 stmtStack = self._assembleRings(knownTrue) |
1631
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
219 except NoOptions: |
1632 | 220 log.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
|
221 return |
1632 | 222 self._debugStmtStack('initial odometer', stmtStack) |
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
|
223 self._assertAllRingsAreValid(stmtStack) |
1631
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
224 |
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
|
225 lastRing = stmtStack[-1] |
1631
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
226 iterCount = 0 |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
227 while True: |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
228 iterCount += 1 |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
229 if iterCount > 10: |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
230 raise ValueError('stuck') |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
231 |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
232 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
|
233 |
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
|
234 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
|
235 |
1632 | 236 self._debugStmtStack('odometer', stmtStack) |
1607
b21885181e35
more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents:
1605
diff
changeset
|
237 |
1631
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
238 done = self._advanceAll(stmtStack) |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
239 |
1632 | 240 self._debugStmtStack('odometer after ({done=})', stmtStack) |
1631
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
241 |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
242 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
|
243 if done: |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
244 break |
1607
b21885181e35
more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents:
1605
diff
changeset
|
245 |
1632 | 246 def _debugStmtStack(self, label, stmtStack): |
247 log.debug(f'{INDENT*5} {label}:') | |
248 for l in stmtStack: | |
249 log.debug(f'{INDENT*6} {l} curbind={l.currentBinding() if not l.pastEnd() else "<end>"}') | |
250 | |
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
|
251 def _assembleRings(self, knownTrue: ReadOnlyWorkingSet) -> List[StmtLooper]: |
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
|
252 """make StmtLooper for each stmt in our LHS graph, but do it in a way that they all |
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
|
253 start out valid (or else 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
|
254 |
1637 | 255 usedByFuncs: Set[Triple] = lhsStmtsUsedByFuncs(self.graph) |
1634
ba59cfc3c747
hack math:sum in there. Test suite is passing except some slow performers
drewp@bigasterisk.com
parents:
1633
diff
changeset
|
256 |
1637 | 257 stmtsToAdd = list(self.graph - usedByFuncs) |
1634
ba59cfc3c747
hack math:sum in there. Test suite is passing except some slow performers
drewp@bigasterisk.com
parents:
1633
diff
changeset
|
258 |
ba59cfc3c747
hack math:sum in there. Test suite is passing except some slow performers
drewp@bigasterisk.com
parents:
1633
diff
changeset
|
259 # sort them by variable dependencies; don't just try all perms! |
ba59cfc3c747
hack math:sum in there. Test suite is passing except some slow performers
drewp@bigasterisk.com
parents:
1633
diff
changeset
|
260 def lightSortKey(stmt): # Not this. Though it helps performance on the big rdf list cases. |
ba59cfc3c747
hack math:sum in there. Test suite is passing except some slow performers
drewp@bigasterisk.com
parents:
1633
diff
changeset
|
261 (s, p, o) = stmt |
ba59cfc3c747
hack math:sum in there. Test suite is passing except some slow performers
drewp@bigasterisk.com
parents:
1633
diff
changeset
|
262 return p == MATH['sum'], p, s, o |
ba59cfc3c747
hack math:sum in there. Test suite is passing except some slow performers
drewp@bigasterisk.com
parents:
1633
diff
changeset
|
263 |
ba59cfc3c747
hack math:sum in there. Test suite is passing except some slow performers
drewp@bigasterisk.com
parents:
1633
diff
changeset
|
264 stmtsToAdd.sort(key=lightSortKey) |
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
|
265 |
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
|
266 for perm in itertools.permutations(stmtsToAdd): |
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
|
267 stmtStack: List[StmtLooper] = [] |
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
|
268 prev: Optional[StmtLooper] = None |
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
|
269 log.debug(f'{INDENT*5} try stmts in this order: {" -> ".join(graphDump([p]) for p in perm)}') |
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
|
270 |
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
|
271 for s in perm: |
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
|
272 try: |
1634
ba59cfc3c747
hack math:sum in there. Test suite is passing except some slow performers
drewp@bigasterisk.com
parents:
1633
diff
changeset
|
273 elem = StmtLooper(s, prev, knownTrue, parent=self) |
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
|
274 except 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
|
275 log.debug(f'{INDENT*6} permutation didnt work, try another') |
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
|
276 break |
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
|
277 stmtStack.append(elem) |
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
|
278 prev = stmtStack[-1] |
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
|
279 else: |
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
|
280 return stmtStack |
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
|
281 log.debug(f'{INDENT*6} no perms worked- rule cannot match anything') |
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
|
282 |
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
|
283 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
|
284 |
1631
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
285 def _advanceAll(self, stmtStack: List[StmtLooper]) -> bool: |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
286 carry = True # 1st elem always must advance |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
287 for i, ring in enumerate(stmtStack): |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
288 # 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
|
289 if carry: |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
290 log.debug(f'{INDENT*5} advanceAll [{i}] {ring} carry/advance') |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
291 ring.advance() |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
292 carry = False |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
293 if ring.pastEnd(): |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
294 if ring is stmtStack[-1]: |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
295 log.debug(f'{INDENT*5} advanceAll [{i}] {ring} says we done') |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
296 return True |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
297 log.debug(f'{INDENT*5} advanceAll [{i}] {ring} restart') |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
298 ring.restart() |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
299 carry = True |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
300 return False |
1607
b21885181e35
more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents:
1605
diff
changeset
|
301 |
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
|
302 def _assertAllRingsAreValid(self, stmtStack): |
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
|
303 if any(ring.pastEnd() for ring in stmtStack): # this is an unexpected debug assertion |
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
|
304 log.debug(f'{INDENT*5} some rings started at pastEnd {stmtStack}') |
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
|
305 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
|
306 |
1592
d7b66234064b
pure reordering of funcs to make the next diffs smaller
drewp@bigasterisk.com
parents:
1591
diff
changeset
|
307 |
1622
38bd8ef9ef67
add CandidateTermMatches, unused so far
drewp@bigasterisk.com
parents:
1621
diff
changeset
|
308 @dataclass |
1607
b21885181e35
more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents:
1605
diff
changeset
|
309 class BoundLhs: |
b21885181e35
more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents:
1605
diff
changeset
|
310 lhs: Lhs |
1610 | 311 binding: CandidateBinding |
1607
b21885181e35
more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents:
1605
diff
changeset
|
312 |
b21885181e35
more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents:
1605
diff
changeset
|
313 |
b21885181e35
more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents:
1605
diff
changeset
|
314 @dataclass |
b21885181e35
more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents:
1605
diff
changeset
|
315 class Rule: |
b21885181e35
more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents:
1605
diff
changeset
|
316 lhsGraph: Graph |
b21885181e35
more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents:
1605
diff
changeset
|
317 rhsGraph: Graph |
1632 | 318 |
1607
b21885181e35
more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents:
1605
diff
changeset
|
319 def __post_init__(self): |
b21885181e35
more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents:
1605
diff
changeset
|
320 self.lhs = Lhs(self.lhsGraph) |
1632 | 321 # |
1631
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
322 self.rhsBnodeMap = {} |
1607
b21885181e35
more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents:
1605
diff
changeset
|
323 |
1612
272f78d4671a
mark skipped tests. move applyRule into Rule. minor cleanups.
drewp@bigasterisk.com
parents:
1611
diff
changeset
|
324 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
|
325 for bound in self.lhs.findCandidateBindings(ReadOnlyGraphAggregate([workingSet]), stats): |
1631
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
326 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
|
327 |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
328 # rhs could have more bnodes, and they just need to be distinct per rule-firing that we do |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
329 existingRhsBnodes = set() |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
330 for stmt in self.rhsGraph: |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
331 for t in stmt: |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
332 if isinstance(t, BNode): |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
333 existingRhsBnodes.add(t) |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
334 # if existingRhsBnodes: |
1632 | 335 # log.debug(f'{INDENT*6} mapping rhs bnodes {existingRhsBnodes} to new ones') |
1631
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
336 |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
337 for b in existingRhsBnodes: |
1612
272f78d4671a
mark skipped tests. move applyRule into Rule. minor cleanups.
drewp@bigasterisk.com
parents:
1611
diff
changeset
|
338 |
1631
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
339 key = tuple(sorted(bound.binding.binding.items())), b |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
340 self.rhsBnodeMap.setdefault(key, BNode()) |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
341 |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
342 bound.binding.addNewBindings(CandidateBinding({b: self.rhsBnodeMap[key]})) |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
343 |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
344 # for lhsBoundStmt in bound.binding.apply(bound.lhsStmtsWithoutEvals()): |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
345 # log.debug(f'{INDENT*6} adding to workingSet {lhsBoundStmt=}') |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
346 # workingSet.add(lhsBoundStmt) |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
347 # log.debug(f'{INDENT*6} rhsGraph is good: {list(self.rhsGraph)}') |
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
348 |
1612
272f78d4671a
mark skipped tests. move applyRule into Rule. minor cleanups.
drewp@bigasterisk.com
parents:
1611
diff
changeset
|
349 for newStmt in bound.binding.apply(self.rhsGraph): |
1631
2c85a4f5dd9c
big rewrite of infer() using statements not variables as the things to iterate over
drewp@bigasterisk.com
parents:
1627
diff
changeset
|
350 # 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
|
351 workingSet.add(newStmt) |
272f78d4671a
mark skipped tests. move applyRule into Rule. minor cleanups.
drewp@bigasterisk.com
parents:
1611
diff
changeset
|
352 implied.add(newStmt) |
272f78d4671a
mark skipped tests. move applyRule into Rule. minor cleanups.
drewp@bigasterisk.com
parents:
1611
diff
changeset
|
353 |
1607
b21885181e35
more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents:
1605
diff
changeset
|
354 |
1587 | 355 class Inference: |
356 | |
357 def __init__(self) -> None: | |
1607
b21885181e35
more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents:
1605
diff
changeset
|
358 self.rules = [] |
1587 | 359 |
360 def setRules(self, g: ConjunctiveGraph): | |
1607
b21885181e35
more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents:
1605
diff
changeset
|
361 self.rules: List[Rule] = [] |
1599
abbf0eb0e640
fix a bug with a slightly moer complicated set of rules
drewp@bigasterisk.com
parents:
1598
diff
changeset
|
362 for stmt in g: |
abbf0eb0e640
fix a bug with a slightly moer complicated set of rules
drewp@bigasterisk.com
parents:
1598
diff
changeset
|
363 if stmt[1] == LOG['implies']: |
1607
b21885181e35
more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents:
1605
diff
changeset
|
364 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
|
365 # other stmts should go to a default working set? |
1587 | 366 |
1601 | 367 @INFER_CALLS.time() |
1587 | 368 def infer(self, graph: Graph): |
369 """ | |
370 returns new graph of inferred statements. | |
371 """ | |
1626 | 372 n = graph.__len__() |
373 INFER_GRAPH_SIZE.observe(n) | |
374 log.info(f'{INDENT*0} Begin inference of graph len={n} with rules len={len(self.rules)}:') | |
1601 | 375 startTime = time.time() |
1607
b21885181e35
more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents:
1605
diff
changeset
|
376 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
|
377 # 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
|
378 workingSet = Graph() |
b0df43d5494c
big rewrite- more classes, smaller methods, more typesafe, all current tests passing
drewp@bigasterisk.com
parents:
1592
diff
changeset
|
379 workingSet += graph |
1587 | 380 |
1594 | 381 # just the statements that came from RHS's of rules that fired. |
1587 | 382 implied = ConjunctiveGraph() |
383 | |
384 bailout_iterations = 100 | |
385 delta = 1 | |
1607
b21885181e35
more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents:
1605
diff
changeset
|
386 stats['initWorkingSet'] = cast(int, workingSet.__len__()) |
1587 | 387 while delta > 0 and bailout_iterations > 0: |
1620 | 388 log.debug('') |
1597 | 389 log.info(f'{INDENT*1}*iteration ({bailout_iterations} left)') |
1587 | 390 bailout_iterations -= 1 |
391 delta = -len(implied) | |
1607
b21885181e35
more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents:
1605
diff
changeset
|
392 self._iterateAllRules(workingSet, implied, stats) |
1587 | 393 delta += len(implied) |
1607
b21885181e35
more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents:
1605
diff
changeset
|
394 stats['iterations'] += 1 |
1597 | 395 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
|
396 stats['timeSpent'] = round(time.time() - startTime, 3) |
b21885181e35
more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents:
1605
diff
changeset
|
397 stats['impliedStmts'] = len(implied) |
b21885181e35
more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents:
1605
diff
changeset
|
398 log.info(f'{INDENT*0} Inference done {dict(stats)}. Implied:') |
1587 | 399 for st in implied: |
1597 | 400 log.info(f'{INDENT*1} {st}') |
1587 | 401 return implied |
402 | |
1607
b21885181e35
more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents:
1605
diff
changeset
|
403 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
|
404 for i, rule in enumerate(self.rules): |
272f78d4671a
mark skipped tests. move applyRule into Rule. minor cleanups.
drewp@bigasterisk.com
parents:
1611
diff
changeset
|
405 self._logRuleApplicationHeader(workingSet, i, rule) |
272f78d4671a
mark skipped tests. move applyRule into Rule. minor cleanups.
drewp@bigasterisk.com
parents:
1611
diff
changeset
|
406 rule.applyRule(workingSet, implied, stats) |
1587 | 407 |
1607
b21885181e35
more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents:
1605
diff
changeset
|
408 def _logRuleApplicationHeader(self, workingSet, i, r: Rule): |
1594 | 409 if not log.isEnabledFor(logging.DEBUG): |
410 return | |
411 | |
412 log.debug('') | |
413 log.debug(f'{INDENT*2} workingSet:') | |
1597 | 414 for j, stmt in enumerate(sorted(workingSet)): |
415 log.debug(f'{INDENT*3} ({j}) {stmt}') | |
1594 | 416 |
417 log.debug('') | |
418 log.debug(f'{INDENT*2}-applying rule {i}') | |
1632 | 419 log.debug(f'{INDENT*3} rule def lhs:') |
1634
ba59cfc3c747
hack math:sum in there. Test suite is passing except some slow performers
drewp@bigasterisk.com
parents:
1633
diff
changeset
|
420 for stmt in sorted(r.lhsGraph, reverse=True): |
1632 | 421 log.debug(f'{INDENT*4} {stmt}') |
1607
b21885181e35
more modules, types. Maybe less repeated computation on BoundLhs
drewp@bigasterisk.com
parents:
1605
diff
changeset
|
422 log.debug(f'{INDENT*3} rule def rhs: {graphDump(r.rhsGraph)}') |
1594 | 423 |
1588
0757fafbfdab
WIP inferencer - partial var and function support
drewp@bigasterisk.com
parents:
1587
diff
changeset
|
424 |
1590
327202020892
WIP inference- getting into more degenerate test cases
drewp@bigasterisk.com
parents:
1589
diff
changeset
|
425 def graphDump(g: Union[Graph, List[Triple]]): |
1637 | 426 # this is very slow- debug only! |
427 if not log.isEnabledFor(logging.DEBUG): | |
428 return "(skipped dump)" | |
1590
327202020892
WIP inference- getting into more degenerate test cases
drewp@bigasterisk.com
parents:
1589
diff
changeset
|
429 if not isinstance(g, Graph): |
327202020892
WIP inference- getting into more degenerate test cases
drewp@bigasterisk.com
parents:
1589
diff
changeset
|
430 g2 = Graph() |
1594 | 431 g2 += g |
1590
327202020892
WIP inference- getting into more degenerate test cases
drewp@bigasterisk.com
parents:
1589
diff
changeset
|
432 g = g2 |
1589
5c1055be3c36
WIP more debugging, working towards bnode-matching support
drewp@bigasterisk.com
parents:
1588
diff
changeset
|
433 g.bind('', ROOM) |
5c1055be3c36
WIP more debugging, working towards bnode-matching support
drewp@bigasterisk.com
parents:
1588
diff
changeset
|
434 g.bind('ex', Namespace('http://example.com/')) |
5c1055be3c36
WIP more debugging, working towards bnode-matching support
drewp@bigasterisk.com
parents:
1588
diff
changeset
|
435 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
|
436 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
|
437 return ' '.join(lines) |