Files
@ dc6095cd7601
Branch filter:
Location: light9/light9/rdfdb/autodepgraphapi.py
dc6095cd7601
6.8 KiB
text/x-python
syncedgraph.contains support
Ignore-this: 64c70222eda0fabec59b446237b17dde
Ignore-this: 64c70222eda0fabec59b446237b17dde
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 | import logging
from rdflib import RDF, RDFS
from light9.rdfdb.currentstategraphapi import contextsForStatementNoWildcards
log = logging.getLogger('syncedgraph')
class AutoDepGraphApi(object):
"""
knockoutjs-inspired API for automatically building a dependency
tree while reading the graph. See addHandler().
This design is more aggressive than knockoutjs, since the obvious
query methods like value() all error if you're not in a watched
section of code. This is supposed to make it easier to notice
dependency mistakes, especially when porting old code to use
SyncedGraph.
This class is a mixin for SyncedGraph, separated here because
these methods work together
"""
def __init__(self):
self._watchers = _GraphWatchers()
self.currentFuncs = [] # stack of addHandler callers
def addHandler(self, func):
"""
run this (idempotent) func, noting what graph values it
uses. Run it again in the future if there are changes to those
graph values. The func might use different values during that
future call, and those will be what we watch for next.
"""
# if we saw this func before, we need to forget the old
# callbacks it wanted and replace with the new ones we see
# now.
# if one handler func calls another, does that break anything?
# maybe not?
# no plan for sparql queries yet. Hook into a lower layer that
# reveals all their statement fetches? Just make them always
# new? Cache their results, so if i make the query again and
# it gives the same result, I don't call the handler?
self.currentFuncs.append(func)
try:
func()
finally:
self.currentFuncs.pop()
def runDepsOnNewPatch(self, p):
"""
patch p just happened to the graph; call everyone back who
might care, and then notice what data they depend on now
"""
for func in self._watchers.whoCares(p):
# todo: forget the old handlers for this func
self.addHandler(func)
def _getCurrentFunc(self):
if not self.currentFuncs:
# this may become a warning later
raise ValueError("asked for graph data outside of a handler")
# we add the watcher to the deepest function, since that
# should be the cheapest way to update when this part of the
# data changes
return self.currentFuncs[-1]
# these just call through to triples() so it might be possible to
# watch just that one.
# if you get a bnode in your response, maybe the answer to
# dependency tracking is to say that you depended on the triple
# that got you that bnode, since it is likely to change to another
# bnode later. This won't work if the receiver stores bnodes
# between calls, but probably most of them don't do that (they
# work from a starting uri)
def value(self, subject=None, predicate=RDF.value, object=None,
default=None, any=True):
if object is not None:
raise NotImplementedError()
func = self._getCurrentFunc()
self._watchers.addSubjPredWatcher(func, subject, predicate)
return self._graph.value(subject, predicate, object=object,
default=default, any=any)
def objects(self, subject=None, predicate=None):
func = self._getCurrentFunc()
self._watchers.addSubjPredWatcher(func, subject, predicate)
return self._graph.objects(subject, predicate)
def label(self, uri):
return self.value(uri, RDFS.label)
def subjects(self, predicate=None, object=None):
func = self._getCurrentFunc()
self._watchers.addPredObjWatcher(func, predicate, object)
return self._graph.subjects(predicate, object)
def contains(self, triple):
func = self._getCurrentFunc()
self._watchers.addTripleWatcher(func, triple)
return triple in self._graph
def contextsForStatement(self, triple):
"""currently this needs to be in an addHandler section, but it
sets no watchers so it won't actually update if the statement
was added or dropped from contexts"""
func = self._getCurrentFunc()
return contextsForStatementNoWildcards(self._graph, triple)
# i find myself wanting 'patch' (aka enter/leave) versions of these calls that tell
# you only what results have just appeared or disappeared. I think
# I'm going to be repeating that logic a lot. Maybe just for the
# subjects(RDF.type, t) call
class _GraphWatchers(object):
"""
store the current handlers that care about graph changes
"""
def __init__(self):
self._handlersSp = {} # (s,p): set(handlers)
self._handlersPo = {} # (p,o): set(handlers)
self._handlersSpo = {} # (s,p,o): set(handlers)
def addSubjPredWatcher(self, func, s, p):
if func is None:
return
key = s, p
try:
self._handlersSp.setdefault(key, set()).add(func)
except Exception:
log.error("with key %r and func %r" % (key, func))
raise
def addPredObjWatcher(self, func, p, o):
self._handlersPo.setdefault((p, o), set()).add(func)
def addTripleWatcher(self, func, triple):
self._handlersSpo.setdefault(triple, set()).add(func)
def whoCares(self, patch):
"""what handler functions would care about the changes in this patch?
this removes the handlers that it gives you
"""
#self.dependencies()
ret = set()
affectedSubjPreds = set([(s, p) for s, p, o, c in patch.addQuads]+
[(s, p) for s, p, o, c in patch.delQuads])
for (s, p), funcs in self._handlersSp.iteritems():
if (s, p) in affectedSubjPreds:
ret.update(funcs)
funcs.clear()
affectedPredObjs = set([(p, o) for s, p, o, c in patch.addQuads]+
[(p, o) for s, p, o, c in patch.delQuads])
for (p, o), funcs in self._handlersPo.iteritems():
if (p, o) in affectedPredObjs:
ret.update(funcs)
funcs.clear()
affectedTriples = set([(s, p, o) for s, p, o, c in patch.addQuads]+
[(s, p, o) for s, p, o, c in patch.delQuads])
for triple, funcs in self._handlersSpo.iteritems():
if triple in affectedTriples:
ret.update(funcs)
funcs.clear()
return ret
def dependencies(self):
"""
for debugging, make a list of all the active handlers and what
data they depend on. This is meant for showing on the web ui
for browsing.
"""
log.info("whocares:")
from pprint import pprint
pprint(self._handlersSp)
|