class GenerationDispatcher:
def __init__(self, config, trainingDataset, holdOutDataset, processingStartIndex, processingEndIndex):
self.config = config
self.logger = config["logger"]
self.esClient = Elasticsearch(config["elasticsearch"]["host"] + ":" + str(config["elasticsearch"]["port"]))
self.trainingDataset = trainingDataset
self.holdOutDataset = holdOutDataset
self.config["processingStartIndex"] = processingStartIndex
self.config["processingEndIndex"] = processingEndIndex
self.bagOfPhrases = {}
self.corpusIndex = config["corpus"]["index"]
self.corpusType = config["corpus"]["type"]
self.corpusFields = config["corpus"]["text_fields"]
self.corpusSize = 0
self.totalPhrasesDispatched = 0
self.phrasesGenerated = 0
self.phrasesNotGenerated = 0
self.timeout = 86400000
self.dispatcherName = "bayzee.generation.dispatcher"
if processingEndIndex != None:
self.dispatcherName += "." + str(processingStartIndex) + "." + str(processingEndIndex)
self.workerName = "bayzee.generation.worker"
self.processorIndex = config["processor"]["index"]
self.processorType = config["processor"]["type"]
self.processorPhraseType = config["processor"]["type"]+"__phrase"
self.processingPageSize = config["processingPageSize"]
config["processor_phrase_type"] = self.processorPhraseType
self.featureNames = map(lambda x: x["name"], config["generator"]["features"])
for module in config["processor"]["modules"]:
self.featureNames = self.featureNames + map(lambda x: x["name"], module["features"])
# creating generation dispatcher
self.generationDispatcher = DurableChannel(self.dispatcherName, config, self.timeoutCallback)
# creating controle channel
self.controlChannel = RemoteChannel(self.dispatcherName, config)
def dispatchToGenerate(self):
processorIndex = self.config["processor"]["index"]
phraseProcessorType = self.config["processor"]["type"] + "__phrase"
nextPhraseIndex = 0
if self.config["processingStartIndex"] != None: nextPhraseIndex = self.config["processingStartIndex"]
endPhraseIndex = -1
if self.config["processingEndIndex"] != None: endPhraseIndex = self.config["processingEndIndex"]
if endPhraseIndex != -1 and self.processingPageSize > (endPhraseIndex - nextPhraseIndex):
self.processingPageSize = endPhraseIndex - nextPhraseIndex + 1
while True:
phrases = self.esClient.search(index=processorIndex, doc_type=phraseProcessorType, body={"from": nextPhraseIndex,"size": self.processingPageSize, "query":{"match_all":{}},"sort":[{"phrase__not_analyzed":{"order":"asc"}}]}, fields=["_id"])
if len(phrases["hits"]["hits"]) == 0: break
self.totalPhrasesDispatched += len(phrases["hits"]["hits"])
floatPrecision = "{0:." + str(self.config["generator"]["floatPrecision"]) + "f}"
self.logger.info("Generating features from " + str(nextPhraseIndex) + " to " + str(nextPhraseIndex+len(phrases["hits"]["hits"])) + " phrases...")
for phraseData in phrases["hits"]["hits"]:
self.logger.info("Dispatching phrase " + phraseData["_id"])
content = {"phraseId": phraseData["_id"], "type": "generate", "count": 1, "from": self.dispatcherName}
self.generationDispatcher.send(content, self.workerName, self.timeout)
nextPhraseIndex += len(phrases["hits"]["hits"])
if endPhraseIndex != -1 and nextPhraseIndex >= endPhraseIndex: break
while True:
message = self.generationDispatcher.receive()
if "phraseId" in message["content"] and message["content"]["phraseId"] > 0:
self.phrasesGenerated += 1
self.generationDispatcher.close(message)
self.logger.info("Generated for " + message["content"]["phraseId"] + str(self.phrasesGenerated) + "/" + str(self.totalPhrasesDispatched))
if (self.phrasesGenerated + self.phrasesNotGenerated) >= self.totalPhrasesDispatched:
self.controlChannel.send("dying")
break
self.__terminate()
def timeoutCallback(self, message):
config.logger.info("Message timed out: " + str(message))
if message["content"]["count"] < 5:
message["content"]["count"] += 1
self.generationDispatcher.send(message["content"], self.workerName, self.timeout)
else:
#log implementation yet to be done for expired phrases
self.phrasesNotGenerated += 1
if self.phrasesNotGenerated == self.totalPhrasesDispatched or (self.phrasesGenerated + self.phrasesNotGenerated) == self.totalPhrasesDispatched:
self.__terminate()
def __terminate(self):
self.logger.info(str(self.totalPhrasesDispatched) + " total dispatched")
self.logger.info(str(self.phrasesGenerated) + " generated")
self.logger.info(str(self.phrasesNotGenerated) + " failed to generate")
self.logger.info("Generation complete")
self.logger.info("Terminating generation dispatcher")
class GenerationDispatcher:
def __init__(self, config, trainingDataset, holdOutDataset,
processingStartIndex, processingEndIndex):
self.config = config
self.logger = config["logger"]
self.esClient = Elasticsearch(config["elasticsearch"]["host"] + ":" +
str(config["elasticsearch"]["port"]))
self.trainingDataset = trainingDataset
self.holdOutDataset = holdOutDataset
self.config["processingStartIndex"] = processingStartIndex
self.config["processingEndIndex"] = processingEndIndex
self.bagOfPhrases = {}
self.corpusIndex = config["corpus"]["index"]
self.corpusType = config["corpus"]["type"]
self.corpusFields = config["corpus"]["text_fields"]
self.corpusSize = 0
self.totalPhrasesDispatched = 0
self.phrasesGenerated = 0
self.phrasesNotGenerated = 0
self.timeout = 86400000
self.dispatcherName = "bayzee.generation.dispatcher"
if processingEndIndex != None:
self.dispatcherName += "." + str(processingStartIndex) + "." + str(
processingEndIndex)
self.workerName = "bayzee.generation.worker"
self.processorIndex = config["processor"]["index"]
self.processorType = config["processor"]["type"]
self.processorPhraseType = config["processor"]["type"] + "__phrase"
self.processingPageSize = config["processingPageSize"]
config["processor_phrase_type"] = self.processorPhraseType
self.featureNames = map(lambda x: x["name"],
config["generator"]["features"])
for module in config["processor"]["modules"]:
self.featureNames = self.featureNames + map(
lambda x: x["name"], module["features"])
# creating generation dispatcher
self.generationDispatcher = DurableChannel(self.dispatcherName, config,
self.timeoutCallback)
# creating controle channel
self.controlChannel = RemoteChannel(self.dispatcherName, config)
def dispatchToGenerate(self):
processorIndex = self.config["processor"]["index"]
phraseProcessorType = self.config["processor"]["type"] + "__phrase"
nextPhraseIndex = 0
if self.config["processingStartIndex"] != None:
nextPhraseIndex = self.config["processingStartIndex"]
endPhraseIndex = -1
if self.config["processingEndIndex"] != None:
endPhraseIndex = self.config["processingEndIndex"]
if endPhraseIndex != -1 and self.processingPageSize > (
endPhraseIndex - nextPhraseIndex):
self.processingPageSize = endPhraseIndex - nextPhraseIndex + 1
while True:
phrases = self.esClient.search(index=processorIndex,
doc_type=phraseProcessorType,
body={
"from":
nextPhraseIndex,
"size":
self.processingPageSize,
"query": {
"match_all": {}
},
"sort": [{
"phrase__not_analyzed": {
"order": "asc"
}
}]
},
fields=["_id"])
if len(phrases["hits"]["hits"]) == 0: break
self.totalPhrasesDispatched += len(phrases["hits"]["hits"])
floatPrecision = "{0:." + str(
self.config["generator"]["floatPrecision"]) + "f}"
self.logger.info("Generating features from " +
str(nextPhraseIndex) + " to " +
str(nextPhraseIndex +
len(phrases["hits"]["hits"])) + " phrases...")
for phraseData in phrases["hits"]["hits"]:
self.logger.info("Dispatching phrase " + phraseData["_id"])
content = {
"phraseId": phraseData["_id"],
"type": "generate",
"count": 1,
"from": self.dispatcherName
}
self.generationDispatcher.send(content, self.workerName,
self.timeout)
nextPhraseIndex += len(phrases["hits"]["hits"])
if endPhraseIndex != -1 and nextPhraseIndex >= endPhraseIndex:
break
while True:
message = self.generationDispatcher.receive()
if "phraseId" in message[
"content"] and message["content"]["phraseId"] > 0:
self.phrasesGenerated += 1
self.generationDispatcher.close(message)
self.logger.info("Generated for " +
message["content"]["phraseId"] +
str(self.phrasesGenerated) + "/" +
str(self.totalPhrasesDispatched))
if (self.phrasesGenerated +
self.phrasesNotGenerated) >= self.totalPhrasesDispatched:
self.controlChannel.send("dying")
break
self.__terminate()
def timeoutCallback(self, message):
config.logger.info("Message timed out: " + str(message))
if message["content"]["count"] < 5:
message["content"]["count"] += 1
self.generationDispatcher.send(message["content"], self.workerName,
self.timeout)
else:
#log implementation yet to be done for expired phrases
self.phrasesNotGenerated += 1
if self.phrasesNotGenerated == self.totalPhrasesDispatched or (
self.phrasesGenerated +
self.phrasesNotGenerated) == self.totalPhrasesDispatched:
self.__terminate()
def __terminate(self):
self.logger.info(
str(self.totalPhrasesDispatched) + " total dispatched")
self.logger.info(str(self.phrasesGenerated) + " generated")
self.logger.info(str(self.phrasesNotGenerated) + " failed to generate")
self.logger.info("Generation complete")
self.logger.info("Terminating generation dispatcher")
class AnnotationDispatcher:
def __init__(self, config, processingStartIndex, processingEndIndex):
self.config = config
self.logger = config["logger"]
self.esClient = Elasticsearch(config["elasticsearch"]["host"] + ":" + str(config["elasticsearch"]["port"]))
self.bagOfPhrases = {}
self.corpusIndex = config["corpus"]["index"]
self.corpusType = config["corpus"]["type"]
self.corpusFields = config["corpus"]["text_fields"]
self.corpusSize = 0
self.processorIndex = config["processor"]["index"]
self.processorType = config["processor"]["type"]
self.processorPhraseType = config["processor"]["type"] + "__phrase"
self.processingPageSize = config["processingPageSize"]
self.analyzerIndex = self.corpusIndex + "__analysis__"
self.config["processingStartIndex"] = processingStartIndex
self.config["processingEndIndex"] = processingEndIndex
self.config["processingPageSize"] = self.processingPageSize
self.totalDocumentsDispatched = 0
self.documentsAnnotated = 0
self.documentsNotAnnotated = 0
self.lastDispatcher = False
self.endProcess = False
self.dispatcherName = "bayzee.annotation.dispatcher"
self.workerName = "bayzee.annotation.worker"
self.timeout = 86400000
if processingEndIndex != None:
self.dispatcherName += "." + str(processingStartIndex) + "." + str(processingEndIndex)
analyzerIndexSettings = {
"index":{
"analysis":{
"analyzer":{
"analyzer_shingle":{
"type": "custom",
"tokenizer": "standard",
"filter": ["standard", "lowercase", "filter_shingle"]
}
},
"filter":{
"filter_shingle":{
"type": "shingle",
"max_shingle_size": config["generator"]["maxShingleSize"],
"min_shingle_size": config["generator"]["minShingleSize"],
"output_unigrams": (config["generator"]["minShingleSize"] == 1)
},
"filter_stop":{
"type": "stop"
}
}
}
}
}
analyzerIndexTypeMapping = {
"properties":{
"phrase":{"type":"string"},
"document_id":{"type":"string", "index": "not_analyzed"},
"phrase__not_analyzed":{"type":"string","index":"not_analyzed"}
}
}
corpusSize = self.esClient.count(index=self.corpusIndex, doc_type=self.corpusType, body={"query":{"match_all":{}}})
self.corpusSize = corpusSize["count"]
self.featureNames = map(lambda x: x["name"], config["generator"]["features"])
for module in config["processor"]["modules"]:
self.featureNames = self.featureNames + map(lambda x: x["name"], module["features"])
if processingStartIndex == 0:
if self.esClient.indices.exists(self.analyzerIndex):
self.esClient.indices.delete(self.analyzerIndex)
data = self.esClient.indices.create(self.analyzerIndex, analyzerIndexSettings)
if "annotateFromScratch" not in self.config or self.config["annotateFromScratch"] == True:
try:
if self.esClient.indices.exists(self.config["processor"]["index"]):
self.esClient.indices.delete(self.config["processor"]["index"])
self.esClient.indices.create(self.config["processor"]["index"])
self.esClient.indices.put_mapping(index=self.config["processor"]["index"],doc_type=self.processorPhraseType,body=analyzerIndexTypeMapping)
if self.esClient.indices.exists(self.analyzerIndex):
self.esClient.indices.delete(self.analyzerIndex)
data = self.esClient.indices.create(self.analyzerIndex, analyzerIndexSettings)
except:
error = sys.exc_info()
self.logger.error("Error occurred during initialization of analyzer index: " + str(error))
sys.exit(1)
else:
sleep(1)
#dispatcher creation
self.annotationDispatcher = DurableChannel(self.dispatcherName, config, self.timeoutCallback)
#remote channel intialisation
self.controlChannel = RemoteChannel(self.dispatcherName, config)
def dispatchToAnnotate(self):
if "indexPhrases" in self.config and self.config["indexPhrases"] == False: return
nextDocumentIndex = 0
if self.config["processingStartIndex"] != None: nextDocumentIndex = self.config["processingStartIndex"]
endDocumentIndex = -1
if self.config["processingEndIndex"] != None: endDocumentIndex = self.config["processingEndIndex"]
if endDocumentIndex != -1 and self.processingPageSize > (endDocumentIndex - nextDocumentIndex):
self.processingPageSize = endDocumentIndex - nextDocumentIndex + 1
self.totalDocumentsDispatched = 0
while True:
documents = self.esClient.search(index=self.corpusIndex, doc_type=self.corpusType, body={"from": nextDocumentIndex,"size": self.processingPageSize,"query":{"match_all":{}}, "sort":[{"_id":{"order":"asc"}}]}, fields=["_id"])
if len(documents["hits"]["hits"]) == 0:
break
self.totalDocumentsDispatched += len(documents["hits"]["hits"])
self.logger.info("Annotating " + str(nextDocumentIndex) + " to " + str(nextDocumentIndex+len(documents["hits"]["hits"])) + " documents...")
for document in documents["hits"]["hits"]:
self.logger.info("Dispatching document " + document["_id"])
content = {"documentId": document["_id"], "type": "annotate", "count": 1, "from":self.dispatcherName}
self.annotationDispatcher.send(content, self.workerName)
nextDocumentIndex += len(documents["hits"]["hits"])
if endDocumentIndex != -1 and endDocumentIndex <= nextDocumentIndex:
break
self.logger.info(str(self.totalDocumentsDispatched) + " documents dispatched")
while True:
message = self.annotationDispatcher.receive()
if "documentId" in message["content"] and message["content"]["documentId"] > 0:
self.documentsAnnotated += 1
self.annotationDispatcher.close(message)
self.logger.info("Annotated document " + message["content"]["documentId"] + " - " + str(self.documentsAnnotated) + "/" + str(self.totalDocumentsDispatched))
if (self.documentsAnnotated + self.documentsNotAnnotated) >= self.totalDocumentsDispatched and not self.lastDispatcher:
self.controlChannel.send("dying")
self.annotationDispatcher.end()
break
self.__terminate()
def timeoutCallback(self, message):
if message["content"]["count"] < 5:
message["content"]["count"] += 1
self.annotationDispatcher.send(message["content"], self.workerName, self.timeout)
else:
#log implementation yet to be done for expired documents
self.documentsNotAnnotated += 1
if self.documentsNotAnnotated == self.totalDocumentsDispatched or (self.documentsAnnotated + self.documentsNotAnnotated) == self.totalDocumentsDispatched:
self.__terminate()
def __terminate(self):
self.logger.info(str(self.totalDocumentsDispatched) + " total dispatched")
self.logger.info(str(self.documentsAnnotated) + " annotated")
self.logger.info(str(self.documentsNotAnnotated) + " failed to annotate")
self.logger.info("Annotation complete")
self.logger.info("Terminating annotation dispatcher")
def __deleteAnalyzerIndex(self):
if self.esClient.indices.exists(self.analyzerIndex):
self.esClient.indices.delete(self.analyzerIndex)
class AnnotationDispatcher:
def __init__(self, config, processingStartIndex, processingEndIndex):
self.config = config
self.logger = config["logger"]
self.esClient = Elasticsearch(config["elasticsearch"]["host"] + ":" +
str(config["elasticsearch"]["port"]))
self.bagOfPhrases = {}
self.corpusIndex = config["corpus"]["index"]
self.corpusType = config["corpus"]["type"]
self.corpusFields = config["corpus"]["text_fields"]
self.corpusSize = 0
self.processorIndex = config["processor"]["index"]
self.processorType = config["processor"]["type"]
self.processorPhraseType = config["processor"]["type"] + "__phrase"
self.processingPageSize = config["processingPageSize"]
self.analyzerIndex = self.corpusIndex + "__analysis__"
self.config["processingStartIndex"] = processingStartIndex
self.config["processingEndIndex"] = processingEndIndex
self.config["processingPageSize"] = self.processingPageSize
self.totalDocumentsDispatched = 0
self.documentsAnnotated = 0
self.documentsNotAnnotated = 0
self.lastDispatcher = False
self.endProcess = False
self.dispatcherName = "bayzee.annotation.dispatcher"
self.workerName = "bayzee.annotation.worker"
self.timeout = 86400000
if processingEndIndex != None:
self.dispatcherName += "." + str(processingStartIndex) + "." + str(
processingEndIndex)
analyzerIndexSettings = {
"index": {
"analysis": {
"analyzer": {
"analyzer_shingle": {
"type": "custom",
"tokenizer": "standard",
"filter":
["standard", "lowercase", "filter_shingle"]
}
},
"filter": {
"filter_shingle": {
"type":
"shingle",
"max_shingle_size":
config["generator"]["maxShingleSize"],
"min_shingle_size":
config["generator"]["minShingleSize"],
"output_unigrams":
(config["generator"]["minShingleSize"] == 1)
},
"filter_stop": {
"type": "stop"
}
}
}
}
}
analyzerIndexTypeMapping = {
"properties": {
"phrase": {
"type": "string"
},
"document_id": {
"type": "string",
"index": "not_analyzed"
},
"phrase__not_analyzed": {
"type": "string",
"index": "not_analyzed"
}
}
}
corpusSize = self.esClient.count(index=self.corpusIndex,
doc_type=self.corpusType,
body={"query": {
"match_all": {}
}})
self.corpusSize = corpusSize["count"]
self.featureNames = map(lambda x: x["name"],
config["generator"]["features"])
for module in config["processor"]["modules"]:
self.featureNames = self.featureNames + map(
lambda x: x["name"], module["features"])
if processingStartIndex == 0:
if self.esClient.indices.exists(self.analyzerIndex):
self.esClient.indices.delete(self.analyzerIndex)
data = self.esClient.indices.create(self.analyzerIndex,
analyzerIndexSettings)
if "annotateFromScratch" not in self.config or self.config[
"annotateFromScratch"] == True:
try:
if self.esClient.indices.exists(
self.config["processor"]["index"]):
self.esClient.indices.delete(
self.config["processor"]["index"])
self.esClient.indices.create(self.config["processor"]["index"])
self.esClient.indices.put_mapping(
index=self.config["processor"]["index"],
doc_type=self.processorPhraseType,
body=analyzerIndexTypeMapping)
if self.esClient.indices.exists(self.analyzerIndex):
self.esClient.indices.delete(self.analyzerIndex)
data = self.esClient.indices.create(self.analyzerIndex,
analyzerIndexSettings)
except:
error = sys.exc_info()
self.logger.error(
"Error occurred during initialization of analyzer index: "
+ str(error))
sys.exit(1)
else:
sleep(1)
#dispatcher creation
self.annotationDispatcher = DurableChannel(self.dispatcherName, config,
self.timeoutCallback)
#remote channel intialisation
self.controlChannel = RemoteChannel(self.dispatcherName, config)
def dispatchToAnnotate(self):
if "indexPhrases" in self.config and self.config[
"indexPhrases"] == False:
return
nextDocumentIndex = 0
if self.config["processingStartIndex"] != None:
nextDocumentIndex = self.config["processingStartIndex"]
endDocumentIndex = -1
if self.config["processingEndIndex"] != None:
endDocumentIndex = self.config["processingEndIndex"]
if endDocumentIndex != -1 and self.processingPageSize > (
endDocumentIndex - nextDocumentIndex):
self.processingPageSize = endDocumentIndex - nextDocumentIndex + 1
self.totalDocumentsDispatched = 0
while True:
documents = self.esClient.search(index=self.corpusIndex,
doc_type=self.corpusType,
body={
"from":
nextDocumentIndex,
"size":
self.processingPageSize,
"query": {
"match_all": {}
},
"sort": [{
"_id": {
"order": "asc"
}
}]
},
fields=["_id"])
if len(documents["hits"]["hits"]) == 0:
break
self.totalDocumentsDispatched += len(documents["hits"]["hits"])
self.logger.info("Annotating " + str(nextDocumentIndex) + " to " +
str(nextDocumentIndex +
len(documents["hits"]["hits"])) +
" documents...")
for document in documents["hits"]["hits"]:
self.logger.info("Dispatching document " + document["_id"])
content = {
"documentId": document["_id"],
"type": "annotate",
"count": 1,
"from": self.dispatcherName
}
self.annotationDispatcher.send(content, self.workerName)
nextDocumentIndex += len(documents["hits"]["hits"])
if endDocumentIndex != -1 and endDocumentIndex <= nextDocumentIndex:
break
self.logger.info(
str(self.totalDocumentsDispatched) + " documents dispatched")
while True:
message = self.annotationDispatcher.receive()
if "documentId" in message[
"content"] and message["content"]["documentId"] > 0:
self.documentsAnnotated += 1
self.annotationDispatcher.close(message)
self.logger.info("Annotated document " +
message["content"]["documentId"] + " - " +
str(self.documentsAnnotated) + "/" +
str(self.totalDocumentsDispatched))
if (self.documentsAnnotated + self.documentsNotAnnotated
) >= self.totalDocumentsDispatched and not self.lastDispatcher:
self.controlChannel.send("dying")
self.annotationDispatcher.end()
break
self.__terminate()
def timeoutCallback(self, message):
if message["content"]["count"] < 5:
message["content"]["count"] += 1
self.annotationDispatcher.send(message["content"], self.workerName,
self.timeout)
else:
#log implementation yet to be done for expired documents
self.documentsNotAnnotated += 1
if self.documentsNotAnnotated == self.totalDocumentsDispatched or (
self.documentsAnnotated + self.documentsNotAnnotated
) == self.totalDocumentsDispatched:
self.__terminate()
def __terminate(self):
self.logger.info(
str(self.totalDocumentsDispatched) + " total dispatched")
self.logger.info(str(self.documentsAnnotated) + " annotated")
self.logger.info(
str(self.documentsNotAnnotated) + " failed to annotate")
self.logger.info("Annotation complete")
self.logger.info("Terminating annotation dispatcher")
def __deleteAnalyzerIndex(self):
if self.esClient.indices.exists(self.analyzerIndex):
self.esClient.indices.delete(self.analyzerIndex)