def merge_rdds(inputRDD, inputPath, baseRDD, joinRDD, removeElements, numPartitions):
#1. Merge baseRDD and joinRDD
#baseRDD: base_uri base_json
#joinRDD: source_uri base_uri
#output: source_uri, base_json
join_rdd_on_base = joinRDD.map(lambda x: load_linking_row(x, False)) #Get back base_uri -> source_uri
base_merge = join_rdd_on_base.join(baseRDD).map(lambda x: (x[1][0], x[1][1]))
#2. Extract the source_uri from inputRDD
#output: source_uri, input_uri
input_source_rdd = inputRDD.flatMapValues(lambda x: JSONUtil.extract_values_from_path(x, inputPath)) \
.map(lambda (x, y): (y, x))
#3. JOIN extracted source_uri with base
#output source_uri, (input_uri, base_json)
if numPartitions != -1:
merge3 = input_source_rdd.join(base_merge, numPartitions)
else:
merge3 = input_source_rdd.join(base_merge)
#4. Make input_uri as the key
#output: input_uri, (source_uri, base_json)
merge4 = merge3.map(lambda (source_uri, join_res): (join_res[0], (source_uri, join_res[1])))
#5. Group results by input_uri
#output: input_uri, list(source_uri, base_json)
merge5 = merge4.groupByKey()
#6 Merge in input_json
#output: input_uri, list(input_json), list(source_uri, base_json)
merge6 = inputRDD.cogroup(merge5)
#7 Replace JSON as necessary
result = merge6.mapValues(lambda x: merge_json(x[0], x[1], inputPath, removeElements))
return result
def clean_rdds(inputRDD, inputPath, baseRDD, numPartitions):
#1. Extract the source_uri from inputRDD
#output: source_uri, input_uri
input_source_rdd = inputRDD.flatMapValues(lambda x: JSONUtil.extract_values_from_path(x, inputPath)) \
.map(lambda (x, y): (y, x))
#2. JOIN extracted source_uri with base
#output source_uri, (input_uri, base_json)
merge3 = input_source_rdd.join(baseRDD, numPartitions)
#3. Make input_uri as the key
#output: input_uri, (source_uri, base_json)
merge4 = merge3.map(lambda (source_uri, join_res): (join_res[0], (source_uri, join_res[1])))
#4. Group results by input_uri
#output: input_uri, list(source_uri, base_json)
merge5 = merge4.groupByKey(numPartitions)
#5 Merge in input_json
#output: input_uri, list(input_json), list(source_uri, base_json)
merge6 = inputRDD.cogroup(merge5)
#6 Replace JSON as necessary
result = merge6.mapValues(lambda x: merge_json(x[0], x[1], inputPath, []))
return result
def merge_rdds(inputRDD, inputPath, baseRDD, numPartitions):
# inputRDD: input_uri, input_json
# base_rdd: merge_uri, base_json
#
# inputPath: path in input_json where merge_uri will be found
#
#2. Extract the source_uri from inputRDD
#output: merge_uri, input_uri
input_source_rdd = inputRDD.flatMapValues(lambda x: JSONUtil.extract_values_from_path(x, inputPath)) \
.map(lambda (x, y): (y, x))
#3. JOIN extracted source_uri with base
#output merge_uri, (input_uri, base_json)
merge3 = input_source_rdd.join(baseRDD)
#4. Make input_uri as the key
#output: input_uri, (merge_uri, base_json)
merge4 = merge3.map(lambda (source_uri, join_res): (join_res[0], [(source_uri, join_res[1])])).partitionBy(numPartitions)
#5. Group results by input_uri
#output: input_uri, list(merge_uri, base_json))
merge5 = merge4.reduceByKey(reduceLists)
#6 Merge in input_json
#output: input_uri, list(input_json), list(merge_uri, base_json)
merge6 = inputRDD.cogroup(merge5)
#7 Replace JSON as necessary
result = merge6.mapValues(lambda x: merge_json(x[0], x[1], inputPath))
return result
