Remove unnecessary counts in graph creation

This commit is contained in:
Ilias Kanellos 2023-05-23 16:48:58 +03:00
parent ec4e010687
commit 6a7e370a21
1 changed files with 6 additions and 0 deletions

View File

@ -196,15 +196,19 @@ oa_objects_df.printSchema()
# cited_by_df.unpersist(True) # cited_by_df.unpersist(True)
# Show total num of unique citations # Show total num of unique citations
'''
num_unique_citations = citations_df.count() num_unique_citations = citations_df.count()
print ("Total unique citations: " + str(num_unique_citations)) print ("Total unique citations: " + str(num_unique_citations))
'''
############################################################################################################################ ############################################################################################################################
# 3. Get any potentially missing 'citing' papers from references (these are dangling nodes w/o any outgoing references) # 3. Get any potentially missing 'citing' papers from references (these are dangling nodes w/o any outgoing references)
dangling_nodes = oa_objects_df.join(citations_df.select('citing').distinct(), citations_df.citing == oa_objects_df.id, 'left_anti')\ dangling_nodes = oa_objects_df.join(citations_df.select('citing').distinct(), citations_df.citing == oa_objects_df.id, 'left_anti')\
.select(F.col('id').alias('citing')).withColumn('cited', F.array([F.lit("0")])).repartition(num_partitions, 'citing') .select(F.col('id').alias('citing')).withColumn('cited', F.array([F.lit("0")])).repartition(num_partitions, 'citing')
# Count dangling nodes # Count dangling nodes
'''
dangling_num = dangling_nodes.count() dangling_num = dangling_nodes.count()
print ("Number of dangling nodes: " + str(dangling_num)) print ("Number of dangling nodes: " + str(dangling_num))
'''
# print ("Dangling nodes sample:") # print ("Dangling nodes sample:")
# dangling_nodes.show(10, False) # dangling_nodes.show(10, False)
############################################################################################################################ ############################################################################################################################
@ -213,8 +217,10 @@ graph = citations_df.groupBy('citing').agg(F.collect_set('cited').alias('cited')
# Free space # Free space
citations_df.unpersist(True) citations_df.unpersist(True)
'''
num_nodes = graph.count() num_nodes = graph.count()
print ("Entries in graph before dangling nodes:" + str(num_nodes)) print ("Entries in graph before dangling nodes:" + str(num_nodes))
'''
# print ("Sample in graph: ") # print ("Sample in graph: ")
# graph.show(10, False) # graph.show(10, False)