From abb936f5d76f7fe5cec1e8948d287da86643d504 Mon Sep 17 00:00:00 2001 From: Terry Truong Date: Wed, 22 Jun 2022 23:16:42 +1000 Subject: Refactor backend scripts --- backend/data/genEnwikiDescData.py | 54 ++++++++++++++++++++++----------------- 1 file changed, 30 insertions(+), 24 deletions(-) (limited to 'backend/data/genEnwikiDescData.py') diff --git a/backend/data/genEnwikiDescData.py b/backend/data/genEnwikiDescData.py index dbc8d6b..d3f93ed 100755 --- a/backend/data/genEnwikiDescData.py +++ b/backend/data/genEnwikiDescData.py @@ -3,10 +3,13 @@ import sys, re, os import sqlite3 -usageInfo = f"usage: {sys.argv[0]}\n" -usageInfo += "Reads Wikimedia enwiki data from enwiki/, and node and name data" -usageInfo += "from a sqlite database, and adds description data for names that\n" -usageInfo += "don't have them.\n" +usageInfo = f""" +Usage: {sys.argv[0]} + +Reads a database containing data from Wikipedia, and tries to associate +wiki pages with nodes in the database, and add descriptions for nodes +that don't have them. +""" if len(sys.argv) > 1: print(usageInfo, file=sys.stderr) sys.exit(1) @@ -15,36 +18,39 @@ enwikiDb = "enwiki/descData.db" dbFile = "data.db" namesToSkipFile = "pickedEnwikiNamesToSkip.txt" pickedLabelsFile = "pickedEnwikiLabels.txt" +# Got about 25k descriptions when testing -# Open dbs +print("Opening databases") enwikiCon = sqlite3.connect(enwikiDb) enwikiCur = enwikiCon.cursor() dbCon = sqlite3.connect(dbFile) dbCur = dbCon.cursor() -# Read name/title files + +print("Checking for names to skip") namesToSkip = set() -nameToPickedTitle = {} # Maps names to titles to be used for them if os.path.exists(namesToSkipFile): with open(namesToSkipFile) as file: for line in file: namesToSkip.add(line.rstrip()) - print(f"Read in {len(namesToSkip)} names to skip") + print(f"Found {len(namesToSkip)}") +print("Checking for picked-titles") +nameToPickedTitle = {} if os.path.exists(pickedLabelsFile): with open(pickedLabelsFile) as file: for line in file: (name, _, title) = line.rstrip().partition("|") nameToPickedTitle[name.lower()] = title -print(f"Read in {len(nameToPickedTitle)} titles to use for certain names") -# Get node names without descriptions -print("Getting node names") +print(f"Found {len(nameToPickedTitle)}") + +print("Getting names of nodes without descriptions") nodeNames = set() query = "SELECT nodes.name FROM nodes LEFT JOIN wiki_ids ON nodes.name = wiki_ids.name WHERE wiki_ids.id IS NULL" -for row in dbCur.execute(query): - nodeNames.add(row[0]) -print(f"Found {len(nodeNames)} names") +for (name,) in dbCur.execute(query): + nodeNames.add(name) +print(f"Found {len(nodeNames)}") nodeNames.difference_update(namesToSkip) -# Find page id for each node name -print("Getting node page-ids") + +print("Associating nodes with page IDs") nodeToPageId = {} iterNum = 0 for name in nodeNames: @@ -63,34 +69,34 @@ for name in nodeNames: nodeToPageId[name] = row[0] else: print("WARNING: Picked title {title} not found", file=sys.stderr) -# Resolve redirects + print("Resolving redirects") redirectingNames = set() iterNum = 0 for (name, pageId) in nodeToPageId.items(): iterNum += 1 - if iterNum % 1000 == 0: + if iterNum % 1e3 == 0: print(f"At iteration {iterNum}") # - row = enwikiCur.execute( - "SELECT pages.id FROM redirects INNER JOIN pages ON redirects.target = pages.title WHERE redirects.id = ?", - (pageId,)).fetchone() + query = "SELECT pages.id FROM redirects INNER JOIN pages ON redirects.target = pages.title WHERE redirects.id = ?" + row = enwikiCur.execute(query, (pageId,)).fetchone() if row != None: nodeToPageId[name] = row[0] redirectingNames.add(name) -# Add descriptions for each node + print("Adding description data") iterNum = 0 for (name, pageId) in nodeToPageId.items(): iterNum += 1 - if iterNum % 1000 == 0: + if iterNum % 1e3 == 0: print(f"At iteration {iterNum}") # row = enwikiCur.execute("SELECT desc FROM descs where descs.id = ?", (pageId,)).fetchone() if row != None: dbCur.execute("INSERT INTO wiki_ids VALUES (?, ?, ?)", (name, pageId, 1 if name in redirectingNames else 0)) dbCur.execute("INSERT OR IGNORE INTO descs VALUES (?, ?, ?)", (pageId, row[0], 0)) -# Close dbs + +print("Closing databases") dbCon.commit() dbCon.close() enwikiCon.close() -- cgit v1.2.3