Homec4science
Diffusion invenio-infoscience (iugw2017) Browse

R3600/modules/bibrank/lib/bibrank_word_indexer.py
e98f0f0028f7iugw2017

Last Change
Raw File

bibrank_word_indexer.py
View Options

# $Id$
## BibRank word frequency indexer utility.
## This file is part of the CERN Document Server Software (CDSware).
## Copyright (C) 2002 CERN.
##
## The CDSware is free software; you can redistribute it and/or
## modify it under the terms of the GNU General Public License as
## published by the Free Software Foundation; either version 2 of the
## License, or (at your option) any later version.
##
## The CDSware is distributed in the hope that it will be useful, but
## WITHOUT ANY WARRANTY; without even the implied warranty of
## MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
## General Public License for more details.
##
## You should have received a copy of the GNU General Public License
## along with CDSware; if not, write to the Free Software Foundation, Inc.,
## 59 Temple Place, Suite 330, Boston, MA 02111-1307, USA.
## read config variables:
#include "config.wml"
#include "configbis.wml"
#include "cdswmllib.wml"
## start Python:
<protect>#!</protect><PYTHON>
<protect># -*- coding: utf-8 -*-</protect>
<protect>## $Id$</protect>
<protect>## DO NOT EDIT THIS FILE! IT WAS AUTOMATICALLY GENERATED FROM CDSware WML SOURCES.</protect>
__version__ = "<: print generate_pretty_version_string('$Id$'); :>"
## fill config variables:
pylibdir = "<LIBDIR>/python"
## programs used to convert fulltext files to text:
conv_programs = {#"ps": ["<PSTOTEXT>","<PSTOASCII>"], # switched off at the moment, since PDF is faster
#"ps.gz": ["<PSTOTEXT>","<PSTOASCII>"],
"pdf": ["<PDFTOTEXT>","<PSTOTEXT>","<PSTOASCII>"],
"doc": ["<ANTIWORD>","<CATDOC>","<WVTEXT>"],
"ppt": ["<PPTHTML>"],
"xls": ["<XLHTML>"]
}
## helper programs used if the above programs convert only to html or other intermediate file formats:
conv_programs_helpers = {"html": "<HTMLTOTEXT>",
"gz": "<GZIP>" }
## okay, rest of the Python code goes below
#######
## import interesting modules:
try:
from zlib import compress,decompress
from string import split,translate,lower,upper
import marshal
import getopt
import getpass
import string
import os
import sre
import sys
import time
import MySQLdb
import Numeric
import urllib
import signal
import tempfile
import unicodedata
import traceback
import cStringIO
import math
import re
import ConfigParser
except ImportError, e:
import sys
try:
import Stemmer
except ImportError, e:
pass
try:
sys.path.append('%s' % pylibdir)
from cdsware.config import *
from cdsware.search_engine_config import cfg_max_recID
from cdsware.search_engine import perform_request_search, strip_accents, HitSet
from cdsware.dbquery import run_sql
from cdsware.access_control_engine import acc_authorize_action
except ImportError, e:
import sys
## safety parameters concerning MySQL thread-multiplication problem:
cfg_check_mysql_threads = 0 # to check or not to check the problem?
cfg_max_mysql_threads = 50 # how many threads (connections) we consider as still safe
cfg_mysql_thread_timeout = 20 # we'll kill threads that were sleeping for more than X seconds
## override urllib's default password-asking behaviour:
class MyFancyURLopener(urllib.FancyURLopener):
def prompt_user_passwd(self, host, realm):
# supply some dummy credentials by default
return ("mysuperuser", "mysuperpass")
def http_error_401(self, url, fp, errcode, errmsg, headers):
# do not bother with protected pages
raise IOError, (999, 'unauthorized access')
return None
#urllib._urlopener = MyFancyURLopener()
## precompile some often-used regexp for speed reasons:
re_subfields = sre.compile('\$\$\w');
nb_char_in_line = 50 # for verbose pretty printing
chunksize = 1000 # default size of chunks that the records will be treated by
wordTables = []
base_process_size = 4500 # process base size
## Dictionary merging functions
def dict_union(list1, list2):
"Returns union of the two dictionaries."
union_dict = {}
for (e, count) in list1.iteritems():
union_dict[e] = count
for (e, count) in list2.iteritems():
if not union_dict.has_key(e):
union_dict[e] = count
else:
union_dict[e] = (union_dict[e][0] + count[0], count[1])
return union_dict
## safety function for killing slow MySQL threads:
def kill_sleepy_mysql_threads(max_threads=cfg_max_mysql_threads, thread_timeout=cfg_mysql_thread_timeout):
"""Check the number of MySQL threads and if there are more than
MAX_THREADS of them, lill all threads that are in a sleeping
state for more than THREAD_TIMEOUT seconds. (This is useful
for working around the the max_connection problem that appears
during indexation in some not-yet-understood cases.) If some
threads are to be killed, write info into the log file.
"""
res = run_sql("SHOW FULL PROCESSLIST")
if len(res) > max_threads:
for row in res:
r_id,r_user,r_host,r_db,r_command,r_time,r_state,r_info = row
if r_command == "Sleep" and int(r_time) > thread_timeout:
run_sql("KILL %s", (r_id,))
if options["verbose"] >= 1:
write_message("WARNING: too many MySQL threads, killing thread %s" % r_id)
return
# tagToFunctions mapping. It offers an indirection level necesary for
# indexing fulltext. The default is get_words_from_phrase
tagToWordsFunctions = {}
def get_words_from_phrase(phrase, weight, lang="",
chars_punctuation=r"[\.\,\:\;\?\!\"]",
chars_alphanumericseparators=r"[1234567890\!\"\#\$\%\&\'\(\)\*\+\,\-\.\/\:\;\<\=\>\?\@\[\\\]\^\_\`\{\|\}\~]",
split=string.split):
"Returns list of words from phrase 'phrase'."
words = {}
#print phrase
phrase = strip_accents(phrase)
phrase = lower(phrase)
#Getting rid of strange characters
phrase = re.sub("&eacute;", 'e', phrase)
phrase = re.sub("&egrave;", 'e', phrase)
phrase = re.sub("&agrave;", 'a', phrase)
phrase = re.sub("&nbsp;", ' ', phrase)
phrase = re.sub("&laquo;", ' ', phrase)
phrase = re.sub("&raquo;", ' ', phrase)
phrase = re.sub("&ecirc;", ' ', phrase)
phrase = re.sub("&amp;", ' ', phrase)
if string.find(phrase, "</") > -1:
#Most likely html, remove html code
phrase = re.sub("(?s)<[^>]*>|&#?\w+;", ' ', phrase)
phrase = re.sub(chars_punctuation, ' ', phrase)
#By doing this like below, characters standing alone, like c a b is not added to the inedx, but when they are together with characters like c++ or c$ they are added.
for word in split(phrase):
if not stopwords.has_key(word) and check_term(word, 0):
if lang and lang !="none" and stemmer.has_key(lang):
word = stemmer[lang].stem(word)
if not words.has_key(word):
words[word] = (0,0)
words[word] = (words[word][0] + weight, 0)
elif not stopwords.has_key(word):
phrase = re.sub(chars_alphanumericseparators, ' ', word)
for word_ in split(phrase):
if lang and lang !="none" and stemmer.has_key(lang):
word_ = stemmer[lang].stem(word_)
if word_:
if not words.has_key(word_):
words[word_] = (0,0)
words[word_] = (words[word_][0] + weight, 0)
#print words
return words
def split_ranges(parse_string):
recIDs = []
ranges = string.split(parse_string, ",")
for range in ranges:
tmp_recIDs = string.split(range, "-")
if len(tmp_recIDs)==1:
recIDs.append([int(tmp_recIDs[0]), int(tmp_recIDs[0])])
else:
if int(tmp_recIDs[0]) > int(tmp_recIDs[1]): # sanity check
tmp = tmp_recIDs[0]
tmp_recIDs[0] = tmp_recIDs[1]
tmp_recIDs[1] = tmp
recIDs.append([int(tmp_recIDs[0]), int(tmp_recIDs[1])])
return recIDs
def get_date_range(var):
"Returns the two dates contained as a low,high tuple"
limits = string.split(var, ",")
if len(limits)==1:
low = get_date(limits[0])
return low,None
if len(limits)==2:
low = get_date(limits[0])
high = get_date(limits[1])
return low,high
def get_datetime(var, format_string="%Y-%m-%d %H:%M:%S"):
"""Returns a date string according to the format string.
It can handle normal date strings and shifts with respect
to now."""
date = time.time()
shift_re=sre.compile("([-\+]{0,1})([\d]+)([dhms])")
factors = {"d":24*3600, "h":3600, "m":60, "s":1}
m = shift_re.match(var)
if m:
sign = m.groups()[0] == "-" and -1 or 1
factor = factors[m.groups()[2]]
value = float(m.groups()[1])
date = time.localtime(date + sign * factor * value)
date = time.strftime(format_string, date)
else:
date = time.strptime(var, format_string)
date = time.strftime(format_string, date)
return date
def create_range_list(res):
"""Creates a range list from a recID select query result contained
in res. The result is expected to have ascending numerical order."""
if not res:
return []
row = res[0]
if not row:
return []
else:
range_list = [[row[0],row[0]]]
for row in res[1:]:
id = row[0]
if id == range_list[-1][1] + 1:
range_list[-1][1] = id
else:
range_list.append([id,id])
return range_list
def beautify_range_list(range_list):
"""Returns a non overlapping, maximal range list"""
ret_list = []
for new in range_list:
found = 0
for old in ret_list:
if new[0] <= old[0] <= new[1] + 1 or new[0] - 1 <= old[1] <= new[1]:
old[0] = min(old[0], new[0])
old[1] = max(old[1], new[1])
found = 1
break
if not found:
ret_list.append(new)
return ret_list
def serialize_via_numeric_array_dumps(arr):
return Numeric.dumps(arr)
def serialize_via_numeric_array_compr(str):
return compress(str)
def serialize_via_numeric_array(arr):
"""Serialize Numeric array into a compressed string."""
return serialize_via_numeric_array_compr(serialize_via_numeric_array_dumps(arr))
def deserialize_via_numeric_array(string):
"""Decompress and deserialize string into a Numeric array."""
return Numeric.loads(decompress(string))
def serialize_via_marshal(obj):
"""Serialize Python object via marshal into a compressed string."""
return MySQLdb.escape_string(compress(marshal.dumps(obj)))
def deserialize_via_marshal(string):
"""Decompress and deserialize string into a Python object via marshal."""
return marshal.loads(decompress(string))
class WordTable:
"A class to hold the words table."
def __init__(self, tablename, fields_to_index, separators="[^\s]"):
"Creates words table instance."
self.tablename = tablename
self.recIDs_in_mem = []
self.fields_to_index = fields_to_index
self.separators = separators
self.value = {}
def get_field(self, recID, tag):
"""Returns list of values of the MARC-21 'tag' fields for the
record 'recID'."""
out = []
bibXXx = "bib" + tag[0] + tag[1] + "x"
bibrec_bibXXx = "bibrec_" + bibXXx
query = """SELECT value FROM %s AS b, %s AS bb
WHERE bb.id_bibrec=%s AND bb.id_bibxxx=b.id
AND tag LIKE '%s'""" % (bibXXx, bibrec_bibXXx, recID, tag);
res = run_sql(query)
for row in res:
out.append(row[0])
return out
def clean(self):
"Cleans the words table."
self.value={}
def put_into_db(self, mode="normal", split=string.split):
"""Updates the current words table in the corresponding MySQL's
idxFOO table. Mode 'normal' means normal execution,
mode 'emergency' means words index reverting to old state.
"""
if options["verbose"]:
write_message("%s %s wordtable flush started" % (self.tablename,mode))
write_message('...updating %d words into %sR started' % \
(len(self.value), self.tablename[:-1]))
task_update_progress("%s flushed %d/%d words" % (self.tablename, 0, len(self.value)))
self.recIDs_in_mem = beautify_range_list(self.recIDs_in_mem)
if mode == "normal":
for group in self.recIDs_in_mem:
query = """UPDATE %sR SET type='TEMPORARY' WHERE id_bibrec
BETWEEN '%d' AND '%d' AND type='CURRENT'""" % \
(self.tablename[:-1], group[0], group[1])
if options["verbose"] >= 9:
write_message(query)
run_sql(query)
nb_words_total = len(self.value)
nb_words_report = int(nb_words_total/10)
nb_words_done = 0
for word in self.value.keys():
self.put_word_into_db(word, self.value[word])
nb_words_done += 1
if nb_words_report!=0 and ((nb_words_done % nb_words_report) == 0):
if options["verbose"]:
write_message('......processed %d/%d words' % (nb_words_done, nb_words_total))
task_update_progress("%s flushed %d/%d words" % (self.tablename, nb_words_done, nb_words_total))
if options["verbose"] >= 9:
write_message('...updating %d words into %s ended' % \
(nb_words_total, self.tablename))
#if options["verbose"]:
# write_message('...updating reverse table %sR started' % self.tablename[:-1])
if mode == "normal":
for group in self.recIDs_in_mem:
query = """UPDATE %sR SET type='CURRENT' WHERE id_bibrec
BETWEEN '%d' AND '%d' AND type='FUTURE'""" % \
(self.tablename[:-1], group[0], group[1])
if options["verbose"] >= 9:
write_message(query)
run_sql(query)
query = """DELETE FROM %sR WHERE id_bibrec
BETWEEN '%d' AND '%d' AND type='TEMPORARY'""" % \
(self.tablename[:-1], group[0], group[1])
if options["verbose"] >= 9:
write_message(query)
run_sql(query)
if options["verbose"] >= 9:
write_message('End of updating wordTable into %s' % self.tablename)
elif mode == "emergency":
write_message("emergency")
for group in self.recIDs_in_mem:
query = """UPDATE %sR SET type='CURRENT' WHERE id_bibrec
BETWEEN '%d' AND '%d' AND type='TEMPORARY'""" % \
(self.tablename[:-1], group[0], group[1])
if options["verbose"] >= 9:
write_message(query)
run_sql(query)
query = """DELETE FROM %sR WHERE id_bibrec
BETWEEN '%d' AND '%d' AND type='FUTURE'""" % \
(self.tablename[:-1], group[0], group[1])
if options["verbose"] >= 9:
write_message(query)
run_sql(query)
if options["verbose"] >= 9:
write_message('End of emergency flushing wordTable into %s' % self.tablename)
#if options["verbose"]:
# write_message('...updating reverse table %sR ended' % self.tablename[:-1])
self.clean()
self.recIDs_in_mem = []
if options["verbose"]:
write_message("%s %s wordtable flush ended" % (self.tablename, mode))
task_update_progress("%s flush ended" % (self.tablename))
def load_old_recIDs(self,word):
"""Load existing hitlist for the word from the database index files."""
query = "SELECT hitlist FROM %s WHERE term=%%s" % self.tablename
res = run_sql(query, (word,))
if res:
return deserialize_via_marshal(res[0][0])
else:
return None
def merge_with_old_recIDs(self,word,recIDs, set):
"""Merge the system numbers stored in memory (hash of recIDs with value[0] > 0 or -1
according to whether to add/delete them) with those stored in the database index
and received in set universe of recIDs for the given word.
Return 0 in case no change was done to SET, return 1 in case SET was changed.
"""
set_changed_p = 0
for recID,sign in recIDs.iteritems():
if sign[0] == -1 and set.has_key(recID):
# delete recID if existent in set and if marked as to be deleted
del set[recID]
set_changed_p = 1
elif sign[0] > -1 and not set.has_key(recID):
# add recID if not existent in set and if marked as to be added
set[recID] = sign
set_changed_p = 1
elif sign[0] > -1 and sign[0] != set[recID][0]:
set[recID] = sign
set_changed_p = 1
return set_changed_p
def put_word_into_db(self, word, recIDs, split=string.split):
"""Flush a single word to the database and delete it from memory"""
set = self.load_old_recIDs(word)
#write_message("%s %s" % (word, self.value[word]))
if set: # merge the word recIDs found in memory:
if self.merge_with_old_recIDs(word, recIDs, set) == 0:
# nothing to update:
if options["verbose"] >= 9:
write_message("......... unchanged hitlist for ``%s''" % word)
pass
else:
# yes there were some new words:
if options["verbose"] >= 9:
write_message("......... updating hitlist for ``%s''" % word)
options["modified_words"][word] = 1
run_sql("UPDATE %s SET hitlist='%s' WHERE term='%s'" % (self.tablename, serialize_via_marshal(set), MySQLdb.escape_string(word)))
else: # the word is new, will create new set:
if options["verbose"] >= 9:
write_message("......... inserting hitlist for ``%s''" % word)
set = self.value[word]
if len(set) > 0:
#new word, add to list
options["modified_words"][word] = 1
run_sql("INSERT INTO %s (term, hitlist) VALUES ('%s', '%s')" % (self.tablename, MySQLdb.escape_string(word), serialize_via_marshal(set)))
if not set: # never store empty words
run_sql("DELETE from %s WHERE term=%%s" % self.tablename,
(word,))
del self.value[word]
def display(self):
"Displays the word table."
keys = self.value.keys()
keys.sort()
for k in keys:
if options["verbose"]:
write_message("%s: %s" % (k, self.value[k]))
def count(self):
"Returns the number of words in the table."
return len(self.value)
def info(self):
"Prints some information on the words table."
if options["verbose"]:
write_message("The words table contains %d words." % self.count())
def lookup_words(self, word=""):
"Lookup word from the words table."
if not word:
done = 0
while not done:
try:
word = raw_input("Enter word: ")
done = 1
except (EOFError, KeyboardInterrupt):
return
if self.value.has_key(word):
if options["verbose"]:
write_message("The word '%s' is found %d times." \
% (word, len(self.value[word])))
else:
if options["verbose"]:
write_message("The word '%s' does not exist in the word file."\
% word)
def update_last_updated(self, rank_method_code, starting_time=None):
"""Update last_updated column of the index table in the database.
Puts starting time there so that if the task was interrupted for record download,
the records will be reindexed next time."""
if starting_time is None:
return None
if options["verbose"] >= 9:
write_message("updating last_updated to %s...", starting_time)
return run_sql("UPDATE rnkMETHOD SET last_updated=%s WHERE name=%s",
(starting_time, rank_method_code,))
def add_recIDs(self, recIDs):
"""Fetches records which id in the recIDs range list and adds
them to the wordTable. The recIDs range list is of the form:
[[i1_low,i1_high],[i2_low,i2_high], ..., [iN_low,iN_high]].
"""
global chunksize
flush_count = 0
records_done = 0
records_to_go = 0
for range in recIDs:
records_to_go = records_to_go + range[1] - range[0] + 1
time_started = time.time() # will measure profile time
for range in recIDs:
i_low = range[0]
chunksize_count = 0
while i_low <= range[1]:
# calculate chunk group of recIDs and treat it:
i_high = min(i_low+options["flush"]-flush_count-1,range[1])
i_high = min(i_low+chunksize-chunksize_count-1, i_high)
try:
self.chk_recID_range(i_low, i_high)
except StandardError, e:
write_message("Exception caught: %s" % e, sys.stderr)
if options["verbose"] >= 9:
traceback.print_tb(sys.exc_info()[2])
task_update_status("ERROR")
task_sig_stop_commands()
sys.exit(1)
if options["verbose"]:
write_message("%s adding records #%d-#%d started" % \
(self.tablename, i_low, i_high))
if cfg_check_mysql_threads:
kill_sleepy_mysql_threads()
task_update_progress("%s adding recs %d-%d" % (self.tablename, i_low, i_high))
self.del_recID_range(i_low, i_high)
just_processed = self.add_recID_range(i_low, i_high)
flush_count = flush_count + i_high - i_low + 1
chunksize_count = chunksize_count + i_high - i_low + 1
records_done = records_done + just_processed
if options["verbose"]:
write_message("%s adding records #%d-#%d ended " % \
(self.tablename, i_low, i_high))
if chunksize_count >= chunksize:
chunksize_count = 0
# flush if necessary:
if flush_count >= options["flush"]:
self.put_into_db()
self.clean()
if options["verbose"]:
write_message("%s backing up" % (self.tablename))
flush_count = 0
self.log_progress(time_started,records_done,records_to_go)
# iterate:
i_low = i_high + 1
if flush_count > 0:
self.put_into_db()
self.log_progress(time_started,records_done,records_to_go)
def add_date(self, date=""):
# If date is not set, then retrieve it from the database.
# Reindex all formats newer than the modification date
if not date:
write_message("Using the last update time for the rank method")
id = self.tablename[len("bibindex"):]
query = """SELECT last_updated FROM rnkMETHOD WHERE name='%s'
""" % options["current_run"]
res = run_sql(query)
if not res:
return
if not res[0][0]:
date = ("0000-00-00",'')
else:
date = (res[0][0],'')
query = """SELECT b.id FROM bibrec AS b WHERE b.modification_date >=
'%s'""" % date[0]
if date[1]:
query += "and b.modification_date <= '%s'" % date[1]
query += "ORDER BY b.id ASC"""
res = run_sql(query)
list = create_range_list(res)
if not list:
if options["verbose"]:
write_message( "No new records added. %s is up to date" % self.tablename)
else:
self.add_recIDs(list)
return list
def add_recID_range(self, recID1, recID2):
empty_list_string = serialize_via_marshal([])
wlist = {}
normalize = {}
self.recIDs_in_mem.append([recID1,recID2])
# secondly fetch all needed tags:
for (tag, weight, lang) in self.fields_to_index:
if tag in tagToWordsFunctions.keys():
get_words_function = tagToWordsFunctions[ tag ]
else: get_words_function = get_words_from_phrase
bibXXx = "bib" + tag[0] + tag[1] + "x"
bibrec_bibXXx = "bibrec_" + bibXXx
query = """SELECT bb.id_bibrec,b.value FROM %s AS b, %s AS bb
WHERE bb.id_bibrec BETWEEN %d AND %d
AND bb.id_bibxxx=b.id AND tag LIKE '%s'""" % (bibXXx, bibrec_bibXXx, recID1, recID2, tag)
res = run_sql(query)
nb_total_to_read = len(res)
verbose_idx = 0 # for verbose pretty printing
for row in res:
recID, phrase = row
if options["validset"].contains(recID):
if not wlist.has_key(recID): wlist[recID] = {}
new_words = get_words_function(phrase, weight, lang) # ,self.separators
wlist[recID] = dict_union(new_words,wlist[recID])
# were there some words for these recIDs found?
if len(wlist) == 0: return 0
recIDs = wlist.keys()
for recID in recIDs:
# was this record marked as deleted?
if "DELETED" in self.get_field(recID, "980__c"):
wlist[recID] = {}
if options["verbose"] >= 9:
write_message("... record %d was declared deleted, removing its word list" % recID)
if options["verbose"] >= 9:
write_message("... record %d, termlist: %s" % (recID, wlist[recID]))
query_factory = cStringIO.StringIO()
qwrite = query_factory.write
qwrite( "INSERT INTO %sR (id_bibrec,termlist,type) VALUES" % self.tablename[:-1])
qwrite( "('" )
qwrite( str(recIDs[0]) )
qwrite( "','" )
qwrite( serialize_via_marshal(wlist[recIDs[0]]) )
qwrite( "','FUTURE')" )
for recID in recIDs[1:]:
qwrite(",('")
qwrite(str(recID))
qwrite("','")
qwrite(serialize_via_marshal(wlist[recID]))
qwrite("','FUTURE')")
query = query_factory.getvalue()
query_factory.close()
run_sql(query)
query_factory = cStringIO.StringIO()
qwrite = query_factory.write
qwrite("INSERT INTO %sR (id_bibrec,termlist,type) VALUES" % self.tablename[:-1])
qwrite("('")
qwrite(str(recIDs[0]))
qwrite("','")
qwrite(serialize_via_marshal(wlist[recIDs[0]]))
qwrite("','CURRENT')")
for recID in recIDs[1:]:
qwrite( ",('" )
qwrite( str(recID) )
qwrite( "','" )
qwrite( empty_list_string )
qwrite( "','CURRENT')" )
query = query_factory.getvalue()
query_factory.close()
try:
run_sql(query)
except MySQLdb.DatabaseError:
pass
put = self.put
for recID in recIDs:
for (w, count) in wlist[recID].iteritems():
put(recID, w, count)
return len(recIDs)
def log_progress(self, start, done, todo):
"""Calculate progress and store it.
start: start time,
done: records processed,
todo: total number of records"""
time_elapsed = time.time() - start
# consistency check
if time_elapsed == 0 or done > todo:
return
time_recs_per_min = done/(time_elapsed/60.0)
if options["verbose"]:
write_message("%d records took %.1f seconds to complete.(%1.f recs/min)"\
% (done, time_elapsed, time_recs_per_min))
if time_recs_per_min:
if options["verbose"]:
write_message("Estimated runtime: %.1f minutes" % \
((todo-done)/time_recs_per_min))
def put(self, recID, word, sign):
"Adds/deletes a word to the word list."
try:
word = lower(word[:50])
if self.value.has_key(word):
# the word 'word' exist already: update sign
self.value[word][recID] = sign
# PROBLEM ?
else:
self.value[word] = {recID: sign}
except:
write_message("Error: Cannot put word %s with sign %d for recID %s." % (word, sign, recID))
def del_recIDs(self, recIDs):
"""Fetches records which id in the recIDs range list and adds
them to the wordTable. The recIDs range list is of the form:
[[i1_low,i1_high],[i2_low,i2_high], ..., [iN_low,iN_high]].
"""
count = 0
for range in recIDs:
self.del_recID_range(range[0],range[1])
count = count + range[1] - range[0]
self.put_into_db()
def del_recID_range(self, low, high):
"""Deletes records with 'recID' system number between low
and high from memory words index table."""
if options["verbose"] > 2:
write_message("%s fetching existing words for records #%d-#%d started" % \
(self.tablename, low, high))
self.recIDs_in_mem.append([low,high])
query = """SELECT id_bibrec,termlist FROM %sR as bb WHERE bb.id_bibrec
BETWEEN '%d' AND '%d'""" % (self.tablename[:-1], low, high)
recID_rows = run_sql(query)
for recID_row in recID_rows:
recID = recID_row[0]
wlist = deserialize_via_marshal(recID_row[1])
for word in wlist:
self.put(recID, word, (-1, 0))
if options["verbose"] > 2:
write_message("%s fetching existing words for records #%d-#%d ended" % \
(self.tablename, low, high))
def report_on_table_consistency(self):
"""Check reverse words index tables (e.g. rnkWORD01R) for
interesting states such as 'TEMPORARY' state.
Prints small report (no of words, no of bad words).
"""
# find number of words:
query = """SELECT COUNT(*) FROM %s""" % (self.tablename)
res = run_sql(query, None, 1)
if res:
nb_words = res[0][0]
else:
nb_words = 0
# find number of records:
query = """SELECT COUNT(DISTINCT(id_bibrec)) FROM %sR""" % (self.tablename[:-1])
res = run_sql(query, None, 1)
if res:
nb_records = res[0][0]
else:
nb_records = 0
# report stats:
if options["verbose"]:
write_message("%s contains %d words from %d records" % (self.tablename, nb_words, nb_records))
# find possible bad states in reverse tables:
query = """SELECT COUNT(DISTINCT(id_bibrec)) FROM %sR WHERE type <> 'CURRENT'""" % (self.tablename[:-1])
res = run_sql(query)
if res:
nb_bad_records = res[0][0]
else:
nb_bad_records = 999999999
if nb_bad_records:
write_message("EMERGENCY: %s needs to repair %d of %d records" % \
(self.tablename, nb_bad_records, nb_records))
else:
if options["verbose"]:
write_message("%s is in consistent state" % (self.tablename))
return nb_bad_records
def repair(self):
"""Repair the whole table"""
# find possible bad states in reverse tables:
query = """SELECT COUNT(DISTINCT(id_bibrec)) FROM %sR WHERE type <> 'CURRENT'""" % (self.tablename[:-1])
res = run_sql(query, None, 1)
if res:
nb_bad_records = res[0][0]
else:
nb_bad_records = 0
# find number of records:
query = """SELECT COUNT(DISTINCT(id_bibrec)) FROM %sR""" % (self.tablename[:-1])
res = run_sql(query)
if res:
nb_records = res[0][0]
else:
nb_records = 0
if nb_bad_records == 0:
return
query = """SELECT id_bibrec FROM %sR WHERE type <> 'CURRENT' ORDER BY id_bibrec""" \
% (self.tablename[:-1])
res = run_sql(query)
recIDs = create_range_list(res)
flush_count = 0
records_done = 0
records_to_go = 0
for range in recIDs:
records_to_go = records_to_go + range[1] - range[0] + 1
time_started = time.time() # will measure profile time
for range in recIDs:
i_low = range[0]
chunksize_count = 0
while i_low <= range[1]:
# calculate chunk group of recIDs and treat it:
i_high = min(i_low+options["flush"]-flush_count-1,range[1])
i_high = min(i_low+chunksize-chunksize_count-1, i_high)
try:
self.fix_recID_range(i_low, i_high)
except StandardError, e:
write_message("Exception caught: %s" % e, sys.stderr)
if options["verbose"] >= 9:
traceback.print_tb(sys.exc_info()[2])
task_update_status("ERROR")
task_sig_stop_commands()
sys.exit(1)
flush_count = flush_count + i_high - i_low + 1
chunksize_count = chunksize_count + i_high - i_low + 1
records_done = records_done + i_high - i_low + 1
if chunksize_count >= chunksize:
chunksize_count = 0
# flush if necessary:
if flush_count >= options["flush"]:
self.put_into_db("emergency")
self.clean()
flush_count = 0
self.log_progress(time_started,records_done,records_to_go)
# iterate:
i_low = i_high + 1
if flush_count > 0:
self.put_into_db("emergency")
self.log_progress(time_started,records_done,records_to_go)
write_message("%s inconsistencies repaired." % self.tablename)
def chk_recID_range(self, low, high):
"""Check if the reverse index table is in proper state"""
## check db
query = """SELECT COUNT(*) FROM %sR WHERE type <> 'CURRENT'
AND id_bibrec BETWEEN '%d' AND '%d'""" % (self.tablename[:-1], low, high)
res = run_sql(query, None, 1)
if res[0][0]==0:
if options["verbose"]:
write_message("%s for %d-%d is in consistent state"%(self.tablename,low,high))
return # okay, words table is consistent
## inconsistency detected!
write_message("EMERGENCY: %s inconsistencies detected..." % self.tablename)
write_message("""EMERGENCY: Errors found. You should check consistency of the %s - %sR tables.\nRunning 'bibindex --repair' is recommended.""" \
% (self.tablename, self.tablename[:-1]))
raise StandardError
def fix_recID_range(self, low, high):
"""Try to fix reverse index database consistency (e.g. table rnkWORD01R) in the low,high doc-id range.
Possible states for a recID follow:
CUR TMP FUT: very bad things have happened: warn!
CUR TMP : very bad things have happened: warn!
CUR FUT: delete FUT (crash before flushing)
CUR : database is ok
TMP FUT: add TMP to memory and del FUT from memory
flush (revert to old state)
TMP : very bad things have happened: warn!
FUT: very bad things have happended: warn!
"""
state = {}
query = "SELECT id_bibrec,type FROM %sR WHERE id_bibrec BETWEEN '%d' AND '%d'"\
% (self.tablename[:-1], low, high)
res = run_sql(query)
for row in res:
if not state.has_key(row[0]):
state[row[0]]=[]
state[row[0]].append(row[1])
ok = 1 # will hold info on whether we will be able to repair
for recID in state.keys():
if not 'TEMPORARY' in state[recID]:
if 'FUTURE' in state[recID]:
if 'CURRENT' not in state[recID]:
write_message("EMERGENCY: Record %d is in inconsistent state. Can't repair it" % recID)
ok = 0
else:
write_message("EMERGENCY: Inconsistency in record %d detected" % recID)
query = """DELETE FROM %sR
WHERE id_bibrec='%d'""" % (self.tablename[:-1], recID)
run_sql(query)
write_message("EMERGENCY: Inconsistency in record %d repaired." % recID)
else:
if 'FUTURE' in state[recID] and not 'CURRENT' in state[recID]:
self.recIDs_in_mem.append([recID,recID])
# Get the words file
query = """SELECT type,termlist FROM %sR
WHERE id_bibrec='%d'""" % (self.tablename[:-1], recID)
if options["verbose"] >= 9:
write_message(query)
res = run_sql(query)
for row in res:
wlist = deserialize_via_marshal(row[1])
if options["verbose"] >= 9:
write_message("Words are %s " % wlist)
if row[0] == 'TEMPORARY':
sign = 1
else:
sign = -1
for word in wlist:
self.put(recID, word, wlist[word])
else:
write_message("EMERGENCY: %s for %d is in inconsistent state. Couldn't repair it." % (self.tablename, recID))
ok = 0
if not ok:
write_message("""EMERGENCY: Unrepairable errors found. You should check consistency
of the %s - %sR tables. Deleting affected records is
recommended.""" % (self.tablename, self.tablename[:-1]))
raise StandardError
def word_index(row, run):
"""Run the indexing task. The row argument is the BibSched task
queue row, containing if, arguments, etc.
Return 1 in case of success and 0 in case of failure.
"""
## import optional modules:
try:
import psyco
psyco.bind(get_words_from_phrase)
psyco.bind(WordTable.merge_with_old_recIDs)
psyco.bind(serialize_via_numeric_array)
psyco.bind(serialize_via_marshal)
psyco.bind(deserialize_via_numeric_array)
psyco.bind(deserialize_via_marshal)
psyco.bind(update_rnkWORD)
psyco.bind(check_rnkWORD)
except StandardError,e:
print "ERROR PSYCO", e
pass
global options, task_id, wordTables, stopwords, stemmer, languages
languages = {'fr': 'french', 'en': 'porter', 'no':'norwegian', 'se':'swedish', 'de': 'german', 'it':'italian', 'pt':'portugese'}
stemmer = {}
# read from SQL row:
task_id = row[0]
task_proc = row[1]
options = marshal.loads(row[6])
task_status = row[7]
# sanity check:
if task_proc != "runbibrank":
write_message("The task #%d does not seem to be a BibRank task." % task_id, sys.stderr)
return 0
if task_status != "WAITING":
write_message("The task #%d is %s. I expected WAITING." % (task_id, task_status), sys.stderr)
return 0
# we can run the task now:
if options["verbose"]:
write_message("Task #%d started." % task_id)
task_update_status("RUNNING")
# install signal handlers
signal.signal(signal.SIGUSR1, task_sig_sleep)
signal.signal(signal.SIGTERM, task_sig_stop)
signal.signal(signal.SIGABRT, task_sig_suicide)
signal.signal(signal.SIGCONT, task_sig_wakeup)
signal.signal(signal.SIGINT, task_sig_unknown)
## go ahead and treat each table:
options["run"] = []
options["run"].append(run)
for rank_method_code in options["run"]:
method_starting_time = time.strftime("%Y-%m-%d %H:%M:%S", time.localtime())
write_message("Running rank method: %s" % getName(rank_method_code))
try:
file = etcdir + "/bibrank/" + rank_method_code + ".cfg"
if options["verbose"] >= 9:
write_message("Getting configuration from file: %s" % file)
config = ConfigParser.ConfigParser()
config.readfp(open(file))
except StandardError, e:
write_message("Cannot find configurationfile: %s" % file, sys.stderr)
raise StandardError
options["current_run"] = rank_method_code
options["modified_words"] = {}
options["table"] = config.get(config.get("rank_method", "function"), "table")
(stopwords, tags) = get_tags(config) #get the tags to include
options["validset"] = get_valid_range(rank_method_code) #get the records from the collections the method is enabled for
wordTable = WordTable(options["table"], tags)
wordTable.report_on_table_consistency()
try:
if options["cmd"] == "del":
if options["id"]:
wordTable.del_recIDs(options["id"])
elif options["collection"]:
l_of_colls = string.split(options["collection"], ",")
recIDs = perform_request_search(c=l_of_colls)
recIDs_range = []
for recID in recIDs:
recIDs_range.append([recID,recID])
wordTable.del_recIDs(recIDs_range)
else:
write_message("Missing IDs of records to delete from index %s.", wordTable.tablename,
sys.stderr)
raise StandardError
elif options["cmd"] == "add":
if options["id"]:
wordTable.add_recIDs(options["id"])
elif options["collection"]:
l_of_colls = string.split(options["collection"], ",")
recIDs = perform_request_search(c=l_of_colls)
recIDs_range = []
for recID in recIDs:
recIDs_range.append([recID,recID])
wordTable.add_recIDs(recIDs_range)
elif options["last_updated"]:
wordTable.add_date("")
wordTable.update_last_updated(rank_method_code, method_starting_time)
else:
wordTable.add_recIDs([[0,cfg_max_recID]])
#wordTable.add_date(options["modified"])
# only update last_updated if run via automatic mode:
elif options["cmd"] == "repair":
wordTable.repair()
check_rnkWORD(options["table"])
elif options["cmd"] == "check":
check_rnkWORD(options["table"])
options["modified_words"] = {}
elif options["cmd"] == "stat":
rank_method_code_statistics(options["table"])
else:
write_message("Invalid command found processing %s" % \
wordTable.tablename, sys.stderr)
raise StandardError
update_rnkWORD(options["table"], options["modified_words"])
except StandardError, e:
write_message("Exception caught: %s" % e, sys.stderr)
if options["verbose"] >= 9:
traceback.print_tb(sys.exc_info()[2])
task_update_status("ERROR")
task_sig_stop_commands()
sys.exit(1)
wordTable.report_on_table_consistency()
# We are done. State it in the database, close and quit
task_update_status("DONE")
if options["verbose"]:
write_message("Task #%d finished." % task_id)
return 1
def get_tags(config):
"""Get the tags that should be used creating the index and each tag's parameter."""
tags = []
function = config.get("rank_method","function")
i = 1
shown_error = 0
try:
while config.has_option(function,"tag%s"% i):
tag = config.get(function, "tag%s" % i)
tag = string.split(tag, ",")
tag[1] = int(string.strip(tag[1]))
tag[2] = string.strip(tag[2])
#get stopwordlist
stopwords = get_stopwords("%s" % config.get(function, "stopword"))
try:
#get stemmer for language if possible
if tag[2] and tag[2] != "none" and languages.has_key(tag[2]) and config.get(function,"stem_if_avail") == "yes" and not stemmer.has_key(tag[2]):
stemmer[tag[2]] = Stemmer.Stemmer(languages[tag[2]])
if options["verbose"] >=9:
write_message("Stemmer for language '%s' initiated" % tag[2])
elif tag[2] != "none" and config.get(function,"stem_if_avail") == "yes" and not stemmer.has_key(tag[2]):
write_message("Warning: Language '%s' not available in PyStemmer." % tag[2])
except NameError:
if shown_error == 0:
write_message("Warning: PyStemmer not found. Please read INSTALL.")
shown_error = 1
tags.append(tag)
i += 1
except Exception:
write_message("Could not read data from configuration file, please check for errors")
raise StandardError
return (stopwords, tags)
def get_valid_range(rank_method_code):
"""Returns which records are valid for this rank method, according to which collections it is enabled for."""
#if options["verbose"] >=9:
# write_message("Getting records from collections enabled for rank method.")
#res = run_sql("SELECT collection.name FROM collection,collection_rnkMETHOD,rnkMETHOD WHERE collection.id=id_collection and id_rnkMETHOD=rnkMETHOD.id and rnkMETHOD.name='%s'" % rank_method_code)
#l_of_colls = []
#for coll in res:
# l_of_colls.append(coll[0])
#if len(l_of_colls) > 0:
# recIDs = perform_request_search(c=l_of_colls)
#else:
# recIDs = []
valid = HitSet(Numeric.ones(cfg_max_recID+1, Numeric.Int0))
#valid.addlist(recIDs)
return valid
def write_message(msg, stream=sys.stdout):
"""Prints message and flush output stream (may be sys.stdout or sys.stderr)."""
if stream == sys.stdout or stream == sys.stderr:
stream.write(time.strftime("%Y-%m-%d %H:%M:%S --> ", time.localtime()))
stream.write("%s\n" % msg)
stream.flush()
else:
sys.stderr.write("Unknown stream %s. [must be sys.stdout or sys.stderr]\n" % stream)
def get_stopwords(file='stopwords.txt'):
"""Get stopword list"""
stopwords = open(file, 'r')
lines = stopwords.readlines()
stopwords.close()
stopwords = {}
for line in lines:
stopwords[string.rstrip(line)] = 1
return stopwords
def check_term(term, termlength):
"""Check if term is a stopword, or for any other reasons for not using this term."""
try:
if stopwords.has_key(lower(term)) or (len(term) <= termlength):
return ""
reg = re.compile(r"[1234567890\!\"\#\$\%\&\'\(\)\*\+\,\-\.\/\:\;\<\=\>\?\@\[\\\]\^\_\`\{\|\}\~]")
if re.search(reg, term):
return ""
term = str.replace(term, "-", "")
term = str.replace(term, ".", "")
term = str.replace(term, ",", "")
if int(term):
return ""
except StandardError, e:
pass
return "true"
def check_rnkWORD(table):
"""Checks for any problems in rnkWORD tables."""
i = 0
errors = {}
termslist = run_sql("SELECT term FROM %s" % table)
N = run_sql("select max(id_bibrec) from %sR" % table[:-1])[0][0]
write_message("Checking integrity of rank values in %s" % table)
terms = map(lambda x: x[0], termslist)
while i < len(terms):
current_terms = ""
for j in range(i, ((i+5000)< len(terms) and (i+5000) or len(terms))):
current_terms += "'%s'," % terms[j]
terms_docs = run_sql("SELECT term, hitlist FROM %s WHERE term in (%s)" % (table, current_terms[:-1]))
for (t, hitlist) in terms_docs:
term_docs = deserialize_via_marshal(hitlist)
if (term_docs.has_key("Gi") and term_docs["Gi"][1] == 0) or not term_docs.has_key("Gi"):
write_message("ERROR: Missing value for term: %s (%s) in %s: %s" % (t, repr(t), table, len(term_docs)))
errors[t] = 1
i += 5000
write_message("Checking integrity of rank values in %sR" % table[:-1])
i = 0
while i < N:
docs_terms = run_sql("SELECT id_bibrec, termlist FROM %sR WHERE id_bibrec>=%s and id_bibrec<=%s" % (table[:-1], i, i+5000))
for (j, termlist) in docs_terms:
termlist = deserialize_via_marshal(termlist)
for (t, tf) in termlist.iteritems():
if tf[1] == 0 and not errors.has_key(t):
errors[t] = 1
write_message("ERROR: Gi missing for record %s and term: %s (%s) in %s" % (j,t,repr(t), table))
terms_docs = run_sql("SELECT term, hitlist FROM %s WHERE term='%s'" % (table, t))
termlist = deserialize_via_marshal(terms_docs[0][1])
i += 5000
if len(errors) == 0:
write_message("No direct errors found, but nonconsistent data may exist.")
else:
write_message("%s errors found during integrity check, repair and rebalancing recommended." % len(errors))
options["modified_words"] = errors
def rank_method_code_statistics(table):
"""Shows some statistics about this rank method."""
maxID = run_sql("select max(id) from %s" % table)
maxID = maxID[0][0]
terms = {}
Gi = {}
write_message("Showing statistics of terms in index:")
write_message("Important: For the 'Least used terms', the number of terms is shown first, and the number of occurences second.")
write_message("Least used terms---Most important terms---Least important terms")
i = 0
while i < maxID:
terms_docs=run_sql("SELECT term, hitlist FROM %s WHERE id>= %s and id < %s" % (table, i, i + 10000))
for (t, hitlist) in terms_docs:
term_docs=deserialize_via_marshal(hitlist)
terms[len(term_docs)] = terms.get(len(term_docs), 0) + 1
if term_docs.has_key("Gi"):
Gi[t] = term_docs["Gi"]
i=i + 10000
terms=terms.items()
terms.sort(lambda x, y: cmp(y[1], x[1]))
Gi=Gi.items()
Gi.sort(lambda x, y: cmp(y[1], x[1]))
for i in range(0, 20):
write_message("%s/%s---%s---%s" % (terms[i][0],terms[i][1], Gi[i][0],Gi[len(Gi) - i - 1][0]))
def update_rnkWORD(table, terms):
"""Updates rnkWORDF and rnkWORDR with Gi and Nj values. For each term in rnkWORDF, a Gi value for the term is added. And for each term in each document, the Nj value for that document is added. In rnkWORDR, the Gi value for each term in each document is added. For description on how things are computed, look in the hacking docs.
table - name of forward index to update
terms - modified terms"""
stime = time.time()
Gi = {}
Nj = {}
N = run_sql("select count(id_bibrec) from %sR" % table[:-1])[0][0]
if len(terms) == 0:
return ""
write_message("Beginning post-processing of %s terms" % len(terms))
#Locating all documents related to the modified/new/deleted terms, if fast update,
#only take into account new/modified occurences
write_message("Phase 1: Finding records containing modified terms")
terms = terms.keys()
i = 0
while i < len(terms):
current_terms = ""
for j in range(i, ((i+5000)< len(terms) and (i+5000) or len(terms))):
current_terms += "'%s'," % terms[j]
terms_docs = run_sql("SELECT term, hitlist FROM %s WHERE term in (%s)" % (table,current_terms[:-1]))
for (t, hitlist) in terms_docs:
term_docs = deserialize_via_marshal(hitlist)
if term_docs.has_key("Gi"):
del term_docs["Gi"]
for (j, tf) in term_docs.iteritems():
if (options["quick"] == "yes" and tf[1] == 0) or options["quick"] == "no":
Nj[j] = 0
write_message("Phase 1: ......processed %s/%s terms" % ((i+5000>len(terms) and len(terms) or (i+5000)), len(terms)))
i += 5000
write_message("Phase 1: Finished finding records containing modified terms")
#Find all terms in the records found in last phase
write_message("Phase 2: Finding all terms in affected records")
records = Nj.keys()
i = 0
while i < len(records):
current_recs = "%s" % records[i:i+5000]
docs_terms = run_sql("SELECT id_bibrec, termlist FROM %sR WHERE id_bibrec in (%s)" % (table[:-1],current_recs[1:-1]))
for (j, termlist) in docs_terms:
doc_terms = deserialize_via_marshal(termlist)
for (t, tf) in doc_terms.iteritems():
Gi[t] = 0
write_message("Phase 2: ......processed %s/%s records " % ((i+5000>len(records) and len(records) or (i+5000)), len(records)))
i += 5000
write_message("Phase 2: Finished finding all terms in affected records")
terms = Gi.keys()
Gi = {}
i = 0
if options["quick"] == "no":
#Calculating Fi and Gi value for each term
write_message("Phase 3: Calculating importance of all affected terms")
while i < len(terms):
current_terms = ""
for j in range(i, ((i+5000)< len(terms) and (i+5000) or len(terms))):
current_terms += "'%s'," % terms[j]
terms_docs = run_sql("SELECT term, hitlist FROM %s WHERE term in (%s)" % (table,current_terms[:-1]))
for (t, hitlist) in terms_docs:
term_docs = deserialize_via_marshal(hitlist)
if term_docs.has_key("Gi"):
del term_docs["Gi"]
Fi = 0
Gi[t] = 1
for (j, tf) in term_docs.iteritems():
Fi += tf[0]
for (j, tf) in term_docs.iteritems():
if tf[0] != Fi:
Gi[t] = Gi[t] + ((float(tf[0]) / Fi) * math.log(float(tf[0]) / Fi, 2)) / math.log(N)
write_message("Phase 3: ......processed %s/%s terms" % ((i+5000>len(terms) and len(terms) or (i+5000)), len(terms)))
i += 5000
write_message("Phase 3: Finished calculating importance of all affected terms")
else:
#Using existing Gi value instead of calculating a new one. Missing some accurancy.
write_message("Phase 3: Getting approximate importance of all affected terms")
while i < len(terms):
current_terms = ""
for j in range(i, ((i+5000)< len(terms) and (i+5000) or len(terms))):
current_terms += "'%s'," % terms[j]
terms_docs = run_sql("SELECT term, hitlist FROM %s WHERE term in (%s)" % (table,current_terms[:-1]))
for (t, hitlist) in terms_docs:
term_docs = deserialize_via_marshal(hitlist)
if term_docs.has_key("Gi"):
Gi[t] = term_docs["Gi"][1]
else:
write_message("Please rebalance these records,(may require deletion before adding again)")
raise Exception("Rebalance")
write_message("Phase 3: ......processed %s/%s terms" % ((i+5000>len(terms) and len(terms) or (i+5000)), len(terms)))
i += 5000
write_message("Phase 3: Finished getting approximate importance of all affected terms")
write_message("Phase 4: Calculating normalization value for all affected records and updating %sR" % table[:-1])
records = Nj.keys()
i = 0
while i < len(records):
#Calculating the normalization value for each document, and adding the Gi value to each term in each document.
current_recs = "%s" % records[i:i+5000]
current_recs = current_recs[1:-1]
docs_terms = run_sql("SELECT id_bibrec, termlist FROM %sR WHERE id_bibrec IN (%s)" % (table[:-1],current_recs))
for (j, termlist) in docs_terms:
doc_terms = deserialize_via_marshal(termlist)
for (t, tf) in doc_terms.iteritems():
if Gi.has_key(t):
Nj[j] = Nj.get(j, 0) + math.pow(Gi[t] * (1 + math.log(tf[0])),2)
Git = int(math.floor(Gi[t]*100))
if Git >= 0:
Git += 1
doc_terms[t] = (tf[0], Git)
else:
Nj[j] = Nj.get(j, 0) + math.pow(tf[1] * (1 + math.log(tf[0])),2)
Nj[j] = 1.0 / math.sqrt(Nj[j])
Nj[j] = int(Nj[j] * 100)
if Nj[j] >= 0:
Nj[j] += 1
run_sql("UPDATE %sR SET termlist='%s' WHERE id_bibrec=%s" % (table[:-1], serialize_via_marshal(doc_terms), j))
write_message("Phase 4: ......processed %s/%s records" % ((i+5000>len(records) and len(records) or (i+5000)), len(records)))
i += 5000
write_message("Phase 4: Finished calculating normalization value for all affected records and updating %sR" % table[:-1])
write_message("Phase 5: Updating %s with new normalization values" % table)
i = 0
terms = Gi.keys()
while i < len(terms):
#Adding the Gi value to each term, and adding the normalization value to each term in each document.
current_terms = ""
for j in range(i, ((i+5000)< len(terms) and (i+5000) or len(terms))):
current_terms += "'%s'," % terms[j]
terms_docs = run_sql("SELECT term, hitlist FROM %s WHERE term IN (%s)" % (table,current_terms[:-1]))
for (t, hitlist) in terms_docs:
term_docs = deserialize_via_marshal(hitlist)
if term_docs.has_key("Gi"):
del term_docs["Gi"]
for (j, tf) in term_docs.iteritems():
if Nj.has_key(j):
term_docs[j] = (tf[0], Nj[j])
Git = int(math.floor(Gi[t]*100))
if Git >= 0:
Git += 1
term_docs["Gi"] = (0, Git)
run_sql("UPDATE %s SET hitlist='%s' WHERE term='%s'" % (table, serialize_via_marshal(term_docs), MySQLdb.escape_string(t)))
write_message("Phase 5: ......processed %s/%s terms" % ((i+5000>len(terms) and len(terms) or (i+5000)), len(terms)))
i += 5000
write_message("Phase 5: Finished updating %s with new normalization values" % table)
write_message("Time used for post-processing: %.1fmin" % ((time.time() - stime) / 60))
write_message("Finished post-processing")
def test_word_separators(phrase="hep-th/0101001"):
"""Tests word separating policy on various input."""
print "%s:" % phrase
gwfp = get_words_from_phrase(phrase)
for (word, count) in gwfp.iteritems():
print "\t-> %s - %s" % (word, count)
def task_sig_sleep(sig, frame):
"""Signal handler for the 'sleep' signal sent by BibSched."""
if options["verbose"] >= 9:
write_message("got signal %d" % sig)
write_message("sleeping...")
task_update_status("SLEEPING")
signal.pause() # wait for wake-up signal
def task_sig_wakeup(sig, frame):
"""Signal handler for the 'wakeup' signal sent by BibSched."""
if options["verbose"] >= 9:
write_message("got signal %d" % sig)
write_message("continuing...")
task_update_status("CONTINUING")
def task_sig_stop(sig, frame):
"""Signal handler for the 'stop' signal sent by BibSched."""
if options["verbose"] >= 9:
write_message("got signal %d" % sig)
write_message("stopping...")
task_update_status("STOPPING")
errcode = 0
try:
task_sig_stop_commands()
write_message("stopped")
task_update_status("STOPPED")
except StandardError, err:
write_message("Error during stopping! %e" % err)
task_update_status("STOPPINGFAILED")
errcode = 1
sys.exit(errcode)
def task_sig_stop_commands():
"""Do all the commands necessary to stop the task before quitting.
Useful for task_sig_stop() handler.
"""
write_message("stopping commands started")
for table in wordTables:
table.put_into_db()
write_message("stopping commands ended")
def task_sig_suicide(sig, frame):
"""Signal handler for the 'suicide' signal sent by BibSched."""
if options["verbose"] >= 9:
write_message("got signal %d" % sig)
write_message("suiciding myself now...")
task_update_status("SUICIDING")
write_message("suicided")
task_update_status("SUICIDED")
sys.exit(0)
def task_sig_unknown(sig, frame):
"""Signal handler for the other unknown signals sent by shell or user."""
if options["verbose"] >= 9:
write_message("got signal %d" % sig)
write_message("unknown signal %d ignored" % sig) # do nothing for other signals
def task_update_progress(msg):
"""Updates progress information in the BibSched task table."""
global task_id, options
if options["verbose"] >= 9:
write_message("Updating task progress to %s." % msg)
return run_sql("UPDATE schTASK SET progress=%s where id=%s", (msg, task_id))
def task_update_status(val):
"""Updates state information in the BibSched task table."""
global task_id, options
if options["verbose"] >= 9:
write_message("Updating task status to %s." % val)
return run_sql("UPDATE schTASK SET status=%s where id=%s", (val, task_id))
def getName(methname, ln=cdslang, type='ln'):
"""Returns the name of the rank method, either in default language or given language.
methname = short name of the method
ln - the language to get the name in
type - which name "type" to get."""
try:
rnkid = run_sql("SELECT id FROM rnkMETHOD where name='%s'" % methname)
if rnkid:
rnkid = str(rnkid[0][0])
res = run_sql("SELECT value FROM rnkMETHODNAME where type='%s' and ln='%s' and id_rnkMETHOD=%s" % (type, ln, rnkid))
if not res:
res = run_sql("SELECT value FROM rnkMETHODNAME WHERE ln='%s' and id_rnkMETHOD=%s and type='%s'" % (cdslang, rnkid, type))
if not res:
return methname
return res[0][0]
else:
raise Exception
except Exception, e:
write_message("Cannot run rank method, either given code for method is wrong, or it has not been added using the webinterface.")
raise Exception
def word_frequency(row, run):
return word_index(row, run)
c4science · Help