"""
Full-text searching of logs
"""
import datetime, calendar, pytz
import os.path
import HyperEstraier as hype
import log_line
class LogSearchError (Exception) :
"""
General search error
"""
pass
class NoResultsFound (LogSearchError) :
"""
No results found
"""
pass
class LogSearchIndex (object) :
"""
An index on the logs for a group of channels.
This uses Hyper Estraier to handle searching, whereby each log line is a document (yes, I have a powerful server).
These log documents have the following attributes:
@uri - channel/date/line
channel - channel code
type - the LogType id
timestamp - UTC timestamp
source_nickname - source nickname
Each document then has a single line of data, which is the log message itself
"""
def __init__ (self, channels, path, mode='r') :
"""
Open the database at the given path, with the given mode:
r - read-only
w - write, create if not exists
a - write, error if not exists
c - write, create, error if exists
* - write, create, truncate if exists
Channels is the ChannelList.
"""
# store
self.channels = channels
self.path = path
self.mode = mode
# check it does not already exist?
if mode in 'c' and os.path.exists(path) :
raise LogSearchError("Index already exists: %s" % (path, ))
# mapping of { mode -> flags }
mode_to_flag = {
'r': hype.Database.DBREADER,
'w': hype.Database.DBWRITER | hype.Database.DBCREAT,
'a': hype.Database.DBWRITER,
'c': hype.Database.DBWRITER | hype.Database.DBCREAT,
'*': hype.Database.DBWRITER | hype.Database.DBCREAT | hype.Database.DBTRUNC,
}
# look up flags
flags = mode_to_flag[mode]
# make instance
self.db = hype.Database()
# open
if not self.db.open(path, flags) :
raise Exception("Index open failed: %s, mode=%s, flags=%#06x: %s" % (path, mode, flags, self.db.err_msg(self.db.error())))
def insert (self, channel, lines) :
"""
Adds a sequence of LogLines from the given LogChannel to the index, and return the number of added items
"""
# validate the LogChannel
assert channel.name
count = 0
# iterate
for line in lines :
# validate the LogLine
assert line.offset
assert line.timestamp
# create new document
doc = hype.Document()
# line date
date = line.timestamp.date()
# convert to UTC timestamp
utc_timestamp = calendar.timegm(line.timestamp.utctimetuple())
# ensure that it's not 1900
assert date.year != 1900
# add URI
doc.add_attr('@uri', "%s/%s/%d" % (channel.id, date.strftime('%Y-%m-%d'), line.offset))
# add channel id
doc.add_attr('channel', channel.id)
# add type
doc.add_attr('type', str(line.type))
# add UTC timestamp
doc.add_attr('timestamp', str(utc_timestamp))
# add source attribute?
if line.source :
source_nickname, source_username, source_hostname, source_chanflags = line.source
# XXX: handle source_nickname is None
if not source_nickname is None :
source_nickname = str(source_nickname)
doc.add_attr('source_nickname', source_nickname)
# add data
if line.data :
doc.add_text(line.data.encode('utf8'))
# put, "clean up dispensable regions of the overwritten document"
if not self.db.put_doc(doc, hype.Database.PDCLEAN) :
raise Exeception("Index put_doc failed")
# count
count += 1
# return
return count
def search_cond (self, cond) :
"""
Search using a raw hype.Condition. Raises NoResultsFound if there aren't any results
"""
# execute search, unused 'flags' arg stays zero
results = self.db.search(cond, 0)
# no results?
if not results :
raise NoResultsFound()
# iterate over the document IDs
for doc_id in results :
# load document, this throws an exception...
# option constants are hype.Database.GDNOATTR/GDNOTEXT
doc = self.db.get_doc(doc_id, 0)
# load the attributes/text
channel = self.channels.lookup(doc.attr('channel'))
type = int(doc.attr('type'))
timestamp = datetime.datetime.fromtimestamp(int(doc.attr('timestamp')), pytz.utc)
source_nickname = doc.attr('source_nickname')
message = doc.cat_texts().decode('utf8')
# build+yield to as LogLine
yield log_line.LogLine(channel, None, type, timestamp, (source_nickname, None, None, None), None, message)
def search (self, options=None, channel=None, phrase=None, order=None, max=None, skip=None) :
"""
Search with flexible parameters
options - bitmask of hype.Condition.*
channel - LogChannel object
phrase - the search query phrase
order - order attribute expression
max - number of results to return
skip - number of results to skip
"""
# build condition
cond = hype.Condition()
if options :
# set options
cond.set_options(options)
if channel :
# add channel attribute
cond.add_attr("@channel STREQ %s" % (channel.id, ))
if phrase :
# add phrase
cond.set_phrase(phrase)
if order :
# set order
cond.set_order(order)
if max :
# set max
cond.set_max(max)
if skip :
# set skip
cond.set_skip(skip)
# execute
return self.search_cond(cond)
def search_simple (self, channel, query, count=None, offset=None) :
"""
Search for lines from the given channel for the given simple query
"""
# use search(), backwards
results = list(self.search(
# simplified phrase
options = hype.Condition.SIMPLE,
# specific channel
channel = channel,
# given phrase
phrase = query,
# order by timestamp
order = "@timestamp NUMD",
# count/offset
max = count,
skip = offset,
))
# reverse
return reversed(results)