mirror of
https://github.com/ail-project/ail-framework.git
synced 2024-11-27 00:07:16 +00:00
359 lines
11 KiB
Python
Executable file
359 lines
11 KiB
Python
Executable file
#!/usr/bin/python2.7
|
|
|
|
"""
|
|
The ``Paste Class``
|
|
===================
|
|
|
|
Use it to create an object from an existing paste or other random file.
|
|
|
|
Conditions to fulfill to be able to use this class correctly:
|
|
-------------------------------------------------------------
|
|
|
|
1/ The paste need to be saved on disk somewhere (have an accessible path)
|
|
2/ The paste need to be gziped.
|
|
3/ The filepath need to look like something like this:
|
|
/directory/source/year/month/day/paste.gz
|
|
|
|
"""
|
|
|
|
import os
|
|
import magic
|
|
import gzip
|
|
import redis
|
|
import operator
|
|
import string
|
|
import re
|
|
import json
|
|
import ConfigParser
|
|
import cStringIO
|
|
from Date import Date
|
|
from Hash import Hash
|
|
|
|
from langid.langid import LanguageIdentifier, model
|
|
|
|
from nltk.tokenize import RegexpTokenizer
|
|
from textblob import TextBlob
|
|
|
|
clean = lambda dirty: ''.join(filter(string.printable.__contains__, dirty))
|
|
"""It filters out non-printable characters from the string it receives."""
|
|
|
|
|
|
class Paste(object):
|
|
"""
|
|
This class representing a Paste as an object.
|
|
When created, the object will have by default some "main attributes"
|
|
such as the size or the date of the paste already calculated, whereas other
|
|
attributes are not set and need to be "asked to be calculated" by their
|
|
methods.
|
|
It was design like this because some attributes take time to be calculated
|
|
such as the langage or the duplicate...
|
|
|
|
:Example:
|
|
|
|
PST = Paste("/home/2013/ZEeGaez5.gz")
|
|
|
|
"""
|
|
|
|
def __init__(self, p_path):
|
|
|
|
configfile = os.path.join(os.environ['AIL_BIN'], 'packages/config.cfg')
|
|
if not os.path.exists(configfile):
|
|
raise Exception('Unable to find the configuration file. \
|
|
Did you set environment variables? \
|
|
Or activate the virtualenv.')
|
|
|
|
cfg = ConfigParser.ConfigParser()
|
|
cfg.read(configfile)
|
|
self.cache = redis.StrictRedis(
|
|
host=cfg.get("Redis_Queues", "host"),
|
|
port=cfg.getint("Redis_Queues", "port"),
|
|
db=cfg.getint("Redis_Queues", "db"))
|
|
self.store = redis.StrictRedis(
|
|
host=cfg.get("Redis_Data_Merging", "host"),
|
|
port=cfg.getint("Redis_Data_Merging", "port"),
|
|
db=cfg.getint("Redis_Data_Merging", "db"))
|
|
|
|
self.p_path = p_path
|
|
self.p_name = os.path.basename(self.p_path)
|
|
self.p_size = round(os.path.getsize(self.p_path)/1024.0, 2)
|
|
self.p_mime = magic.from_buffer(self.get_p_content(), mime=True)
|
|
|
|
# Assuming that the paste will alway be in a day folder which is itself
|
|
# in a month folder which is itself in a year folder.
|
|
# /year/month/day/paste.gz
|
|
var = self.p_path.split('/')
|
|
self.p_date = Date(var[-4], var[-3], var[-2])
|
|
self.p_source = var[-5]
|
|
|
|
self.p_encoding = None
|
|
self.p_hash_kind = {}
|
|
self.p_hash = {}
|
|
self.p_langage = None
|
|
self.p_nb_lines = None
|
|
self.p_max_length_line = None
|
|
self.p_duplicate = None
|
|
|
|
def get_p_content(self):
|
|
"""
|
|
Returning the content of the Paste
|
|
|
|
:Example:
|
|
|
|
PST.get_p_content()
|
|
|
|
"""
|
|
|
|
paste = self.cache.get(self.p_path)
|
|
if paste is None:
|
|
try:
|
|
with gzip.open(self.p_path, 'rb') as f:
|
|
paste = f.read()
|
|
self.cache.set(self.p_path, paste)
|
|
self.cache.expire(self.p_path, 300)
|
|
except:
|
|
return ''
|
|
pass
|
|
return paste
|
|
|
|
def get_p_content_as_file(self):
|
|
return cStringIO.StringIO(self.get_p_content())
|
|
|
|
def get_lines_info(self):
|
|
"""
|
|
Returning and setting the number of lines and the maximum lenght of the
|
|
lines of the paste.
|
|
|
|
:return: tuple (#lines, max_length_line)
|
|
|
|
:Example: PST.get_lines_info()
|
|
|
|
"""
|
|
if self.p_nb_lines is None or self.p_max_length_line is None:
|
|
max_length_line = 0
|
|
f = self.get_p_content_as_file()
|
|
line_id = 0
|
|
for line_id, line in enumerate(f):
|
|
length = len(line)
|
|
if length >= max_length_line:
|
|
max_length_line = length
|
|
f.close()
|
|
self.p_nb_lines = line_id
|
|
self.p_max_length_line = max_length_line
|
|
return (self.p_nb_lines, self.p_max_length_line)
|
|
|
|
def _get_p_encoding(self):
|
|
"""
|
|
Setting the encoding of the paste.
|
|
|
|
:Example: PST._set_p_encoding()
|
|
|
|
"""
|
|
return self.p_mime
|
|
|
|
def _set_p_hash_kind(self, hashkind):
|
|
"""
|
|
Setting the hash (as an object) used for futur operation on it.
|
|
|
|
:Example: PST._set_p_hash_kind("md5")
|
|
|
|
.. seealso:: Hash.py Object to get the available hashs.
|
|
|
|
"""
|
|
self.p_hash_kind[hashkind] = (Hash(hashkind))
|
|
|
|
def _get_p_hash(self):
|
|
"""
|
|
Setting the hash of the paste as a kind of "uniq" identificator
|
|
|
|
:return: a dictionnary of hash string (md5, sha1....)
|
|
|
|
:Example: PST._get_p_hash()
|
|
|
|
.. note:: You need first to "declare which kind of hash you want to use
|
|
before using this function
|
|
.. seealso:: _set_p_hash_kind("md5")
|
|
|
|
"""
|
|
for hash_name, the_hash in self.p_hash_kind.iteritems():
|
|
self.p_hash[hash_name] = the_hash.Calculate(self.get_p_content())
|
|
return self.p_hash
|
|
|
|
def _get_p_language(self):
|
|
"""
|
|
Returning and setting the language of the paste (guessing)
|
|
|
|
:Example: PST._get_p_language()
|
|
|
|
..note:: The language returned is purely guessing and may not be accurate
|
|
if the paste doesn't contain any human dictionnary words
|
|
..seealso: git@github.com:saffsd/langid.py.git
|
|
|
|
FIXME: This procedure is using more than 20% of CPU
|
|
|
|
"""
|
|
identifier = LanguageIdentifier.from_modelstring(model, norm_probs=True)
|
|
return identifier.classify(self.get_p_content())
|
|
|
|
def _get_p_hash_kind(self):
|
|
return self.p_hash_kind
|
|
|
|
def _get_p_date(self):
|
|
return self.p_date
|
|
|
|
def _get_p_size(self):
|
|
return self.p_size
|
|
|
|
def is_duplicate(self, obj, min=1, percent=50, start=1, jump=10):
|
|
"""
|
|
Returning the percent of similarity with another paste.
|
|
( Using the previous hashing method )
|
|
|
|
:param obj: (Paste) The paste to compare with
|
|
:param min: -- (int) Minimum line length to be hashed.
|
|
:param percent: -- (int)
|
|
:param start: -- (int) Number the line where to start.
|
|
:param jump: -- (int) Granularity of the hashing 0 or 1 means no jumps
|
|
(Maximum Granularity)
|
|
|
|
:return: (tuple) (bool, percent)
|
|
|
|
:Example:
|
|
PST.is_duplicate(PST)
|
|
|
|
>>> return (True, 100.0)
|
|
|
|
..seealso: _get_hash_lines()
|
|
|
|
"""
|
|
|
|
set1 = self._get_hash_lines(min, start, jump)
|
|
set2 = obj._get_hash_lines(min, start, jump)
|
|
|
|
inter = set.intersection(set1, set2)
|
|
|
|
numerator = len(inter)
|
|
denominator = float((len(set1) + len(set2)) / 2)
|
|
|
|
try:
|
|
var = round((numerator / denominator)*100, 2)
|
|
except ZeroDivisionError:
|
|
var = 0.0
|
|
|
|
if var >= percent:
|
|
return True, var
|
|
else:
|
|
return False, var
|
|
|
|
def _get_p_duplicate(self):
|
|
self.p_duplicate = self.store.hget(self.p_path, "p_duplicate")
|
|
return self.p_duplicate if self.p_duplicate is not None else []
|
|
|
|
def save_all_attributes_redis(self, key=None):
|
|
"""
|
|
Saving all the attributes in a "Redis-like" Database (Redis, LevelDB)
|
|
|
|
:param r_serv: -- Connexion to the Database.
|
|
:param key: -- Key of an additionnal set.
|
|
|
|
Example:
|
|
import redis
|
|
|
|
r_serv = redis.StrictRedis(host = 127.0.0.1, port = 6739, db = 0)
|
|
|
|
PST = Paste("/home/Zkopkmlk.gz")
|
|
PST.save_all_attributes_redis(r_serv)
|
|
|
|
"""
|
|
# LevelDB Compatibility
|
|
p = self.store.pipeline(False)
|
|
p.hset(self.p_path, "p_name", self.p_name)
|
|
p.hset(self.p_path, "p_size", self.p_size)
|
|
p.hset(self.p_path, "p_mime", self.p_mime)
|
|
# p.hset(self.p_path, "p_encoding", self.p_encoding)
|
|
p.hset(self.p_path, "p_date", self._get_p_date())
|
|
p.hset(self.p_path, "p_hash_kind", self._get_p_hash_kind())
|
|
p.hset(self.p_path, "p_hash", self.p_hash)
|
|
# p.hset(self.p_path, "p_langage", self.p_langage)
|
|
# p.hset(self.p_path, "p_nb_lines", self.p_nb_lines)
|
|
# p.hset(self.p_path, "p_max_length_line", self.p_max_length_line)
|
|
# p.hset(self.p_path, "p_categories", self.p_categories)
|
|
p.hset(self.p_path, "p_source", self.p_source)
|
|
if key is not None:
|
|
p.sadd(key, self.p_path)
|
|
else:
|
|
pass
|
|
p.execute()
|
|
|
|
def save_attribute_redis(self, attr_name, value):
|
|
"""
|
|
Save an attribute as a field
|
|
"""
|
|
if type(value) == set:
|
|
self.store.hset(self.p_path, attr_name, json.dumps(list(value)))
|
|
else:
|
|
self.store.hset(self.p_path, attr_name, json.dumps(value))
|
|
|
|
def _get_from_redis(self, r_serv):
|
|
ans = {}
|
|
for hash_name, the_hash in self.p_hash:
|
|
ans[hash_name] = r_serv.hgetall(the_hash)
|
|
return ans
|
|
|
|
def _get_top_words(self, sort=False):
|
|
"""
|
|
Tokenising method: Returning a sorted list or a set of paste's words
|
|
|
|
:param sort: Selecting the output: sorted list or a set. (set by default)
|
|
|
|
:return: set or sorted list of tuple [(word, occurency)...]
|
|
|
|
:Example: PST._get_top_words(False)
|
|
|
|
"""
|
|
words = {}
|
|
tokenizer = RegexpTokenizer('[\&\~\:\;\,\.\(\)\{\}\|\[\]\\\\/\-/\=\'\"\%\$\?\@\+\#\_\^\<\>\!\*\n\r\t\s]+',
|
|
gaps=True, discard_empty=True)
|
|
|
|
blob = TextBlob(clean(self.get_p_content()), tokenizer=tokenizer)
|
|
|
|
for word in blob.tokens:
|
|
if word in words.keys():
|
|
num = words[word]
|
|
else:
|
|
num = 0
|
|
words[word] = num + 1
|
|
if sort:
|
|
var = sorted(words.iteritems(), key=operator.itemgetter(1), reverse=True)
|
|
else:
|
|
var = words
|
|
|
|
return var
|
|
|
|
def _get_word(self, word):
|
|
"""
|
|
Returning a specific word and his occurence if present in the paste
|
|
|
|
:param word: (str) The word
|
|
|
|
:return: (tuple) ("foo", 1337)
|
|
|
|
"""
|
|
return [item for item in self._get_top_words() if item[0] == word]
|
|
|
|
def get_regex(self, regex):
|
|
"""
|
|
Returning matches with the regex given as an argument.
|
|
|
|
:param regex: -- (str) a regex
|
|
|
|
:return: (list)
|
|
|
|
:Example: PST.get_regex("4[0-9]{12}(?:[0-9]{3})?")
|
|
|
|
|
|
"""
|
|
matchs = []
|
|
for match in re.findall(regex, self.get_p_content()):
|
|
if match != '' and len(match) < 100:
|
|
matchs.append(match)
|
|
return matchs
|