Python实现大文件排序的方法

yipeiwu_com5年前Python基础

本文实例讲述了Python实现大文件排序的方法。分享给大家供大家参考。具体实现方法如下:

import gzip
import os
from multiprocessing import Process, Queue, Pipe, current_process, freeze_support
from datetime import datetime
def sort_worker(input,output):
 while True:
  lines = input.get().splitlines()
  element_set = {}
  for line in lines:
    if line.strip() == 'STOP':
      return
    try:
      element = line.split(' ')[0]
      if not element_set.get(element): element_set[element] = ''
    except:
      pass
  sorted_element = sorted(element_set)
  #print sorted_element
  output.put('\n'.join(sorted_element))
def write_worker(input, pre):
  os.system('mkdir %s'%pre)
  i = 0
  while True:
    content = input.get()
    if content.strip() == 'STOP':
      return
    write_sorted_bulk(content, '%s/%s'%(pre, i))
    i += 1
def write_sorted_bulk(content, filename):
  f = file(filename, 'w')
  f.write(content)
  f.close()
def split_sort_file(filename, num_sort = 3, buf_size = 65536*64*4):
  t = datetime.now()
  pre, ext = os.path.splitext(filename)
  if ext == '.gz':
    file_file = gzip.open(filename, 'rb')
  else:
    file_file = open(filename)
  bulk_queue = Queue(10)
  sorted_queue = Queue(10)
  NUM_SORT = num_sort
  sort_worker_pool = []
  for i in range(NUM_SORT):
    sort_worker_pool.append( Process(target=sort_worker, args=(bulk_queue, sorted_queue)) )
    sort_worker_pool[i].start()
  NUM_WRITE = 1
  write_worker_pool = []
  for i in range(NUM_WRITE):
    write_worker_pool.append( Process(target=write_worker, args=(sorted_queue, pre)) )
    write_worker_pool[i].start()
  buf = file_file.read(buf_size)
  sorted_count = 0
  while len(buf):
    end_line = buf.rfind('\n')
    #print buf[:end_line+1]
    bulk_queue.put(buf[:end_line+1])
    sorted_count += 1
    if end_line != -1:
      buf = buf[end_line+1:] + file_file.read(buf_size)
    else:
      buf = file_file.read(buf_size)
  for i in range(NUM_SORT):
    bulk_queue.put('STOP')
  for i in range(NUM_SORT):
    sort_worker_pool[i].join()
   
  for i in range(NUM_WRITE):
    sorted_queue.put('STOP')
  for i in range(NUM_WRITE):
    write_worker_pool[i].join()
  print 'elasped ', datetime.now() - t
  return sorted_count
from heapq import heappush, heappop
from datetime import datetime
from multiprocessing import Process, Queue, Pipe, current_process, freeze_support
import os
class file_heap:
  def __init__(self, dir, idx = 0, count = 1):
    files = os.listdir(dir)
    self.heap = []
    self.files = {}
    self.bulks = {}
    self.pre_element = None
    for i in range(len(files)):
      file = files[i]
      if hash(file) % count != idx: continue
      input = open(os.path.join(dir, file))
      self.files[i] = input
      self.bulks[i] = ''
      heappush(self.heap, (self.get_next_element_buffered(i), i))
  def get_next_element_buffered(self, i):
    if len(self.bulks[i]) < 256:
      if self.files[i] is not None:
        buf = self.files[i].read(65536)
        if buf:
          self.bulks[i] += buf
        else:
          self.files[i].close()
          self.files[i] = None
    end_line = self.bulks[i].find('\n')
    if end_line == -1:
      end_line = len(self.bulks[i])
    element = self.bulks[i][:end_line]
    self.bulks[i] = self.bulks[i][end_line+1:]
    return element
  def poppush_uniq(self):
    while True:
      element = self.poppush()
      if element is None:
        return None
      if element != self.pre_element:
        self.pre_element = element
        return element
  def poppush(self):
    try:
      element, index = heappop(self.heap)
    except IndexError:
      return None
    new_element = self.get_next_element_buffered(index)
    if new_element:
      heappush(self.heap, (new_element, index))
    return element
def heappoppush(dir, queue, idx = 0, count = 1):
  heap = file_heap(dir, idx, count)
  while True:
    d = heap.poppush_uniq()
    queue.put(d)
    if d is None: return
def heappoppush2(dir, queue, count = 1):
  heap = []
  procs = []
  queues = []
  pre_element = None
  for i in range(count):
    q = Queue(1024)
    q_buf = queue_buffer(q)
    queues.append(q_buf)
    p = Process(target=heappoppush, args=(dir, q_buf, i, count))
    procs.append(p)
    p.start()
  queues = tuple(queues)
  for i in range(count):
    heappush(heap, (queues[i].get(), i))
  while True:
    try:
      d, i= heappop(heap)
    except IndexError:
      queue.put(None)
      for p in procs:
        p.join()
      return
    else:
      if d is not None:
        heappush(heap,(queues[i].get(), i))
        if d != pre_element:
          pre_element = d
          queue.put(d)
def merge_file(dir):
  heap = file_heap( dir )
  os.system('rm -f '+dir+'.merge')
  fmerge = open(dir+'.merge', 'a')
  element = heap.poppush_uniq()
  fmerge.write(element+'\n')
  while element is not None:
    element = heap.poppush_uniq()
    fmerge.write(element+'\n')
class queue_buffer:
  def __init__(self, queue):
    self.q = queue
    self.rbuf = []
    self.wbuf = []
  def get(self):
    if len(self.rbuf) == 0:
      self.rbuf = self.q.get()
    r = self.rbuf[0]
    del self.rbuf[0]
    return r
  def put(self, d):
    self.wbuf.append(d)
    if d is None or len(self.wbuf) > 1024:
      self.q.put(self.wbuf)
      self.wbuf = []
def diff_file(file_old, file_new, file_diff, buf = 268435456):
  print 'buffer size', buf
  from file_split import split_sort_file
  os.system('rm -rf '+ os.path.splitext(file_old)[0] )
  os.system('rm -rf '+ os.path.splitext(file_new)[0] )
  t = datetime.now()
  split_sort_file(file_old,5,buf)
  split_sort_file(file_new,5,buf)
  print 'split elasped ', datetime.now() - t
  os.system('cat %s/* | wc -l'%os.path.splitext(file_old)[0])
  os.system('cat %s/* | wc -l'%os.path.splitext(file_new)[0])
  os.system('rm -f '+file_diff)
  t = datetime.now()
  zdiff = open(file_diff, 'a')
  old_q = Queue(1024)
  new_q = Queue(1024)
  old_queue = queue_buffer(old_q)
  new_queue = queue_buffer(new_q)
  h1 = Process(target=heappoppush2, args=(os.path.splitext(file_old)[0], old_queue, 3))
  h2 = Process(target=heappoppush2, args=(os.path.splitext(file_new)[0], new_queue, 3))
  h1.start(), h2.start()
  old = old_queue.get()
  new = new_queue.get()
  old_count, new_count = 0, 0
  while old is not None or new is not None:
    if old > new or old is None:
      zdiff.write('< '+new+'\n')
      new = new_queue.get()
      new_count +=1
    elif old < new or new is None:
      zdiff.write('> '+old+'\n')
      old = old_queue.get()
      old_count +=1
    else:
      old = old_queue.get()
      new = new_queue.get()
  print 'new_count:', new_count
  print 'old_count:', old_count
  print 'diff elasped ', datetime.now() - t
  h1.join(), h2.join()

希望本文所述对大家的Python程序设计有所帮助。

相关文章

python中yaml配置文件模块的使用详解

简述 和GNU一样,YAML是一个递归着说“不”的名字。不同的是,GNU对UNIX说不,YAML说不的对象是XML。 YAML不是XML。 为什么不是XML呢?因为: YAML...

在Python的一段程序中如何使用多次事件循环详解

背景 本文主要给大家介绍了关于在Python一段程序中使用多次事件循环的相关内容,我们在Python异步程序编写中经常要用到如下的结构 import asyncio async d...

python3.4控制用户输入与输出的方法

一、输入 1.函数格式:input() 2.函数功能:接受一个标准输入数据,返回string类型。ctrl+z结束输入。 3.实例: 默认input():等待一个任意字符的输入 s...

获取Pytorch中间某一层权重或者特征的例子

问题:训练好的网络模型想知道中间某一层的权重或者看看中间某一层的特征,如何处理呢? 1、获取某一层权重,并保存到excel中; 以resnet18为例说明: import t...

Python微信企业号开发之回调模式接收微信端客户端发送消息及被动返回消息示例

本文实例讲述了Python微信企业号开发之回调模式接收微信端客户端发送消息及被动返回消息。分享给大家供大家参考,具体如下: 说明:此代码用于接收手机微信端发送的消息 #-*- cod...