# -*- coding: utf-8 -*-
# @Time  : 2019/06/10 18:56
# @Author : TuanZhangSama
import os
import xml.etree.ElementTree as ET
from multiprocessing import Pool,freeze_support,cpu_count
import imghdr
import logging
def get_all_xml_path(xml_dir:str,filter=['.xml']):
  #遍歷文件夾下所有xml
  result=[]
  #maindir是當(dāng)前搜索的目錄 subdir是當(dāng)前目錄下的文件夾名 file是目錄下文件名
  for maindir,subdir,file_name_list in os.walk(xml_dir):
    for filename in file_name_list:
      ext=os.path.splitext(filename)[1]#返回?cái)U(kuò)展名
      if ext in filter:
        result.append(os.path.join(maindir,filename))
  return result
def analysis_xml(xml_path:str):
  tree=ET.parse(xml_path)
  root=tree.getroot()
  result_dict={}
  for obj in root.findall('object'):
    obj_name = obj.find('name').text
    obj_num=result_dict.get(obj_name,0)+1
    result_dict[obj_name]=obj_num
  if imghdr.what(xml_path.replace('.xml','.jpg')) != 'jpeg':
    print(xml_path.replace('.xml','.jpg'),'is worng')
    # logging.info(xml_path.replace('.xml','.jpg'))
  if is_valid_jpg(xml_path.replace('.xml','.jpg')):
    pass
  return result_dict
def analysis_xmls_batch(xmls_path_list:list):
  result_list=[]
  for i in xmls_path_list:
    result_list.append(analysis_xml(i))
  return result_list
def collect_result(result_list:list):
  all_result_dict={}
  for result_dict in result_list:
    for key,values in result_dict.items():
      obj_num=all_result_dict.get(key,0)+values
      all_result_dict[key]=obj_num
  return all_result_dict
def main(xml_dir:str,result_save_path:str =None):
  r'''根據(jù)xml文件統(tǒng)計(jì)所有樣本的數(shù)目.對(duì)于文件不完整的圖片和有xml但無圖片的樣本,直接進(jìn)行刪除.默認(rèn)跑滿所有的cpu核心
  Parameters
  ----------
  xml_dir : str
    xml所在的文件夾.用的遞歸形式,因此只需保證xml在此目錄的子目錄下即可.對(duì)應(yīng)的圖片和其xml要在同一目錄
  result_save_path : str
    分析結(jié)果的日志保存路徑.默認(rèn) None 無日志
  '''
  if result_save_path is not None:
    assert isinstance(result_save_path,str),'{} is illegal path'.format(result_save_path)
  else:
    logging.basicConfig(filename=result_save_path,filemode='w',level=logging.INFO)
  freeze_support()#windows 上用
  xmls_path=get_all_xml_path(xml_dir)
  worker_num=cpu_count()
  print('your CPU num is',cpu_count())
  length=float(len(xmls_path))/float(worker_num)
  #計(jì)算下標(biāo)，盡可能均勻地劃分輸入文件的列表
  indices=[int(round(i*length)) for i in range(worker_num+1)]
  #生成每個(gè)進(jìn)程要處理的子文件列表
  sublists=[xmls_path[indices[i]:indices[i+1]] for i in range(worker_num)]
  pool=Pool(processes=worker_num)
  all_process_result_list=[]
  for i in range(worker_num):
    all_process_result_list.append(pool.apply_async(analysis_xmls_batch,args=(sublists[i],)))
  pool.close()
  pool.join()
  print('analysis done!')
  _temp_list=[]
  for i in all_process_result_list:
    _temp_list=_temp_list+i.get()
  result=collect_result(_temp_list)
  logging.info(result)
  print(result)
def is_valid_jpg(jpg_file):
  """判斷JPG文件下載是否完整   """
  if not os.path.exists(jpg_file):
    print(jpg_file,'is not existes')
    os.remove(jpg_file.replace('.jpg','.xml'))
  with open(jpg_file, 'rb') as fr:
    fr.seek(-2, 2)
    if fr.read() == b'\xff\xd9':
      return True
    else:
      os.remove(jpg_file)
      os.remove(jpg_file.replace('.jpg','.xml'))
      print(jpg_file)
      logging.error(jpg_file,'is imperfect img')
      return False
if __name__=='__main__':
  test_dir='/home/chiebotgpuhq/Share/winshare/origin'
  save_path='/home/chiebotgpuhq/MyCode/python/pytorch/mmdetection-master/result.log'
  main(test_dir,save_path)