From e07784559a7c34e9797fececa09777f36e5a7428 Mon Sep 17 00:00:00 2001
From: yinjimmy <jimmy.yin5@gmail.com>
Date: Wed, 23 May 2018 21:35:28 +0800
Subject: [PATCH] fix url join and png file checking.

---
 aiss-spider/download_pictures.py | 16 ++++++++++++++--
 aiss-spider/run.py               |  5 ++++-
 2 files changed, 18 insertions(+), 3 deletions(-)

diff --git a/aiss-spider/download_pictures.py b/aiss-spider/download_pictures.py
index 522e781..662cf17 100644
--- a/aiss-spider/download_pictures.py
+++ b/aiss-spider/download_pictures.py
@@ -3,6 +3,7 @@
 import json
 import requests
 import time
+from urlparse import urljoin
 
 
 def get_info():
@@ -24,7 +25,8 @@ def get_info_imgs(info,host='http://aiss-1254466972.costj.myqcloud.com/picture')
         issue = item["issue"]
         pictureCount = item["pictureCount"]
         for pic_idx in range(pictureCount):
-            url = "%s/%s/%s/%s.jpg" % (host, catalog, issue, pic_idx)
+            url = "%s/%s/%s.jpg" % (catalog, issue, pic_idx)
+            url = urljoin(host, url)
             directory = os.path.join("data", name, "%s-%s" % (issue, nickname))
             filepath = os.path.join(directory, "%s.jpg" % pic_idx)
             # 每张图片一组，包含 图片url，所在目录，存储路径
@@ -45,12 +47,22 @@ def setup_download_dir(directory):
 
 from multiprocessing import Process, Queue, Pool
 
+def check_png(filepath):
+    from PIL import Image
+    try:
+        Image.open(filepath)
+        return True
+    except IOError:
+        # filename not an image file
+        return False
+    return False
 
 def download_one(img):
     """ 下载一张图片 """
     url, directory, filepath = img
     # 如果文件已经存在，放弃下载
-    if os.path.exists(filepath):
+    # if os.path.exists(filepath):
+    if check_png(filepath):
         print('exists:', filepath)
         return
 
diff --git a/aiss-spider/run.py b/aiss-spider/run.py
index 3c687bb..cd28b05 100644
--- a/aiss-spider/run.py
+++ b/aiss-spider/run.py
@@ -10,6 +10,9 @@
     info = get_info()
     # 获取每张图片的url，存储文件夹，本地文件名
     imgs = get_info_imgs(info,host=fetch_picture_url_header())
-    # test imgs = imgs[-2:]
+
+    # test
+    # imgs = imgs[-11:]
+    #
     # 以10个进程并发下载图片
     download(imgs, processes=10)