|
...
|
...
|
@@ -2,105 +2,69 @@ from userio import *
|
|
2
|
2
|
import requests
|
|
3
|
3
|
import re
|
|
4
|
4
|
import newsParser
|
|
|
5
|
+from requests_html import HTML
|
|
|
6
|
+from requests_html import HTMLSession
|
|
5
|
7
|
|
|
6
|
|
-def localArticleTitle(content):
|
|
7
|
|
- articleElementBegin="<meta property=\"og:title\" content=\""
|
|
8
|
|
- articleElementEnd ="\"/>"
|
|
9
|
|
- indexElementBegin = content.index(articleElementBegin)
|
|
10
|
|
- indexElementEnd = content.index(articleElementEnd,indexElementBegin)
|
|
11
|
|
- return content[indexElementBegin+len(articleElementBegin):indexElementEnd]
|
|
12
|
|
-
|
|
13
|
|
-def localArticleDescription(content):
|
|
14
|
|
- articleElementBegin="<meta property=\"og:description\" content=\""
|
|
15
|
|
- articleElementEnd ="\"/>"
|
|
16
|
|
- indexElementBegin = content.index(articleElementBegin)
|
|
17
|
|
- indexElementEnd = content.index(articleElementEnd,indexElementBegin)
|
|
18
|
|
- return content[indexElementBegin+len(articleElementBegin):indexElementEnd]
|
|
19
|
|
- return ""
|
|
20
|
|
-
|
|
21
|
|
-def localArticleImage(content):
|
|
22
|
|
- articleElementBegin="<meta property=\"og:image\" content=\""
|
|
23
|
|
- articleElementEnd ="\"/>"
|
|
24
|
|
- indexElementBegin = content.index(articleElementBegin)
|
|
25
|
|
- indexElementEnd = content.index(articleElementEnd,indexElementBegin)
|
|
26
|
|
- return content[indexElementBegin+len(articleElementBegin):indexElementEnd]
|
|
27
|
|
- return ""
|
|
28
|
8
|
|
|
29
|
9
|
def article(url):
|
|
30
|
10
|
say("Article: "+url)
|
|
31
|
|
- r = requests.get(url, allow_redirects=True)
|
|
32
|
|
- r.encoding = r.apparent_encoding
|
|
33
|
|
- content = r.text
|
|
34
|
|
-
|
|
35
|
|
- articleStrImageUrl = localArticleImage(content)
|
|
36
|
|
- articleStrTitle = localArticleTitle(content)
|
|
37
|
|
- articleStrDescription = localArticleDescription(content)
|
|
38
|
|
- articleStrImageUrl = re.sub(r"https://www\.washingtonpost\.com/wp-apps/imrs\.php\?src=(.+)&.+", r"\g<1>", articleStrImageUrl)
|
|
39
|
|
-
|
|
40
|
|
- pageContent = ""
|
|
41
|
|
- pageContent += "<meta charset=\"utf-8\"/>"
|
|
42
|
|
- pageContent += "<meta property=\"og:type\" content=\"article\">\n"
|
|
43
|
|
- pageContent += "<meta property=\"og:title\" content=\""+articleStrTitle+"\">\n"
|
|
44
|
|
- pageContent += "<meta property=\"og:description\" content=\""+articleStrDescription+"\">\n"
|
|
45
|
|
- pageContent += "<meta property=\"og:url\" content=\""+url+"\">\n"
|
|
46
|
|
- pageContent += "<meta property=\"og:image\" content=\""+articleStrImageUrl+"\">\n"
|
|
47
|
|
- pageContent += "<meta property=\"og:image:type\" content=\"image/jpeg\">"
|
|
48
|
|
-
|
|
49
|
|
- articleCstBegin = "<article"
|
|
50
|
|
- articleCstEnd = "<div class=\"mt-md\">"
|
|
51
|
|
- articleCstEnd2 = "</article>"
|
|
52
|
|
- articleCstEnd3 = "<div class=\"flex mt-md\">"
|
|
53
|
|
- indexBegin = content.index(articleCstBegin)
|
|
54
|
|
- try:
|
|
55
|
|
- indexEnd = content.index(articleCstEnd)
|
|
56
|
|
- except:
|
|
57
|
|
- try:
|
|
58
|
|
- indexEnd = content.index(articleCstEnd2)
|
|
59
|
|
- except:
|
|
60
|
|
- indexEnd = content.index(articleCstEnd3)
|
|
61
|
|
- debug("indexBegin: "+str(indexBegin))
|
|
62
|
|
- debug("indexEnd : "+str(indexEnd))
|
|
63
|
|
- say("Title: "+articleStrTitle)
|
|
64
|
|
- say("Image: "+articleStrImageUrl)
|
|
65
|
|
-
|
|
|
11
|
+ session = HTMLSession()
|
|
|
12
|
+ response = session.get(url,timeout=20)
|
|
|
13
|
+ pageContent=""
|
|
|
14
|
+ article_only=""
|
|
|
15
|
+ with response as r:
|
|
|
16
|
+ articleStrTitle = r.html.xpath('//meta[@property="og:title"]/@content')[0]
|
|
|
17
|
+ articleStrDescription = r.html.xpath('//meta[@property="og:description"]/@content')[0]
|
|
|
18
|
+ articleStrImageUrl = r.html.xpath('//meta[@property="og:image"]/@content')[0]
|
|
|
19
|
+ articleStrAuthor = r.html.xpath('//div[@class="author_wrapper"]/@content')
|
|
|
20
|
+ article=r.html.find("article")[0]
|
|
|
21
|
+ article_only+=article.html
|
|
|
22
|
+ lenBefore=len(article_only)
|
|
|
23
|
+ say("LengthBefore: "+str(lenBefore))
|
|
|
24
|
+ pageContent += "<meta property=\"og:type\" content=\"article\" />\n"
|
|
|
25
|
+ pageContent += "<meta property=\"og:title\" content=\""+articleStrTitle+"\" />\n"
|
|
|
26
|
+ pageContent += "<meta property=\"og:description\" content=\""+articleStrDescription+"\" />\n"
|
|
|
27
|
+ pageContent += "<meta property=\"og:url\" content=\""+url+"\" />\n"
|
|
|
28
|
+ pageContent += "<meta property=\"og:image\" content=\""+articleStrImageUrl+"\" />\n"
|
|
|
29
|
+ pageContent += "<meta property=\"og:image:type\" content=\"image/jpeg\" />\n"
|
|
|
30
|
+ #pageContent += "<meta name=\"author\" content=\""+articleStrAuthor+"\" />\n"
|
|
66
|
31
|
|
|
67
|
|
- article_only = "<h2>"+articleStrTitle+"</h2>"
|
|
68
|
|
- article_only += "<img src=\""+articleStrImageUrl+"\">"
|
|
69
|
|
- article_only += "<em>"+articleStrDescription+"</em>"
|
|
70
|
|
-
|
|
71
|
|
- with open("titi.html", "w") as f2:
|
|
72
|
|
- f2.write(content[indexBegin:indexEnd])
|
|
73
|
|
- f2.close
|
|
74
|
|
- article_only += content[indexBegin:indexEnd]
|
|
75
|
32
|
article_only = re.sub(r"<amp-img", '<img', article_only)
|
|
76
|
33
|
article_only = re.sub(r"</amp-img>", '', article_only)
|
|
77
|
34
|
article_only = re.sub(r"<h2", '<h3', article_only)
|
|
78
|
35
|
article_only = re.sub(r"</h2>", '</h3>', article_only)
|
|
79
|
36
|
article_only = re.sub(r"<h1", '<h2', article_only)
|
|
80
|
37
|
article_only = re.sub(r"</h1>", '</h2>', article_only)
|
|
81
|
|
- # ~ article_only = re.sub(r"<div data-sc-v=\"4.23.4\" data-sc-c=\"placeholder\">Advertisement</div>", '</h2>', article_only)
|
|
82
|
|
- article_only = re.sub(r"<div data-sc-v=\"4\.24\.3\" data-sc-c=\"placeholder\">Advertisement</div>", '', article_only)
|
|
83
|
|
- #article_only = re.sub(r"<div class=\"dib bg-white pl-xs pr-xs font-sans-serif light font-xxxxs lh-md gray-dark\" data-sc-v=\"4\.24\.3\" data-sc-c=\"adslot\">Story continues below advertisement</div>", '', article_only)
|
|
84
|
|
- article_only = re.sub(r"<div class=\"dib flex divider.+?data-sc-c=\"adslot\">Story continues below advertisement</div>","", article_only)
|
|
85
|
|
-
|
|
86
|
|
- article_only = re.sub(r"style=\"width:300px;height:250px\"", 'style=\"width:1px;height:1px\"', article_only)
|
|
87
|
|
- article_only = re.sub(r"style=\"width:120px;height:32px\"", 'style=\"width:1px;height:1px\"', article_only)
|
|
88
|
|
- article_only = re.sub(r"style=\"width:136px;height:20px\"", 'style=\"width:1px;height:1px\"', article_only)
|
|
89
|
|
- article_only = re.sub(r"style=\"width:300px;height:600px\"", 'style=\"width:1px;height:1px\"', article_only)
|
|
90
|
|
- article_only = re.sub(r"style=\"min-height:250px\"", 'style=\"min-height:1px\"', article_only)
|
|
91
|
|
- article_only = re.sub(r"style=\"min-height:298px\"", 'style=\"min-height:1px\"', article_only)
|
|
92
|
|
- article_only = re.sub(r"style=\"min-height:600px\"", 'style=\"min-height:1px\"', article_only)
|
|
93
|
|
- article_only = re.sub(r"class=\"center absolute w-100\" style=\"top:-12px\"", '', article_only)
|
|
94
|
|
- article_only = re.sub(r"class=\"center absolute w-100\" style=\"top:-12px\"", '', article_only)
|
|
95
|
|
- article_only = re.sub(r"<div data-qa=\"drop-cap-letter\">", '<div>', article_only)
|
|
96
|
|
- article_only = re.sub(r"filter:blur\(10px\);", '', article_only)
|
|
97
|
|
- article_only = re.sub(r"<div class=\"bg-pattern-1\".+?>", '<div>', article_only)
|
|
98
|
|
- article_only = re.sub(r"<div class=\"bg-pattern-2\".+?>", '<div>', article_only)
|
|
99
|
|
- article_only = re.sub(r"<img class=\"dn canvas-foreground\" src=\".+?\"/>", '', article_only)
|
|
100
|
|
- article_only = re.sub(r"<div class=\"subhead .+?>", '<div>', article_only)
|
|
101
|
|
- #article_only = re.sub(r"<canvas id=\"artboard\" style=\".+\">", '<canvas>', article_only)
|
|
102
|
|
- #article_only = re.sub(r"", '', article_only)
|
|
103
|
|
- article_only = article_only.replace("><", ">\n<")
|
|
|
38
|
+ article_only = re.sub(r'<script(.+?)</script>','',article_only,flags=re.M|re.S)
|
|
|
39
|
+ article_only = re.sub(r'<script(.+?)/>','',article_only)
|
|
|
40
|
+ article_only = re.sub(r'<button class="wpds(.+?)</button>','',article_only,flags=re.M|re.S)
|
|
|
41
|
+ article_only = re.sub(r'<div>Advertisement</div>','',article_only)
|
|
|
42
|
+ article_only = re.sub(r'<div class="flex flex-column(.+?)</div>','',article_only,flags=re.M|re.S)
|
|
|
43
|
+ article_only = re.sub(r'<div data-testid="placeholder-box" class="w-100 h-100 absolute flex flex-column justify-center border-box bg-offwhite" style="width:300px;height:250px">','<div>',article_only)
|
|
|
44
|
+ article_only = re.sub(r'<div aria-hidden="true" class="hide-for-print relative flex justify-center content-box items-center b bh mb-md mt-sm pt-sm pb-sm" style="min-height:250px;border-top-color:;border-bottom-color:">','<div>',article_only)
|
|
|
45
|
+ article_only = re.sub(r'<div aria-hidden="true" class="hide-for-print relative flex justify-center content-box items-center b bh mb-md mt-none pt-lg pb-lg" style="min-height:250px;border-top-color:;border-bottom-color:">','<div>',article_only)
|
|
|
46
|
+ article_only = re.sub(r'<button class="inline-flex(.+?)</button>','',article_only,flags=re.M|re.S)
|
|
|
47
|
+ article_only = re.sub(r'<div class="cb dn db-ns" data-qa="article-body-ad">','<div>',article_only)
|
|
|
48
|
+ article_only = re.sub(r'<svg (.+?)</svg>','',article_only,flags=re.M|re.S)
|
|
|
49
|
+ article_only = re.sub(r'<div class="center absolute w-100 border-box" style="top:">','<div>',article_only)
|
|
|
50
|
+ article_only = re.sub(r'<p data-qa="drop-cap-letter" data-el="text" class="font-copy font--article-body gray-darkest ma-0 pb-md">','<p>',article_only)
|
|
|
51
|
+ article_only = re.sub(r'<div class="article-body" data-qa="article-body">','<div>',article_only)
|
|
|
52
|
+ article_only = re.sub(r'<div class="cb db dn-ns" data-qa="article-body-ad">','<div>',article_only)
|
|
|
53
|
+ #article_only = re.sub(r'<img style="(.+?)" (.+?)>','<img src="'+articleStrImageUrl+'">',article_only)
|
|
|
54
|
+ article_only = re.sub(r'<div style="filter:blur(.+?)" class="w-100 mw-100 h-auto" width="600" height="(.+?)">','<div>',article_only)
|
|
|
55
|
+ article_only = re.sub(r'<div style="min-height:358px"/>','<div>',article_only)
|
|
|
56
|
+ #article_only = re.sub(r'','',article_only)
|
|
|
57
|
+ article_only = re.sub(r"href=\"/",'href="https://xxxxx/',article_only)
|
|
|
58
|
+ article_only = re.sub(r"src=\"/",'src="https://xxxxx/',article_only)
|
|
|
59
|
+ article_only = re.sub(r"^$",'',article_only)
|
|
|
60
|
+ article_only = re.sub(r'^\s*$', '',article_only,flags=re.M|re.S)
|
|
|
61
|
+ article_only = re.sub(r"><",'>\n<',article_only)
|
|
104
|
62
|
|
|
105
|
|
- pageContent += "<article>"+article_only+"</article>"
|
|
|
63
|
+ #pageContent += "\n"+article_only+"\n"
|
|
|
64
|
+ pageContent += "<article>\n"+article_only+"\n</article>\n"
|
|
|
65
|
+ lenAfter=len(article_only)
|
|
|
66
|
+ lenGain=float(10000-int(float(100*lenAfter/lenBefore)*100))/100
|
|
|
67
|
+ say("LengthAfter : "+str(lenAfter))
|
|
|
68
|
+ say("Gain : "+str(lenGain)+"%")
|
|
106
|
69
|
return pageContent
|
|
|
70
|
+
|