Python + pyspider某小说站的爬虫,入数据库,火车头发布,资源下载到本地,另可写爬虫!
- g; Y1 t) X+ {6 c) C, w0 W- #!/usr/bin/env python6 e( S- y# M$ [" R, ~' ^/ }
- # -*- encoding: utf-8 -*-
7 S6 p6 e8 }: X2 M+ A - # Created on 2019-05-05 21:43:118 N) \3 K6 W; J3 Q/ T) A0 ~/ ~9 a
- # Project: XiaoShuo
, R/ \' S0 `* T2 s - 6 \# r1 ]# \1 |2 C) \) [& V
- from pyspider.libs.base_handler import *5 k0 P1 B9 w1 x
- import pymysql
, e9 I; \/ H! e! L( r+ C - import random
d' r' x/ k; e - import datetime
2 T3 H5 Y/ x* o/ T" w2 s7 z) U - import urllib2,HTMLParser,re/ d; J5 b- F- Y$ ?3 r7 G
- import os# s6 i; @$ ~: E v
- import sys
& Y& w7 ]% S( f6 c( V - import re
, w4 y+ K! o# p3 f* u# m - import codecs
$ `& D& @4 h+ }3 y) J( O/ F3 C - import requests
- c7 j8 x) h& b) ?2 x# m- g' q - import json
: c O. }% W2 r, T& X# T' R- b/ z, r - 7 Y# P" T9 J' H8 C
- class Handler(BaseHandler):
% n; z! |: B9 N5 O6 b/ _$ R5 S - global Datos
* n- N+ R. N0 R$ z# X - global P_dir
. ]* G2 S: [; R# F% L - P_dir = '/Tools/Debug/' #采集时候图片保持到本地的路径
) y( X G/ \0 ~. n" |# G6 E - global Datos# [9 O B* s+ @. t
- Datos = {}
; |# g7 P' V/ G% ]8 M0 w2 @ - headers= {, q2 M5 ?0 ~# \/ o
- 'Accept':'text/html,application/xhtml+xml,application/xml;q=0.9,image/webp,*/*;q=0.8',
+ y, S) I/ J$ k& ? - 'Accept-Encoding':'gzip, deflate, sdch'," n% l2 K9 \ F _) R+ z! O3 c
- 'Accept-Language':'zh-CN,zh;q=0.8',
3 m0 X. x' G9 V {5 s6 N - 'Cache-Control':'max-age=0',% |/ [8 t! t% {6 v2 @+ c
- 'Connection':'keep-alive',
. a- @$ c9 q( X( H - 'User-Agent':'Mozilla/5.0 (Windows NT 6.1; WOW64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/41.0.2272.101 Safari/537.36'- {1 L9 X/ e+ @
- }0 m1 S K, _9 B& f7 _
- crawl_config = {
1 V* _! n" S; K% g2 I h - 'headers' : headers,
* C$ \9 {4 z' j9 K" ]( N - 'timeout' : 300* {) l. ^. [: M. Z+ ^9 {
- }! P5 ?! j" A3 r8 j* J# \, _
- def add_BookFile(self,Bookname, BookIDs, img, Locaimg, Book_Dates):
- U: B ?. X- z: Y( G" W - db = pymysql.connect(host="localhost",user="数据库用户名",password="密码",db="数据库名",charset="utf8")* i# L* ~& j' [1 |; \9 l
- try:2 D1 | P+ K E( {* x. C# \$ {
- cursor = db.cursor()
2 j( P! N- g" L9 q9 R4 N t - #注意此处字符串的占位符要加双引号"%s"; p; A5 x! A& }5 m- i1 h! _( I9 x
- sql = 'insert into BookFile(Bookname, BookID, img, Locaimg, Book_Date) values ("%s","%s","%s","%s","%s")' % (Bookname, BookIDs, img, Locaimg, Book_Dates);
) V9 j! o- V3 [+ u! b) ] - # print(sql)6 d2 b1 a7 ?% X! i9 G% ^
- cursor.execute(sql)$ P% u. ]! |& g# ]
- ! H) _& ^/ E0 i6 }
- #qid = cursor.lastrowid2 m( j/ w" b- q
- #print(qid)
) M0 B0 m! ^5 [1 C" D2 q& G8 U( L - ) d7 q: ]) d- b- e' ~. f. L
- db.commit()
; U) O+ f0 k; k - except Exception as err:
8 f! m5 D- N9 l- R0 M: M6 k1 v - print("Error %s for execute sql: %s" % (err, sql))
; r. E- o1 l" z D. S - db.rollback()
6 S' o2 r4 Y; C1 j, g6 I7 Q - def add_comment(self,Bookname, Booktitle, BookID, Titleid, Book_Date):- E6 D/ Y ]. d
- db = pymysql.connect(host="localhost",user="数据库用户名",password="密码",db="数据库名",charset="utf8")- C# L( Z! D* C, u, j
- try:
; h. D+ N* |" r: a" Y' t - cursor = db.cursor()% t4 A+ l! w- m- Y! S
- #注意此处字符串的占位符要加双引号"%s"
' d* e' D' `3 { - sql = 'insert into BookTitle(Bookname, Booktitle, BookID, Titleid, Book_Date) values ("%s","%s","%s","%s","%s")' % (Bookname, Booktitle, BookID, Titleid, Book_Date);9 k- K% Z4 @1 I+ v7 S5 M
- # print(sql)
, P; b7 p' B5 z4 V/ ?' u - cursor.execute(sql)9 \* G9 Z# T. |0 b0 p( r
-
' o& }; _5 Y* [. Z8 m2 ~ - #qid = cursor.lastrowid, d" X4 r7 V6 q
- #print(qid). P! l7 I& b" w) y8 q" C* w" G, \
-
; v% s) g) x. H) w - db.commit()7 }$ s* d# ~2 S
- except Exception as err:
6 l+ {" O: g8 ~" X* z1 h - print("Error %s for execute sql: %s" % (err, sql))' E. c9 A' T+ u) w- H& w3 q
- db.rollback()7 V) l( z5 ?1 P' V c R
- def add_question(self,Bookname,Cater_Name,Book_author,Book_Introduction,Book_Synopsis,Book_Palabras,Bookurl,Booktitle,BookID,BookConte,Titleid,abover,Book_Date):
- G+ m' ~/ M- c; M7 I2 ]9 A - db = pymysql.connect(host="localhost",user="数据库用户名",password="密码",db="数据库名",charset="utf8")
3 J* S% g9 q( g - try:
3 R3 z- V8 Q" l0 d( Z - cursor = db.cursor()% J; y/ X/ ?# ~1 u
- #注意此处字符串的占位符要加双引号"%s"% h N$ b% R+ c; J/ Y" O D
- sql = ' INSERT INTO BookConte (Bookname,Cater_Name,Book_author,Book_Introduction,Book_Synopsis,Book_Palabras,Bookurl,Booktitle,BookID,BookConte,Titleid,abover,Book_Date) VALUES("%s","%s","%s","%s","%s","%s","%s","%s","%s","%s","%s","%s","%s")' % (Bookname,Cater_Name,Book_author,Book_Introduction,Book_Synopsis,Book_Palabras,Bookurl,Booktitle,BookID,BookConte,Titleid,abover,Book_Date);
+ E- p# ]) n3 l( n+ O$ f9 J - print(sql)
" n5 F7 E# `$ G% E4 n - cursor.execute(sql)
' b" I4 n1 y: k' [7 h% K: ^4 a# Y - print(cursor.lastrowid)
1 x( L1 h0 O. _# I - db.commit()
7 A1 Y0 u( Q' K$ ]- d: S5 z - except Exception as err:
j) N8 i+ Z# \2 r: ?6 F - # except:+ i* ^+ d& `5 u" }
- # print('Failed')/ ]* a+ U# Q% b- X' x; ~$ h" s2 j& q
- print("Error %s for execute sql: %s" % (err, sql))
) t9 P$ F C4 G3 d- V$ { - db.rollback()
; c8 |' L8 e0 e/ Y# q' t -
5 E/ h* z# }) V5 T' N5 | - def add_locoy(self,Bookname,Cater_Name,Book_author,Book_Introduction,Book_Palabras,Book_img,Booktitle,BookConte,abover): , Z) d" |0 r! l$ `6 Q8 N4 D
- reload(sys)
8 ^8 E7 {$ P# f; c& ~ - sys.setdefaultencoding("gbk")6 a! x m2 q/ y( w6 K6 C: T
- locoy_url = 'http://www.******.net/locoy/?my=book' #697火车头发接口地址* C* ?. X2 H$ U+ Y
- locoy_data = {
$ y, G; e: ^9 i3 }+ _ - 'my_u':'用户名', #后台用户名6 ?( l* }! `' c4 u
- 'my_p':'密码', #后台密码
3 S" C1 }; T* W4 L2 w - 'subject_669977_net':Bookname.encode('gbk', 'ignore'),
& D& t" @) O0 h( P - 'caid':Cater_Name.encode('gbk', 'ignore'),! d1 t. m. e& z+ a! ^
- 'title_669977_net':Booktitle.encode('gbk', 'ignore'),
6 @& J4 X x3 D8 Z' S X9 } - 'article':BookConte.encode('gbk', 'ignore'),- z- Q' F- ^4 ~4 a4 B
- 'author':Book_author.encode('gbk', 'ignore'),
# n) V5 v2 M9 i6 |8 N - 'ready_1':Book_Palabras.encode('gbk', 'ignore'),+ i! J+ @4 X! n, N: `6 Y) h
- 'thumb':Book_img," R- F4 `! l& ?$ `( e
- 'content':Book_Introduction.encode('gbk', 'ignore'),
2 I6 T# z+ a2 m/ Z3 { l& {9 @ - 'abover':abover.encode('gbk', 'ignore')
/ ~* r, H( P8 C - }! Q- ?% L! `, r/ f5 F
- res = requests.post(locoy_url, data=locoy_data) [- F% d! z1 b4 z
- print res.text& ^, R3 _' E& o( l
- print res.content
; g! }8 A5 z. u5 [1 D - # print Dsd/ m$ f2 X7 A% A1 Y/ r: v
- return res
- H" `+ |/ ^; R) z -
7 }6 W5 V9 C ]4 f8 K! p - def __init__(self):6 j( S4 C7 _* X- ~6 f' \9 S$ o* ?
- self.base_url1 = 'https://www.****.cc/'4 i+ ~8 r. g" `- }. Q3 d
- self.base_url2 = '/'
6 h3 A7 N2 O; \/ g5 E- W - self.CaterId = []
M0 k% e1 }3 E/ ~ - self.CaterIds = ['xuanhuan', 'wuxia', 'yanqing', 'lishi', 'kehuan', 'kongbu', 'nvsheng']
7 v! z- u! d1 b - self.page_num = 1* Z' K: }0 g. M. h
- self.total_num = 200 0 E) G, O4 ^- Q( d7 o
- 3 J% b. ]' K9 w' p6 g! V
- @every(minutes=8 * 60)
+ Z4 y- ]0 N8 s - def on_start(self):" {, @, v/ h+ T8 L; D2 S" Q ?8 V
- global Cater_Name
( Y0 y: N, Z9 W4 O3 h' j2 ` - Cater_Name = []
0 L9 J; o- x$ z$ z* ] - while self.page_num <= self.total_num:
* g+ T. e, d: |$ S9 V$ { - for self.CaterId in self.CaterIds:/ V; v' N8 Y0 ^* U
- if self.CaterId == 'xuanhuan':
" K" J& ?- N Y5 O7 u4 R, L; o - Cater_Name = '玄幻'* C) M5 W) j, \9 D; U- O4 p& c7 \8 N
- if self.CaterId == 'wuxia':
. G# r g7 {! L6 c' N - Cater_Name = '武侠'
& Z2 }$ T" y0 Q3 T1 v - if self.CaterId == 'lishi':
$ P& z* D% f. A) Z5 `+ |& D - Cater_Name = '历史' 6 H/ I3 j$ f" y, K7 B& r
- if self.CaterId == 'yanqing':0 L F) B! C; e3 [) g% c
- Cater_Name = '都市' * g7 I0 u9 L7 i6 g2 f* d
- if self.CaterId == 'nvsheng':* \/ m# }/ }3 b! r. U; x
- Cater_Name = '都市'
7 @5 Y# W6 a. c* r4 p! h' N7 ` - if self.CaterId == 'kehuan':
! \ N, l4 H( F7 w% Y4 G - Cater_Name = '科幻'
' Q# R+ T. f$ O" f& r - if self.CaterId == 'kongbu':
8 d+ w. r! D( x# P1 N) W - Cater_Name = '游戏' " U0 k3 k# [8 b. i9 G
- print self.CaterId
7 X/ i- z! Z6 C% L+ @( V1 y - url = self.base_url1 + str(self.CaterId) + self.base_url2 + str(self.page_num) + "/"
( h$ Q3 k' m* n; [' f5 U4 Z7 ? - self.crawl(url, callback=self.list_Caterg,save=Cater_Name)) e- u! l" \# l8 q c! y
- self.page_num += 1
' X9 H- x6 F& P+ @ -
8 I* p* L. a. Y2 B1 ~, j! T e6 y - def list_Caterg(self, response):
8 s& V# W" S3 t2 q5 Y - Cater_Name = response.save
6 R" ~4 m4 h8 z' o- g" m k - for each in response.doc('.pic-list a[href^="http"]').items():( G% N+ u( A2 ?9 H( i
- self.crawl(each.attr.href, callback=self.list_Caterg_detail,save=Cater_Name)
4 f. O7 M. n% L# s& G4 u -
0 n( _2 A2 e- Y - def list_Caterg_detail(self, response):
1 @3 |: |! X" Z0 J$ ~& ]: D - Cater_Name = response.save) b* I/ {4 [$ q. l$ f8 R5 f
- # print Cater_Name; @5 G$ U* p7 t2 m! O2 U: P
- Bookname = response.doc('h1').text()5 O6 D7 n: L2 z0 I
- print Bookname
8 Z1 I2 o4 j: x: [ - Book_author = response.doc('.authorname > a').text(); I9 X+ i4 ?/ k
- # print Book_author
9 F% x' |2 H2 r) a6 m6 j0 V - Book_Introduction = response.doc('.book-intro > div').text()7 m; e( U. @& Y
- # print Book_Introduction
, G1 x$ X" m: o8 Y( O1 @ - Book_Synopsis = response.doc('b').eq(1).text()
4 D* M+ U% L5 d$ P$ W, y4 \# I - # print Book_Synopsis ^6 G2 c3 G( y+ t: w9 `
- Book_Palabras = response.doc('.booktitle p').text().split(' ')[1].split('|')[0]
1 S9 [* m) z% [! Z - # print Book_Palabras) p7 a& b& f+ g" X# f
- BookIDs = response.url.split("xiaoshuo/")[-1].split("/")[0] #小说ID
. e$ D a1 e6 E. t9 _" m - # print BookIDs0 b& D) |+ [$ ]+ m% D/ ~
- Book_Dates = str(datetime.datetime.now()) ( }% _& M- Y4 S) G3 R
- for imgs in response.doc('.bigpic > img[src^="http"]').items():
7 {0 t$ D5 m( }4 o( Q2 }! [6 m - img = imgs.attr.src
w% B, e0 J7 K7 ^( P P5 N - print img
. y. u' p$ \+ f - #小说封面下载
9 }5 W" Z" P; J; z s - extension = self.getExtension(img)
0 F$ a' ?3 `- ~3 t* G( } - name = self.getname(img), @0 P+ P4 N2 j' y# ?4 ?
- file_name = name + "." + extension
3 S# J5 w/ K/ V" H- V - imgDir = P_dir + name
) [ H3 B6 ^% i/ o. R! ~ - Locaimg = imgDir + "/" + file_name
0 U5 V' g* @5 Z - print Locaimg
9 Q2 D8 Y# l% _7 N: N X' Y - if(self.download(P_dir, imgDir, file_name, img)): #这2行可注译,图片下载到本地
7 V8 C5 V5 \9 P; [0 k& B M - print('attachment url is ' + img) #
5 {# E5 c2 I6 R5 `; T3 W% C$ L1 v - Datos = {
6 T1 R/ z4 D8 d" c4 q( I - "Cater_Name":Cater_Name,& ]& i; t7 x( g/ E8 z6 d! _
- "Book_author":Book_author,
; d$ K+ I9 {; `( w: m - "Book_Introduction":Book_Introduction,
( k0 _. V6 N1 c/ u* S - "Book_Synopsis":Book_Synopsis,4 X# I! w/ O6 U( T" U6 ^
- "Book_Palabras":Book_Palabras,8 e5 y& I! r0 W1 o: X
- "img":img,
' Q, j; o+ [% N" x# q - }
4 E9 ^6 M/ I; Q/ k, R- t* n - self.add_BookFile(Bookname, BookIDs, img, Locaimg, Book_Dates) #这行可注译,数据库发布接口,方便其他系统的发布8 U& f: O- T0 d
- for each in response.doc('div[class="bookbtn-txt"] a[class="catalogbtn"]').items():
( `. W) s7 O, X" F$ M1 j - self.crawl(each.attr.href, callback=self.index_page,save=Datos)$ t: l+ p8 m+ t6 h8 c. b9 ~0 Y4 O
-
, x) x1 R7 e! s" k# b - @config(age=8 * 60 * 60)
$ x/ C" ?5 r F8 d- e - def index_page(self, response):
8 E5 |4 W+ _' d5 q2 N9 x$ k - Datos = {
/ m5 `" @5 M0 f! Q - "Cater_Name":response.save['Cater_Name'],
- M5 R8 E7 c7 V) G9 h" { - "Book_author":response.save['Book_author'],7 b* J# `4 ~/ x9 f- J6 O! Y
- "Book_Introduction":response.save['Book_Introduction'],
5 V( r1 i) w N) ]$ q* M" m$ n- _ - "Book_Synopsis":response.save['Book_Synopsis'],7 j+ h( \" s% O# T3 @
- "Book_Palabras":response.save['Book_Palabras'],; `$ ]1 y# O8 q4 i* L( A/ h
- "img":response.save['img'],' K, s) j. d) s& |; O+ a
- }* k3 q! {( Z# q. ]$ A ?2 f4 r
- for each in response.doc('.chapter-list li:first-child a[href^="http"]').items():2 z! Y) F% ]$ f: m p
- # for each in response.doc('.chapter-list a[href^="http"]').items(): # a- p, K+ R4 N+ S: s& {6 t6 p
- self.crawl(each.attr.href, callback=self.detail_page,save=Datos)
9 L" a! l7 b0 K: z/ [- Q. A - @config(priority=2)# \* J9 Z. P# q
- @catch_status_code_error
) l- H; z8 Z( O: T3 r - def detail_page(self, response): 4 r. J, H9 H$ T& z
- NewRe1 = u'哈书'
! T+ A, y" T( N$ f - NewRe2 = u'huhjsd.CC'% V' ^6 z7 ^0 E. u: {2 U9 V. e
- NewRe3 = r'^\\n\\n'
( |9 j" b/ W8 c, I - NewRe5 = u'小说网'$ L/ L# a% [' U% s5 K- X1 u
- NewRe6 = u'fgdfgf'* h. j1 A8 ?! Q5 B5 o
- NewRe7 = u'fgfgf'9 F1 o" e6 `) x: l5 }; P6 D/ G7 x
- NewRe8 = u'ffhgf'
6 n0 F( y8 g9 C, t4 v2 R - NewRe4 = r'[\f\t\v+\.\{\(\)\}\!\/_,$%^*(+"\')]+|[+——()?【】“”!,。?、~@#¥%……&*()]+'3 q. `6 e8 W3 K6 h( X% S, p
- ReC1 = u'静思'
6 Z- c7 Z: A \% \9 N - ReC2 = u'aghgf.com'
1 F0 r' l# i& j - ReC3 = u'aghgfh.com' [# i6 j7 o/ M
- ReC4 = u''4 `; F4 a& k/ k. _3 N
- ReC5 = u'文学网': o! v& w. a: @( d% W
- ReC6 = r'<BR>'
, J! a, b7 m- T( j - Bookname = response.doc('.readlocation a').eq(2).text() #小说名称' [" s d: ]/ ?, x" e. z) A/ w7 z/ m
- print Bookname( Z# ]$ Q$ J: F* I& s( H
- Cater_Name = response.save['Cater_Name'] # 小说分类
( d0 L8 K/ s/ V. Y6 R - Book_author = response.save['Book_author'] #小说作者* Z' I1 c% Q3 \+ I% W5 d
- Book_Introduction1 = response.save['Book_Introduction'] #小说简介
0 g- }, M2 I! P0 [5 w - Book_Synopsis = response.save['Book_Synopsis'] #最近更新$ c2 p( N3 a; k A" ^/ b
- Book_Palabras = response.save['Book_Palabras'] #小说字数
0 O& d8 a6 [' Y - Bookurl = response.url #小说网址1 T. |6 t4 ?9 V" y Y
- Booktitle = response.doc('.article-title').text() #章节名称
5 T4 J. p8 G8 i! d - BookID = response.doc('.readset-r span').text() #小说ID" ~# y$ b b* |' F* L# ]! `8 G
- BookConte1 = response.doc('.article-con').text() #小说章节内容
$ f9 b' a e# B! p - abover = response.doc('.article-title').text() + response.save['Book_Synopsis'] + response.save['Book_Palabras'] + response.save['Book_Introduction'] #小说状态(连载还是完成)) q+ X- ^3 ]3 Q" Q6 v
- Book_Date = str(datetime.datetime.now()) # 采集时间/ c, {1 u( Y6 j/ G
- BookConte2 = BookConte1.replace(NewRe1 , ReC1)% O9 R3 M$ y! I' R5 o+ |4 ^
- BookConte3 = BookConte2.replace(NewRe2 , ReC2)& V- U4 Z9 a. f! ~* w8 L4 s
- BookConte5 = BookConte3.replace(NewRe5 , ReC5)1 |- E! q: s. ~: H( u8 Q
- BookConte6 = BookConte5.replace(NewRe6 , ReC2)
. i7 u7 I1 w1 g: Z- m - BookConte7 = BookConte6.replace(NewRe7 , ReC2)
( A, W! h1 @; H* ~) } - BookConte8 = BookConte7.replace(NewRe3 , ReC6)
# u8 R" C8 q8 r+ ]: U0 C - BookConte4 = re.sub(NewRe4 , ReC4 , BookConte8)
9 K6 W) o+ ?9 n* _; Z- y - BookConte = BookConte4.replace("\n\n","<br>")$ m5 N9 k' I4 a6 K
- print BookConte
% V7 m5 o" `+ s+ t1 V* R5 e - Book_Introduction2 = Book_Introduction1.replace(NewRe1 , ReC1)
2 |( l. H& }; D6 v. z - Book_Introduction3 = Book_Introduction2.replace(NewRe2 , ReC2)8 Q+ Y4 j m# w6 K, F
- Book_Introduction4 = Book_Introduction3.replace(NewRe3 , ReC3)+ X% Q% n2 b2 O* g: W$ P& k- q
- Book_Introduction = re.sub(NewRe4 , ReC4 , Book_Introduction4)8 a% w5 S0 y# W! \
- Titleid = response.url.split(BookID + "/")[-1].split("/")[0] : `' ?( P1 B3 j3 P. g
- Book_img = response.save['img'], #小说图片! V/ J# a- l" v) g6 T! N
-
# t3 c# {+ P; Z6 K - #insert into MySQL 小说入库' D: [1 b/ A- n( Y
- self.add_question(Bookname,Cater_Name,Book_author,Book_Introduction,Book_Synopsis,Book_Palabras,Bookurl,Booktitle,BookID,BookConte,Titleid,abover,Book_Date) #这行可注译,数据库发布接口,方便其他系统的发布2 R+ `) j, B! u: K8 M$ o6 x8 s
- self.add_comment(Bookname, Booktitle, BookID, Titleid, Book_Date) #这行可注译,数据库发布接口,方便其他系统的发布; r0 U1 f5 F( Z ?6 ~9 ]0 H' M
- #post提交发布
6 f5 A" |$ K, y8 X$ W* u1 `6 k - self.add_locoy(Bookname,Cater_Name,Book_author,Book_Introduction,Book_Palabras,Book_img,Booktitle,BookConte,abover) #这行可注译,火车头发布接口,不需要可取消% e* c9 e1 ?1 n& S
- Datos = {
3 a) R3 o/ K" x8 o - "Cater_Name":response.save['Cater_Name'],; ^2 A: w: P. c. D. C3 }
- "Book_author":response.save['Book_author'],
5 c* h9 f; L6 @, B, G* H/ t# U: @ - "Book_Introduction":response.save['Book_Introduction'],
: _: Z$ G; g( o( `+ J - "Book_Synopsis":response.save['Book_Synopsis'],
- {2 P- i, F/ f - "Book_Palabras":response.save['Book_Palabras'],2 u/ v6 v* l% l- M, l3 M
- "img":response.save['img'],
* ~& u5 L1 @2 \% L! ^; e - }$ J( B: H/ m( a5 x' X7 d- |
- for each in response.doc('.articlebtn > a:nth-child(4)[href*="/xiaoshuo"]').items():+ d0 O6 U; O" h J
- self.crawl(each.attr.href, callback=self.detail_page,save=Datos)
e. q/ }. a, R+ n3 b$ B1 ^9 F" l4 v/ Q - return {" f8 Y' i% t9 ~" L) R
- "Cater_Name":Cater_Name,
$ w2 L3 ~$ H. u/ U- f+ W7 r - "Bookname":Bookname,5 F. q( t( t( ^% l' j( k7 x
- "Book_author":Book_author,1 {' _2 ?4 p- _- m4 t3 \+ i- ~
- "Book_Introduction":Book_Introduction,8 D6 F" y. N8 F# q f
- "Book_Synopsis":Book_Synopsis,
+ V a- x6 U4 ^7 |) c/ r - "Book_Palabras":Book_Palabras,
1 M u/ Z- R( z$ d; ^0 K - "Book_img":Book_img,4 Y- t+ ?& H- E* S" Z6 P
- "Bookurl": response.url,6 [! F$ }* E3 ?, e! A
- "Booktitle": Booktitle,9 K2 H1 E! P) R7 G
- "BookID": BookID,
# W, g- d0 M! ]# m; f" W - "BookConte": BookConte,' j" r6 _2 D5 }' [
- "Titleid": Titleid,8 J' D- P% j; x( u2 f1 P- Y
- "abover":abover,4 N4 S' S3 t- V6 w/ y* |/ D
- # "Book_Date" = str(datetime.datetime.now()),
* m7 P% N0 @7 i7 r( H - }
5 G5 i: z9 ~/ K& s - def download(self, P_dir, imgDir, file_name, Book_img):+ t- b/ ]1 }; {8 J
- if not os.path.exists(imgDir): p) n" Q+ V% f$ O4 ?8 y* k
- os.makedirs(imgDir)
: [1 K& y, `4 y+ _ - file = imgDir + "/" + file_name6 w- \( |0 |1 x) [: d. n
- # print file
; w W% A. p& Y/ [1 n - f = open(file, 'wb+')
' x M+ _; n% L, G7 h - imag = requests.get(Book_img)
/ M' }$ R A0 c$ O& ~7 F0 h - f.write(imag.content)
2 x& [ E4 }% @3 G% S- S" Q6 R - f.close()6 B4 e A7 u' H5 _" I; W
- #保存图片前4 ]" n" Q4 G. T' {
- def save_imgs(self,response):
- O9 Z/ T) g# w. D7 {& G* s - content = response.content
6 S1 ?+ S+ ?2 f; [( ~5 i6 x. Y! i - file_name = response.save["file_name"]1 G e9 h9 V# V% K+ s& g0 N _0 Y1 t
- imgDir = response.save["imgDir"]
4 w. d, L `4 z- V! ~- C8 S6 E - file_path = imgDir + file_name: y% N% i! }! H2 ~+ O( U2 N9 W
- self.save_img(content,imgDir,file_path)
l' y0 p+ N( q1 }/ { - #保存图片+ [/ V/ f$ D" W3 j
- def save_img(self,content,imgDir,path):6 N3 h/ V# F' Q4 a( Y
- if not os.path.exists(imgDir):
6 ?5 p6 l5 ^+ V8 W7 D - os.makedirs(imgDir)6 [, }; a/ }- _2 ]) T3 x* J
- f = open(path,"wb" )( G( }) k4 s& e. t
- f.write(content)
B: D4 S9 N8 V$ y, T! M& x/ E$ @ - f.close(): b# ?1 x! \0 c. z& ^% f
- #获取url后缀名
9 f& C5 B$ R! P) Z; ]% L - def getExtension(self,url):
+ c2 T4 f e: |& W5 k) z7 {, p - extension = url.split(".")[-1] w# n' D4 s5 i) Y
- return extension
* C% D7 f$ K& u+ p# q: k -
7 s9 C* Z# l9 B" X' P: M# M - #获取图片名1 H8 a5 ]. h! T$ Q* D5 Z: `" N
- def getname(self,url):8 s1 [" `8 E% {& Z7 U1 _
- name=url.split("/")[-1].split(".")[0]
- A" E7 M/ f3 i) G! c8 k4 V. [ - return name
复制代码
5 Q$ Z8 ^% B9 p1 n: V / G( Y9 V' G2 ~% s
|