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!
0 `) y* h& h7 x" P# T H, G- #!/usr/bin/env python. u. `# I$ y: ]: ^( q
- # -*- encoding: utf-8 -*-
/ A7 x d; m/ w - # Created on 2019-05-05 21:43:11- P6 i( {- B7 y0 C3 t, C: e0 [
- # Project: XiaoShuo- c6 p; _2 S5 f' o
-
; {) ^1 o2 y" Q - from pyspider.libs.base_handler import *, P5 z5 s8 Z0 u+ q) d: f
- import pymysql
* q3 H* N% r p M+ I/ N - import random
; N* j0 b$ F2 l) ] - import datetime1 o/ }% q/ J2 {
- import urllib2,HTMLParser,re
0 ?5 A2 Y3 t7 V; f; a2 t: [2 h) a1 V - import os
5 ?" c' O& {$ K - import sys
! J; M( e1 @; ^1 w1 N - import re
+ @9 }, Z" w& \/ |3 u' |' n% c& L - import codecs1 `: g4 j9 W. m+ y
- import requests9 g* ~# }5 {( x0 ?( D7 ^
- import json
; n6 f$ H! s4 d -
! l0 m' @% h. M2 i {7 L- \ - class Handler(BaseHandler):
; C k" d. s. k- x - global Datos* N9 \# L8 ]* f) a1 L
- global P_dir 6 a5 T! C X2 o" T, e
- P_dir = '/Tools/Debug/' #采集时候图片保持到本地的路径
! W# i! { Q' v' }9 c% D/ S, i - global Datos9 ~8 m- C+ \6 P9 Z) o' o4 v0 q
- Datos = {}
( t5 i+ L; U9 O5 O0 M- I3 F) c - headers= {4 i" ~7 i" ^" z2 ~9 D
- 'Accept':'text/html,application/xhtml+xml,application/xml;q=0.9,image/webp,*/*;q=0.8',, L" S g Z5 p
- 'Accept-Encoding':'gzip, deflate, sdch',; \. B- S8 e# N
- 'Accept-Language':'zh-CN,zh;q=0.8',
( Y& _6 y% y) H - 'Cache-Control':'max-age=0',
2 l+ B8 s. R' s. z8 B) h( a7 _- O) y - 'Connection':'keep-alive',+ [7 [( g- l% b0 h8 {% M+ e' F
- 'User-Agent':'Mozilla/5.0 (Windows NT 6.1; WOW64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/41.0.2272.101 Safari/537.36'9 F* R+ F& O6 c6 W9 u; F% z: Z
- }
0 B; e0 z! r# v6 H; e - crawl_config = {
x' v( x1 ?& K2 _4 I - 'headers' : headers,/ p5 b( v% x# a2 D. G; {
- 'timeout' : 300( d- N7 c, t! B. y3 _* P
- }
( N& N4 \4 |2 B8 \+ \- n - def add_BookFile(self,Bookname, BookIDs, img, Locaimg, Book_Dates):
: t6 r" W2 d' C6 n$ l - db = pymysql.connect(host="localhost",user="数据库用户名",password="密码",db="数据库名",charset="utf8")$ l& {6 \, f2 e5 P+ p( [6 F
- try:+ T) P {1 g9 A, t1 b) y
- cursor = db.cursor()/ r2 m) I/ P8 |! t3 V# j: t* m5 m
- #注意此处字符串的占位符要加双引号"%s"$ Q0 v! e' e1 |4 `5 p- O
- sql = 'insert into BookFile(Bookname, BookID, img, Locaimg, Book_Date) values ("%s","%s","%s","%s","%s")' % (Bookname, BookIDs, img, Locaimg, Book_Dates);
7 l P; r. H$ e' h5 E1 c - # print(sql)
: L7 [5 f8 |8 g5 [ - cursor.execute(sql)6 J* x) z% X- b
-
( O0 x: h4 S: q' [2 _# D( ` - #qid = cursor.lastrowid
" |, t. v& g; \% m4 D - #print(qid)
4 N! R6 o. T. T/ j - 6 X7 ?2 J8 s3 ^ _7 ^
- db.commit()
9 i- ]" e* ?' s8 I4 U, |0 } - except Exception as err:
8 H3 J7 G4 U3 V" h, ^. V' p - print("Error %s for execute sql: %s" % (err, sql))
) W* {) Y- C0 g2 d - db.rollback()3 S( b6 f- q I5 z' B
- def add_comment(self,Bookname, Booktitle, BookID, Titleid, Book_Date):3 f) ^- [6 E6 y% p/ Z! o$ F7 i
- db = pymysql.connect(host="localhost",user="数据库用户名",password="密码",db="数据库名",charset="utf8")
, J3 ^- `/ J O: m q' ] - try:
/ Q/ d! r) U0 D8 s9 e p/ V' g7 [ - cursor = db.cursor()% P6 U# n7 ~8 i* N2 x$ n
- #注意此处字符串的占位符要加双引号"%s"- \. w3 A, S4 w2 ?$ B- [
- sql = 'insert into BookTitle(Bookname, Booktitle, BookID, Titleid, Book_Date) values ("%s","%s","%s","%s","%s")' % (Bookname, Booktitle, BookID, Titleid, Book_Date);
/ `( E, Y5 o5 c6 c4 @2 i9 X - # print(sql)& |# ]3 l* R! ?8 z7 o$ S I
- cursor.execute(sql)5 b, k$ C5 f, ], ]6 b
- $ Y% @/ O! F8 ~; V
- #qid = cursor.lastrowid
, X" u; |+ j# e- X# v - #print(qid)
" ^$ i9 U9 ]1 ]$ h) ]0 Q6 U, Z - 2 u3 |$ r$ M9 f' _( h
- db.commit()
0 a: s$ M# q1 b* ^3 } F( a( x - except Exception as err:
8 Y% ` W8 d' z5 j& L" T9 {5 E - print("Error %s for execute sql: %s" % (err, sql)); ?7 M" E+ u# d$ y$ d0 e
- db.rollback()" x8 F) d* k( \7 L
- def add_question(self,Bookname,Cater_Name,Book_author,Book_Introduction,Book_Synopsis,Book_Palabras,Bookurl,Booktitle,BookID,BookConte,Titleid,abover,Book_Date):
. A- l- C- q& N& R) ?! c! T - db = pymysql.connect(host="localhost",user="数据库用户名",password="密码",db="数据库名",charset="utf8"): a+ u2 V) V: _% X! a! Z
- try:
0 [- I9 }$ e3 L2 U* A) D$ y - cursor = db.cursor()& s0 {' |/ ]2 R* `6 V& P
- #注意此处字符串的占位符要加双引号"%s"6 [* t. p* ?8 E6 i8 b* K9 n9 @
- 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);2 R1 k9 y" q+ h1 ]& q
- print(sql)# j1 S; ~* h" E9 C ]$ D$ e1 I
- cursor.execute(sql)
7 z/ _) g# X+ `, K4 ^& B - print(cursor.lastrowid)& x7 [2 {) p- A9 c5 t) k
- db.commit()4 K& x5 r0 b2 p! X4 e8 Y- C, w, b
- except Exception as err:
^8 E0 l9 M |+ n/ {' x5 |" z# a - # except:
+ x+ f- q% Z W6 y; S+ u! A - # print('Failed')
! H( n( D4 w! |4 } - print("Error %s for execute sql: %s" % (err, sql))
. X. w3 s$ `: U2 |) i5 C b - db.rollback()
; r) N/ ]& A3 w' |0 Q$ l+ E - , P) y) r# s- l
- def add_locoy(self,Bookname,Cater_Name,Book_author,Book_Introduction,Book_Palabras,Book_img,Booktitle,BookConte,abover):
; l% y$ E6 ~5 c( Y+ O% I7 k4 N - reload(sys)
' | @; D, C! R - sys.setdefaultencoding("gbk")' W3 t* o3 ~1 c8 r. r. `! ?; l
- locoy_url = 'http://www.******.net/locoy/?my=book' #697火车头发接口地址
+ o7 Y6 m6 X7 I: u# e0 ^0 k1 t - locoy_data = {% H; @& R' O# V
- 'my_u':'用户名', #后台用户名: U* {- j% t- R0 G( h; w: }$ z
- 'my_p':'密码', #后台密码
; C; q R# N1 C3 x - 'subject_669977_net':Bookname.encode('gbk', 'ignore'),
# X" E% Z8 a# r1 ?6 X+ X2 g - 'caid':Cater_Name.encode('gbk', 'ignore'),
0 k* a, Z' j. s% K# f, i0 d! P - 'title_669977_net':Booktitle.encode('gbk', 'ignore'),, N. E2 x3 I3 b
- 'article':BookConte.encode('gbk', 'ignore'),
1 f0 V1 K9 Z/ n9 Z. r9 u - 'author':Book_author.encode('gbk', 'ignore'),$ w* c& M, |) ]/ R |* P* i( \. h
- 'ready_1':Book_Palabras.encode('gbk', 'ignore'),. {$ ^- H' A X
- 'thumb':Book_img," j z& z+ e4 T1 v
- 'content':Book_Introduction.encode('gbk', 'ignore'),
% n/ i" l% Y+ _+ M$ Y4 D - 'abover':abover.encode('gbk', 'ignore') ; J# q' v- y9 X% G, W; I2 X
- }+ @0 v; I; H m3 q! P5 Y' a- L% }
- res = requests.post(locoy_url, data=locoy_data)
/ J4 L G- F/ s' {- [ - print res.text0 f) p/ r% i5 ]' n
- print res.content- ]( C7 {$ o$ S' ~; d8 ~
- # print Dsd- |( m( u R* f* M% X5 a0 A4 x
- return res# E9 E% B- \8 i2 H
- 7 z% |4 V' q( m& ` @4 z
- def __init__(self):7 T3 Y' w5 C5 v' g( t6 G8 M
- self.base_url1 = 'https://www.****.cc/'0 G8 [$ n) o/ k! G8 c6 f/ O) ]' i
- self.base_url2 = '/'. J/ @" F* x t4 h; x! l# M5 b4 ] @
- self.CaterId = []
- P U& m+ L6 P# o# [, b - self.CaterIds = ['xuanhuan', 'wuxia', 'yanqing', 'lishi', 'kehuan', 'kongbu', 'nvsheng']* O+ z) q# |2 S8 W# Q
- self.page_num = 1
% @( n/ `; ?' S% e - self.total_num = 200
$ I+ y1 `8 @$ I1 t. D3 y; I9 \3 _, M -
4 {' }2 J+ I: Z; D5 S9 c - @every(minutes=8 * 60)4 K, i, P7 d. S1 `! E
- def on_start(self):1 l2 _8 \2 ` e! n0 R
- global Cater_Name1 K9 X8 |0 k3 i+ q
- Cater_Name = []) `; @- [% c0 Q7 @7 {* u) P
- while self.page_num <= self.total_num: # B# x" ~' R [. L+ q
- for self.CaterId in self.CaterIds:
* ?' t: T2 Y3 y. }8 f: G) X9 E - if self.CaterId == 'xuanhuan':- o( y$ s r) T) o- I
- Cater_Name = '玄幻'
8 u9 p: r' s. d! A - if self.CaterId == 'wuxia':" |" X& r5 W3 a& {# e
- Cater_Name = '武侠'; T& a7 G0 n2 L5 ?2 l. @
- if self.CaterId == 'lishi':
2 Y% | w p6 g* x! x6 y9 \4 E$ E5 k - Cater_Name = '历史' ( U3 @7 Z; M4 H6 }. G% D9 c. T- u
- if self.CaterId == 'yanqing':/ x! v9 x6 c4 Y! _7 L0 o
- Cater_Name = '都市' T6 X N0 y, G$ N9 f) I6 D' _
- if self.CaterId == 'nvsheng':
h/ A4 l8 p4 y1 B, L0 [7 d3 M - Cater_Name = '都市'
1 ]' |9 J/ g/ K! Q# s j - if self.CaterId == 'kehuan':
" E8 w, ^: j( y( [ - Cater_Name = '科幻' ( [% s0 b- S- [5 ?& U6 S
- if self.CaterId == 'kongbu':
4 F+ j2 f1 _9 P) Q) R/ W - Cater_Name = '游戏' 0 ^% G% v% u( i( m3 W
- print self.CaterId
$ }& R$ g; A. J4 J8 G) v2 e* a B - url = self.base_url1 + str(self.CaterId) + self.base_url2 + str(self.page_num) + "/"
# ^4 |5 I' E z- e - self.crawl(url, callback=self.list_Caterg,save=Cater_Name)
3 d M( k4 p& R% Y6 ?& y( R2 A! v - self.page_num += 1
3 R. a2 {' A4 C% }' {! ? - ! O( U$ \; M3 j k' _" ^
- def list_Caterg(self, response):
5 s* y- C* B) S6 Z - Cater_Name = response.save
+ x2 r. N& B/ F5 N. d4 j+ e7 Q" q - for each in response.doc('.pic-list a[href^="http"]').items():2 U# q. W- E- {& d3 `/ U' n
- self.crawl(each.attr.href, callback=self.list_Caterg_detail,save=Cater_Name)% {4 Y' g4 ~; b! O' ]" ?$ q; n# ~
-
6 ^* p$ ?" o% F9 v: c4 E - def list_Caterg_detail(self, response):
1 Z/ V& [1 ?" m- ~$ _, k1 } - Cater_Name = response.save' y4 \! Q5 \0 z _6 x9 N. r0 Y
- # print Cater_Name, b6 M' Q! E5 I2 w0 P- [0 K# [
- Bookname = response.doc('h1').text()
0 a5 D- s5 Q$ T - print Bookname$ |+ ^3 Y9 J, b# {; a% i, }
- Book_author = response.doc('.authorname > a').text()
9 t9 \- B8 l# V - # print Book_author
0 ^( e6 z4 h; T- ~) c2 T - Book_Introduction = response.doc('.book-intro > div').text()
9 Y) ~3 b) V4 E - # print Book_Introduction7 h. V: g5 V6 u: o
- Book_Synopsis = response.doc('b').eq(1).text()
! W) N+ ~" k8 V/ A# t3 `+ `! \ - # print Book_Synopsis
4 B7 q& O" k' F6 [/ P) Y3 t - Book_Palabras = response.doc('.booktitle p').text().split(' ')[1].split('|')[0]: u! G }1 V7 w1 `
- # print Book_Palabras
% w7 j+ S- j* f) R$ C) X - BookIDs = response.url.split("xiaoshuo/")[-1].split("/")[0] #小说ID
. P l, v# B0 N& v/ l - # print BookIDs9 x0 i8 q& R; o: |% }
- Book_Dates = str(datetime.datetime.now())
+ J/ ^' _ Z5 F5 _% p- C4 ? - for imgs in response.doc('.bigpic > img[src^="http"]').items():
9 }8 j+ |5 E2 l M# K* U% H - img = imgs.attr.src
7 S, _1 p Y6 C& m - print img
* M, x5 t G& O* E4 m - #小说封面下载
* b+ T: e4 |+ v, j - extension = self.getExtension(img)
/ t! Z* G% H# Q( w - name = self.getname(img)
# ?1 }* t3 n/ C- U# ]8 a- F' B - file_name = name + "." + extension; h. e4 K7 t7 U+ D; Z: N
- imgDir = P_dir + name% ~( S" K/ J2 g5 T' b
- Locaimg = imgDir + "/" + file_name
; [+ C; r# b% W9 i# ^ - print Locaimg
: o, z; e! X" F4 t- f8 c' T* ~! g - if(self.download(P_dir, imgDir, file_name, img)): #这2行可注译,图片下载到本地
- r8 z5 m, t7 @% m$ z - print('attachment url is ' + img) #3 ~# u( a, ]$ B: l, y
- Datos = {
$ P4 Q6 [: E0 O. G q4 L - "Cater_Name":Cater_Name,
( K! T# D$ S( y' X) z - "Book_author":Book_author,7 A8 }2 j2 i- A" }! K) L) B
- "Book_Introduction":Book_Introduction,# V' I( O9 R7 R
- "Book_Synopsis":Book_Synopsis,
* B) L: @* x- j: D* w - "Book_Palabras":Book_Palabras,* E! T% q( t5 K% {$ a) ?/ t! M6 A i" C
- "img":img,
3 j2 {, I8 t$ s. s2 \: K [ - }
$ j; R* Z3 t! r8 ` - 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
" K- x* |; D6 Q9 A' o9 i v7 y - for each in response.doc('div[class="bookbtn-txt"] a[class="catalogbtn"]').items(): c- s. ~3 L9 Y2 u
- self.crawl(each.attr.href, callback=self.index_page,save=Datos)7 l' G' C$ e* f4 W
-
4 Y! i2 N+ e% l2 ], d- k" O8 G - @config(age=8 * 60 * 60) . e5 B9 H0 \; @7 E& h' |
- def index_page(self, response): $ L+ w# Y( c6 B$ N- r& P9 A; ^" c% Y
- Datos = {
+ [: U& w( w# p2 K0 j t - "Cater_Name":response.save['Cater_Name'],
' p Z4 }5 {/ q! V8 m - "Book_author":response.save['Book_author'],
7 `, I' k" ]# D5 b! X - "Book_Introduction":response.save['Book_Introduction'],
% M4 Q7 G/ B. x- R/ Y, O - "Book_Synopsis":response.save['Book_Synopsis'],. A l, H7 x; f% H, a
- "Book_Palabras":response.save['Book_Palabras'],
! V* h) i) L* j7 ~2 l2 Q - "img":response.save['img'],
% W0 g" K& U: w8 { T0 t- H+ T" U& o - }" a2 |, u) c X k t* S* q- \
- for each in response.doc('.chapter-list li:first-child a[href^="http"]').items():6 j6 `( s& p. G4 O
- # for each in response.doc('.chapter-list a[href^="http"]').items(): , a' I, N4 l! c, v4 g# }) w$ b
- self.crawl(each.attr.href, callback=self.detail_page,save=Datos): E% V% d- _8 v, [ K
- @config(priority=2)
O5 f8 z: g' J. F+ v0 i9 F - @catch_status_code_error
6 U9 E6 C8 e0 P5 U6 `# [ - def detail_page(self, response):
" L4 B, A" M& ]1 E( N5 @ P c - NewRe1 = u'哈书'& o7 P" \; H a: t6 H4 T' P
- NewRe2 = u'huhjsd.CC'
. M$ E; e- [ ?% h8 x$ n9 o: | - NewRe3 = r'^\\n\\n'
5 Q4 h" V/ p% }& M+ H; w - NewRe5 = u'小说网'
3 [4 y# y8 ]' q$ z- W# d8 e2 e* } - NewRe6 = u'fgdfgf'
/ \; _3 c) Y/ ], } I - NewRe7 = u'fgfgf'
" B" L+ S; N, c6 O3 N1 n - NewRe8 = u'ffhgf'
+ ?$ ^" p% Z3 n$ i+ i0 w2 N9 g& m - NewRe4 = r'[\f\t\v+\.\{\(\)\}\!\/_,$%^*(+"\')]+|[+——()?【】“”!,。?、~@#¥%……&*()]+'
# r/ }% T. |/ o3 j - ReC1 = u'静思'6 N! g9 n1 F: P. \
- ReC2 = u'aghgf.com'8 X" Q& h& e, B: ^3 E5 n2 h
- ReC3 = u'aghgfh.com'
% K7 M! P o7 @ - ReC4 = u''
) f) O, r7 K: J v+ O - ReC5 = u'文学网'
4 R0 E0 `* A& @4 i! [, @) i - ReC6 = r'<BR>' y$ p8 ?" j/ b8 n( h; X5 ~( t j) _6 a; g
- Bookname = response.doc('.readlocation a').eq(2).text() #小说名称( ^3 Z3 H2 V. U! S2 H5 b1 C( l
- print Bookname* N! p* L% T2 a8 }: B
- Cater_Name = response.save['Cater_Name'] # 小说分类 f+ I( _( R2 m n
- Book_author = response.save['Book_author'] #小说作者
& I4 d5 X3 e7 ?4 d& Z - Book_Introduction1 = response.save['Book_Introduction'] #小说简介
) E+ q: c0 T+ _ - Book_Synopsis = response.save['Book_Synopsis'] #最近更新# Z3 s3 ^* E) t5 s. |/ F0 j* y
- Book_Palabras = response.save['Book_Palabras'] #小说字数, V' ~1 m0 d3 H2 ~
- Bookurl = response.url #小说网址3 K5 R# \) k) h) M4 r* _
- Booktitle = response.doc('.article-title').text() #章节名称
) E$ a3 |, J( Y1 Q2 | - BookID = response.doc('.readset-r span').text() #小说ID! C3 Y: O- ^ c$ t5 c
- BookConte1 = response.doc('.article-con').text() #小说章节内容) V) g2 E& l2 H, D6 c
- 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)
8 h: K2 T( u: p - Book_Date = str(datetime.datetime.now()) # 采集时间
: x) C# I$ H" _ I4 k) ? _ - BookConte2 = BookConte1.replace(NewRe1 , ReC1)' D8 x1 ?6 {' M$ a; q
- BookConte3 = BookConte2.replace(NewRe2 , ReC2)
$ ?6 I) A2 L- N9 A% p - BookConte5 = BookConte3.replace(NewRe5 , ReC5)
+ X8 G8 f' f" ]* | - BookConte6 = BookConte5.replace(NewRe6 , ReC2)
# Y$ L) W) U0 Q. N6 s: P+ F - BookConte7 = BookConte6.replace(NewRe7 , ReC2): O5 [6 [2 @: }5 V/ W3 r& i
- BookConte8 = BookConte7.replace(NewRe3 , ReC6)$ \3 V/ Q3 ^+ V& q# z& L
- BookConte4 = re.sub(NewRe4 , ReC4 , BookConte8)$ F3 E7 X/ S8 x9 t7 `: y- g
- BookConte = BookConte4.replace("\n\n","<br>")
- n/ e( J- l# @ - print BookConte
: p& A! O+ V1 G' F2 ^8 ]3 O/ \ - Book_Introduction2 = Book_Introduction1.replace(NewRe1 , ReC1)+ m# Q* s4 O5 [" J* j1 e% X
- Book_Introduction3 = Book_Introduction2.replace(NewRe2 , ReC2)
8 W( n( s |8 D - Book_Introduction4 = Book_Introduction3.replace(NewRe3 , ReC3) j. U. f. X8 f
- Book_Introduction = re.sub(NewRe4 , ReC4 , Book_Introduction4)" i; A" u/ R" T t) R
- Titleid = response.url.split(BookID + "/")[-1].split("/")[0] 9 j, A- M: x$ P3 P" j( Z
- Book_img = response.save['img'], #小说图片
4 d7 Z$ S" W( v -
4 i, Y4 {% o( @: `$ ^9 _7 ~ - #insert into MySQL 小说入库
! q# U6 W' K. I- x - 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
3 s' ^- {% Y& ?2 B$ z8 W6 Y - 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
% i+ w+ X, z/ d( s- B8 R+ g j, q' | - #post提交发布# e" |3 H' E @8 x- {1 b3 E
- 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
: R% [: k& x: Z5 A1 |1 S4 } - Datos = {' e2 x) L8 g& Z0 p
- "Cater_Name":response.save['Cater_Name'],
# i: X5 n0 X' C: x - "Book_author":response.save['Book_author']," M* N, O$ S8 k: p7 @* p
- "Book_Introduction":response.save['Book_Introduction'],' A# O" M7 ]2 f' C. H
- "Book_Synopsis":response.save['Book_Synopsis'],
+ F! t k4 [- ?3 `. Z8 q z1 O0 N: e - "Book_Palabras":response.save['Book_Palabras'],
: D. D1 ~0 I, ` - "img":response.save['img'],: Z- T' y; v- }9 E) v. u8 b
- }
6 K$ |5 E9 W; |( z% y$ F* Y! O" _ - for each in response.doc('.articlebtn > a:nth-child(4)[href*="/xiaoshuo"]').items():# Z9 X+ a4 q( m" @
- self.crawl(each.attr.href, callback=self.detail_page,save=Datos) & P2 i5 `6 I' _* k
- return {7 T2 A2 R" b! {# _! W
- "Cater_Name":Cater_Name,
. Z F2 M# M- S( y' K - "Bookname":Bookname,
$ b0 S! w: k2 b' i' i - "Book_author":Book_author,
0 }3 _, ?! G% Q- m; ` - "Book_Introduction":Book_Introduction,/ Z: o' o- q. u
- "Book_Synopsis":Book_Synopsis,# b# ?5 Z/ P: r
- "Book_Palabras":Book_Palabras,
6 D7 d. A2 P# G# ~- Q! R) X5 n$ L* z - "Book_img":Book_img, {- T, t" k! @5 m& }
- "Bookurl": response.url,
6 j" z& k8 f/ E/ L; Z% }- v$ k9 |* l - "Booktitle": Booktitle,- b( j' G' q4 C. V
- "BookID": BookID,, G' _1 r* K( X. ~
- "BookConte": BookConte,- L7 i. Q d6 r: o+ Z9 V
- "Titleid": Titleid,( G: I! G8 y- w8 t& y
- "abover":abover,# j( b. I! R8 \+ [$ a# B. i7 C
- # "Book_Date" = str(datetime.datetime.now()),
: v, d2 i& w- @7 o; U9 Y5 x7 j; T - }4 H/ `; U4 h# d0 l8 s* @
- def download(self, P_dir, imgDir, file_name, Book_img):
/ P5 M+ X! s8 x2 Q8 m8 b4 Y. } - if not os.path.exists(imgDir): 8 H/ X; c8 z1 d! J& c6 h9 h
- os.makedirs(imgDir)6 n. m0 Q3 h& |
- file = imgDir + "/" + file_name( x6 |% ~4 ~, i8 K; }8 D$ T
- # print file4 X3 o5 c& H- E! ^9 y) ?1 f1 A
- f = open(file, 'wb+')
/ {: f3 Y1 o8 P4 ]$ h: H: \& T - imag = requests.get(Book_img)
5 a `3 g7 l+ Z+ v9 X; ^- j( Q - f.write(imag.content). g" M# _! t! i5 A
- f.close()
6 L9 M0 R' E1 S( I4 f' j. I* g - #保存图片前* g, N- N2 o2 m8 Y0 I
- def save_imgs(self,response):6 A7 N6 C2 b& K" D0 v( N. O) e
- content = response.content2 h# }& \+ p" u
- file_name = response.save["file_name"]
& P: O: X6 a, i) [( a, A& q- \ - imgDir = response.save["imgDir"]
+ Z! [9 A: o7 ~; p# |# k& [ - file_path = imgDir + file_name) X( |2 h1 u, b; U" k( d
- self.save_img(content,imgDir,file_path)
& Q4 q+ g5 u3 @; q - #保存图片
7 Q7 X' \: A0 p$ @5 W' @ - def save_img(self,content,imgDir,path):( U5 q& n( M: E* U3 X
- if not os.path.exists(imgDir): 3 {. g1 L! z1 s
- os.makedirs(imgDir)
& k. }2 b4 r9 n( E2 {5 ^ - f = open(path,"wb" )5 t& t; B$ g1 P0 w
- f.write(content)8 X* Q, x9 M! m3 O' w
- f.close()
3 Y/ T0 c5 e4 |* z' r% ?2 S# v6 ^ - #获取url后缀名
, j6 Q% @2 n* Y( P: Q" b - def getExtension(self,url):
1 j- _2 F9 z' U: ] - extension = url.split(".")[-1]% M7 k6 Y9 U# e$ ? \
- return extension
H, i( w4 B: c5 ^* X2 Z - ( N' ?% e+ Y d2 U' w
- #获取图片名. F6 ]2 C4 e4 z5 F
- def getname(self,url):
8 a" z# e% W6 p/ j3 r - name=url.split("/")[-1].split(".")[0]
, g' S7 w6 F3 n" A2 ? - return name
复制代码 ( C+ \ G: [! g6 D% e
( \% r1 |* ?$ W5 @" y( e
|