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!
6 M! o$ K9 P' x3 D0 v- #!/usr/bin/env python( v7 {3 @ V( x! t0 A
- # -*- encoding: utf-8 -*-/ i' a2 z/ D" r; t/ p' b0 q6 S
- # Created on 2019-05-05 21:43:11
3 G- _) V" z# m/ y9 T - # Project: XiaoShuo
6 e7 L% ?$ n" v6 |2 c8 F `7 t7 \8 @ - # Y# d8 M7 F+ U: S+ d9 @
- from pyspider.libs.base_handler import *; z1 R( K8 @; W* ?9 O
- import pymysql
& Z- C2 p/ l. L; O- M - import random& v) ]2 R7 |. R: s1 a+ x- O( V& k
- import datetime
o9 R( O: I' b' {; ~ - import urllib2,HTMLParser,re) l7 k6 Y3 P% ]. k
- import os( `, K- \& C9 W! _
- import sys
$ D+ `! {+ g7 x) s& U- a* d1 Q - import re
# e7 @! z7 X: {+ g' z5 t4 U - import codecs
( T @; [# o* I! e5 j$ ?) K- G/ g - import requests
! u1 m3 N! H* i# y - import json
* o4 o* z7 |; k+ ], B/ } - : T' q# }% |8 _: m, _9 X1 m* b
- class Handler(BaseHandler):, O' u, g' g- A" o& u% n% R
- global Datos7 n# `% M1 g' h7 @0 X% w
- global P_dir 3 B4 r4 w; ]8 `: @- W8 Q
- P_dir = '/Tools/Debug/' #采集时候图片保持到本地的路径
5 }6 C9 `5 b" h( u/ x+ r1 ~ - global Datos
+ n' v9 [7 F# p Y0 z* i& c% e, U - Datos = {}0 L x3 y! G9 ~
- headers= {
- \9 C0 Y' j1 |2 @. t$ e6 ?6 ]! l - 'Accept':'text/html,application/xhtml+xml,application/xml;q=0.9,image/webp,*/*;q=0.8',, ?! l" o r$ p* i; x0 a0 |# i/ D
- 'Accept-Encoding':'gzip, deflate, sdch',
! n5 S9 i1 X$ Z9 }' R! O2 u - 'Accept-Language':'zh-CN,zh;q=0.8',2 @6 L( e1 |/ Y7 C& M' e$ R
- 'Cache-Control':'max-age=0',/ ]) W. f+ v6 V6 I
- 'Connection':'keep-alive',( N& R. g. Q' }
- 'User-Agent':'Mozilla/5.0 (Windows NT 6.1; WOW64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/41.0.2272.101 Safari/537.36'
, A% y' m8 {: \ - }6 H4 ?2 H& ^3 D- v- ]; c
- crawl_config = {
' i) C3 m$ C; \- _ - 'headers' : headers," |* ^1 B/ j9 P
- 'timeout' : 300& S/ \# |8 |4 X% t4 ]# ]
- }$ n( I: @/ n. p* L$ K
- def add_BookFile(self,Bookname, BookIDs, img, Locaimg, Book_Dates):, G) h0 P/ j3 Y! p& \
- db = pymysql.connect(host="localhost",user="数据库用户名",password="密码",db="数据库名",charset="utf8")
/ c, s* S8 i1 Q0 E5 R' ~4 e - try:$ [! p; s% T9 E) P
- cursor = db.cursor()
$ b9 ^" K5 v* Z$ p1 E, U, Q - #注意此处字符串的占位符要加双引号"%s"
2 x0 x% k# s. y. N* Z - sql = 'insert into BookFile(Bookname, BookID, img, Locaimg, Book_Date) values ("%s","%s","%s","%s","%s")' % (Bookname, BookIDs, img, Locaimg, Book_Dates);
- V/ w( |4 U- ^7 h# y - # print(sql)9 X5 N; b _# ~. ?# Y. q# z1 r
- cursor.execute(sql)
8 j6 a* T' z4 N: ^5 u; T/ P - $ j' s* C3 `' h6 Z$ w9 d3 ]1 y
- #qid = cursor.lastrowid6 ]0 E/ a5 \8 ]0 p
- #print(qid)
; o& o' a6 s f4 m N" n( p - ) `5 b/ [$ s* ^ Z6 |9 d$ [$ f7 p
- db.commit()
$ r3 i$ V6 A; S- E/ y, L - except Exception as err:
8 K2 s6 A/ a9 n1 x$ l3 p - print("Error %s for execute sql: %s" % (err, sql))
. k6 \5 @" S* h+ }6 b9 b1 j" V - db.rollback()/ P& f. D1 m, W6 r3 P
- def add_comment(self,Bookname, Booktitle, BookID, Titleid, Book_Date):
3 ^& J$ n5 D5 f( g+ R - db = pymysql.connect(host="localhost",user="数据库用户名",password="密码",db="数据库名",charset="utf8")! P/ E& l, _9 e; K) x% C6 l
- try:5 f, D2 U3 ` U1 C: I
- cursor = db.cursor()
/ f5 o% h. ^ A7 |: n, i, Z v - #注意此处字符串的占位符要加双引号"%s"' t5 \+ c- v* m8 Z2 q
- sql = 'insert into BookTitle(Bookname, Booktitle, BookID, Titleid, Book_Date) values ("%s","%s","%s","%s","%s")' % (Bookname, Booktitle, BookID, Titleid, Book_Date);
4 ]3 A# z, ?$ h! ~0 \9 ?# Z4 k - # print(sql)
9 H4 c! k* r( C+ S* j - cursor.execute(sql)
7 ~% r) _& Q) l - + t7 K6 N* J0 `* O& w! ^+ Z# M
- #qid = cursor.lastrowid. ^5 p2 l' c. K( i( @' W
- #print(qid)
9 ]1 G- r. Y! b$ y# ~; H; H, u/ n+ V - 0 z# M7 }$ Z# J& L6 C8 @
- db.commit()
+ S5 P: ?1 R( M Q" L. ?2 G - except Exception as err:
0 G& T( g% e6 {/ v" E- o; p - print("Error %s for execute sql: %s" % (err, sql))! s& d* u4 ~& q0 ~( _0 z( C7 i
- db.rollback()
# {1 @0 ]& z% j0 z2 A9 }1 U/ C - def add_question(self,Bookname,Cater_Name,Book_author,Book_Introduction,Book_Synopsis,Book_Palabras,Bookurl,Booktitle,BookID,BookConte,Titleid,abover,Book_Date):* i+ N2 R* G1 f; o. G. H0 a
- db = pymysql.connect(host="localhost",user="数据库用户名",password="密码",db="数据库名",charset="utf8"): m5 u6 O: ?: [( P6 ^( N
- try:: P, z3 c' i6 k- E6 D% l
- cursor = db.cursor()- n: l S$ `2 Z# F e6 N
- #注意此处字符串的占位符要加双引号"%s"
) q2 B8 t4 B% {) C - 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);
) |5 p b% L1 b6 {) s - print(sql)
% [/ @& c# b( u - cursor.execute(sql)/ y6 K+ Y# E3 B0 U& x+ H; h
- print(cursor.lastrowid)! C" @7 P0 y. j
- db.commit()
5 I9 W% s8 o8 d! g* Z- F' }5 z - except Exception as err:# h. s8 G/ P" }2 D: O" H. _
- # except:
" V& s- V; n) K - # print('Failed')
* T3 _/ B- v6 ?( ?) s - print("Error %s for execute sql: %s" % (err, sql))
/ A: c* Y; K8 a. x$ G - db.rollback()8 C" K) }. X' q, S1 d
-
) S( u4 K5 n2 F; R/ T# x - def add_locoy(self,Bookname,Cater_Name,Book_author,Book_Introduction,Book_Palabras,Book_img,Booktitle,BookConte,abover): 7 B. I5 x3 i+ x
- reload(sys)
3 O, l3 {9 V7 | - sys.setdefaultencoding("gbk")$ ?8 J5 P8 L+ U' ]
- locoy_url = 'http://www.******.net/locoy/?my=book' #697火车头发接口地址
( K, h5 w* S: o b3 u: {, L+ H - locoy_data = {5 b& m4 L. O1 W- z# a* D5 {- W* D) D
- 'my_u':'用户名', #后台用户名 A, v: m5 ~2 m1 @# h( a
- 'my_p':'密码', #后台密码, [0 W! m0 V* p- i. V* E/ [. D
- 'subject_669977_net':Bookname.encode('gbk', 'ignore'),
! A! g# l( R/ L" x; C- y$ ~8 j - 'caid':Cater_Name.encode('gbk', 'ignore'),! s' X T _" u' [. _% e j( T
- 'title_669977_net':Booktitle.encode('gbk', 'ignore'),
6 c) {; v0 }# a - 'article':BookConte.encode('gbk', 'ignore'),& L4 [* j$ q/ Z) z2 B( i
- 'author':Book_author.encode('gbk', 'ignore'),4 K, w& }5 d6 @% X, F
- 'ready_1':Book_Palabras.encode('gbk', 'ignore'),
3 v9 C- Q8 ]& o2 n4 T9 B7 c5 X* d - 'thumb':Book_img,
# z I( L7 Q" D8 } - 'content':Book_Introduction.encode('gbk', 'ignore'),
) `0 ^$ P- U, M" p- Y - 'abover':abover.encode('gbk', 'ignore') : `2 y: W% _" _, p. W& `
- }5 \0 o9 N6 _4 D# b
- res = requests.post(locoy_url, data=locoy_data)* N/ Y0 c, G! i/ g8 |
- print res.text; Y w$ D) o+ U7 k9 F6 c- F
- print res.content* x& l* x8 @1 ?
- # print Dsd' g6 y9 ~% h. {/ M, w
- return res
% g' {1 i5 d; c2 B - . i0 p3 h0 A6 Y$ D( @; |) p
- def __init__(self):
4 m2 ~0 c2 N1 Q4 z/ a4 M - self.base_url1 = 'https://www.****.cc/'4 O- G5 R3 _! d* L
- self.base_url2 = '/'4 N4 a( U" y2 F2 V% s
- self.CaterId = []
3 h2 y5 _: c% H - self.CaterIds = ['xuanhuan', 'wuxia', 'yanqing', 'lishi', 'kehuan', 'kongbu', 'nvsheng']0 B9 ] d. B# t" z4 n& z
- self.page_num = 1
3 _/ Q8 d0 C7 U - self.total_num = 200 ( j. B* l0 ?2 T/ L0 D
- 0 @8 _! O6 A$ ~3 N7 @
- @every(minutes=8 * 60)
- F' ~3 {0 V) l - def on_start(self):
; X6 P' D1 {) E% b" d - global Cater_Name
' q3 u- o* f& }, K) v1 |1 v6 \1 [ - Cater_Name = []
6 r& b5 |$ M. q- K7 X: E6 L4 j; o - while self.page_num <= self.total_num: 0 K5 {7 ]% W3 T+ G
- for self.CaterId in self.CaterIds:
) U; ]& a/ m; e$ r; }$ R7 [3 M - if self.CaterId == 'xuanhuan':# X8 Y% X8 d8 c2 z
- Cater_Name = '玄幻'* l+ l a) j* A8 O
- if self.CaterId == 'wuxia':
! ?6 p& j6 r* p& Q - Cater_Name = '武侠'
3 x' V6 I+ @. b( }2 H; u- ~+ f6 w# G - if self.CaterId == 'lishi':- a) s% c: z- V( P0 G% @% Z
- Cater_Name = '历史' % ?9 q' [& f& O# L3 E9 i
- if self.CaterId == 'yanqing':, ~9 Y9 t' l8 e& u
- Cater_Name = '都市' % w7 ?+ r( y6 t5 ^
- if self.CaterId == 'nvsheng':
1 U7 a5 D5 z" I7 M; i: ?) u& q' R - Cater_Name = '都市' 4 N' l0 C( W( f
- if self.CaterId == 'kehuan':
: x5 c# J" c& L0 @& Y8 s$ p - Cater_Name = '科幻'
# `# w1 @- ]( {9 M* l1 f - if self.CaterId == 'kongbu':9 ?" s3 ?. J. l. q3 {# f
- Cater_Name = '游戏' 3 h/ S/ J! i( P1 |# f: `( z. F
- print self.CaterId' X. ~( N7 T0 V$ X2 V- l: X8 k
- url = self.base_url1 + str(self.CaterId) + self.base_url2 + str(self.page_num) + "/" & T; r# ^( c3 k! I \
- self.crawl(url, callback=self.list_Caterg,save=Cater_Name)& u5 P. x) B2 f
- self.page_num += 1
. f1 M$ c2 {. E6 O) `# r - 5 v) m6 k. @5 @. C N
- def list_Caterg(self, response):
9 w' t: O7 r4 G7 E - Cater_Name = response.save, i6 y! @" o# t4 Q2 Z: X9 Q+ r$ ~
- for each in response.doc('.pic-list a[href^="http"]').items():' b7 x. q* Q. L+ K! g$ H
- self.crawl(each.attr.href, callback=self.list_Caterg_detail,save=Cater_Name)
. u, v/ b: S& c( x -
) B1 n& c5 b9 V+ W8 V# ?7 x) a" S) K - def list_Caterg_detail(self, response):
6 G" f( E; g' F; J! X6 B - Cater_Name = response.save
* E0 F' f8 S P, _' w1 F - # print Cater_Name! b7 l' \6 K& S
- Bookname = response.doc('h1').text()
2 @. ]4 V4 Y- X - print Bookname5 h2 d; _7 f% @' b3 x+ Z/ y
- Book_author = response.doc('.authorname > a').text()4 m! w" D" G2 B! J
- # print Book_author
$ T0 Z- B" {- C ^( H6 ?- s - Book_Introduction = response.doc('.book-intro > div').text()
, Z" I% b# j+ ?7 Z* v* ? - # print Book_Introduction
e, R) U6 L8 z( k: ~ - Book_Synopsis = response.doc('b').eq(1).text()
. O# s7 ?, x) t) |! ? - # print Book_Synopsis
7 U" |) Y. F& i% W, D/ o- [6 j - Book_Palabras = response.doc('.booktitle p').text().split(' ')[1].split('|')[0]
, `, }9 u) K% } - # print Book_Palabras
: z; I9 X8 R8 |/ h- o - BookIDs = response.url.split("xiaoshuo/")[-1].split("/")[0] #小说ID, X+ G w1 x- \) |0 s1 t
- # print BookIDs1 W N0 h* H9 c) R
- Book_Dates = str(datetime.datetime.now()) & I, `! z6 ~: K3 I! f
- for imgs in response.doc('.bigpic > img[src^="http"]').items():0 r* h2 ?: Z% \5 z& { @ ^
- img = imgs.attr.src
4 f& O0 ?0 I# F5 s( v - print img
7 s9 ?+ h0 e! a. v; @" Y6 v9 o2 h - #小说封面下载
" o; |+ H; I* _3 d' |- q- S - extension = self.getExtension(img)7 |5 H# s7 f C- _, A+ t8 ^' H
- name = self.getname(img)
P' N s, N: x& R2 R - file_name = name + "." + extension
3 @4 }' y' V `' V: a% s5 R - imgDir = P_dir + name
+ ?; d `6 h; ~, w& q% t - Locaimg = imgDir + "/" + file_name
5 ~9 n. s7 Z6 \ - print Locaimg) D" W! F2 X- u5 T
- if(self.download(P_dir, imgDir, file_name, img)): #这2行可注译,图片下载到本地& V3 L" R0 {8 d9 X# O
- print('attachment url is ' + img) #
$ `# s( S* h9 n1 | - Datos = {
9 M% U/ T( E$ F# p; `& ~ - "Cater_Name":Cater_Name,
, {- l3 z- u" |( N3 O. _ - "Book_author":Book_author,
: {5 p k: y& I) T' T - "Book_Introduction":Book_Introduction,
! x$ w0 J1 Y; S8 O - "Book_Synopsis":Book_Synopsis,
# y' q$ k5 b4 Z2 Z1 N- y - "Book_Palabras":Book_Palabras,0 r `) b/ l" B' P/ j4 r
- "img":img,# O* Q& {& M! K l$ W" D
- }
- ?3 ?* ~! |9 G - 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& A& U1 S0 j& `0 Q
- for each in response.doc('div[class="bookbtn-txt"] a[class="catalogbtn"]').items():5 s3 I4 V/ J# A |3 M
- self.crawl(each.attr.href, callback=self.index_page,save=Datos)' g, w/ V9 Z1 b
-
" [6 X3 M1 T8 M6 Z6 f3 L; i - @config(age=8 * 60 * 60)
o. w" Y! J# ^ b* T" l - def index_page(self, response):
- }9 z$ r+ w) X9 }5 L" l& V9 c7 _ - Datos = {) r0 r9 H3 ?( k$ I8 P0 I3 _; R
- "Cater_Name":response.save['Cater_Name'],; K1 ]4 l$ m5 a! m9 x
- "Book_author":response.save['Book_author'],# [# B: ~3 W: D$ w: X [
- "Book_Introduction":response.save['Book_Introduction'],# v* P& g( C( h7 Q9 E+ a3 F
- "Book_Synopsis":response.save['Book_Synopsis'],% K0 v; S2 Z: g( W" w
- "Book_Palabras":response.save['Book_Palabras'],3 c* o: z! n, N# _; f4 f. d
- "img":response.save['img'],
/ P9 R( R( ]; X - }
% ?) e. \$ W3 ]3 s$ W6 g" f1 [ - for each in response.doc('.chapter-list li:first-child a[href^="http"]').items():" p) @7 B' ^9 X; X/ v
- # for each in response.doc('.chapter-list a[href^="http"]').items(): ( T% q5 f4 D7 N) h1 x
- self.crawl(each.attr.href, callback=self.detail_page,save=Datos). Z: n. v1 }) z+ j6 |& F# `
- @config(priority=2)1 K1 J7 f- m) K: x4 P7 h
- @catch_status_code_error2 Y; o+ Q4 d! p6 ^
- def detail_page(self, response):
( D& |% a8 c) A( p- M8 P6 M - NewRe1 = u'哈书'& y8 ^6 I! V( C Z0 r J
- NewRe2 = u'huhjsd.CC'6 B6 v/ _& N, f7 o9 |6 E" `' h
- NewRe3 = r'^\\n\\n'
G- Q1 Y, ^1 S! y8 a7 J - NewRe5 = u'小说网'; f) Q, g- }* y* g6 b
- NewRe6 = u'fgdfgf'
1 X( X H* E& \7 C. y8 w - NewRe7 = u'fgfgf'
; G0 G/ _( b! J" c. ~7 J: _9 x - NewRe8 = u'ffhgf': I5 F7 G2 x l1 }2 x) o4 w
- NewRe4 = r'[\f\t\v+\.\{\(\)\}\!\/_,$%^*(+"\')]+|[+——()?【】“”!,。?、~@#¥%……&*()]+'9 q8 I9 x9 T% H' i* G8 Z
- ReC1 = u'静思'
2 B% f4 S3 w! j - ReC2 = u'aghgf.com'
/ R- P) q+ A; n( j$ i4 L6 P I - ReC3 = u'aghgfh.com'# s4 n: D8 T7 V" f1 e. {$ T
- ReC4 = u''9 q: a) p- U' l ^) V+ H
- ReC5 = u'文学网'
* Q6 p( [9 i$ M3 ?7 V0 j9 A - ReC6 = r'<BR>'" _# i2 L( v. h
- Bookname = response.doc('.readlocation a').eq(2).text() #小说名称
4 S, I5 r/ u% b; i. D, d - print Bookname
, x; a$ Y2 Q _5 Y - Cater_Name = response.save['Cater_Name'] # 小说分类8 G, o* E5 C5 q6 [. u P: C
- Book_author = response.save['Book_author'] #小说作者1 R9 q% D+ V1 L+ y: D$ O' h3 H
- Book_Introduction1 = response.save['Book_Introduction'] #小说简介
" K5 \) M& D3 d) i/ R - Book_Synopsis = response.save['Book_Synopsis'] #最近更新- \& h7 E9 B- K+ B
- Book_Palabras = response.save['Book_Palabras'] #小说字数; z1 @& f8 Z# T6 Z! z% u
- Bookurl = response.url #小说网址2 o6 G0 v1 N7 P$ M6 b
- Booktitle = response.doc('.article-title').text() #章节名称9 L- H. U' `! Y- x+ Y) ` r i
- BookID = response.doc('.readset-r span').text() #小说ID
1 w. k. r( V9 e# b% w5 M - BookConte1 = response.doc('.article-con').text() #小说章节内容- y0 V7 o" L2 y) {( ~6 H
- 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)
/ o5 ^0 V- x6 F - Book_Date = str(datetime.datetime.now()) # 采集时间 \' v" g! r+ Y" U& q* D
- BookConte2 = BookConte1.replace(NewRe1 , ReC1)
9 Q3 j- U& _6 B( C - BookConte3 = BookConte2.replace(NewRe2 , ReC2)8 a& \( R u0 q' u8 r$ q
- BookConte5 = BookConte3.replace(NewRe5 , ReC5)
* a6 A$ v* J: ^# @+ v; s7 V( s - BookConte6 = BookConte5.replace(NewRe6 , ReC2)
1 o! M& h( ?& W% q9 B! U, r! K* O - BookConte7 = BookConte6.replace(NewRe7 , ReC2)) @9 {9 j) O, P) g7 ^+ ]
- BookConte8 = BookConte7.replace(NewRe3 , ReC6)
7 [$ a+ p" S7 J; Q - BookConte4 = re.sub(NewRe4 , ReC4 , BookConte8)
6 q3 g* O" z: z! }& O - BookConte = BookConte4.replace("\n\n","<br>")/ x1 F4 A# A+ U4 t3 b6 u& S/ E+ F
- print BookConte
! o A/ N! }. Y Q - Book_Introduction2 = Book_Introduction1.replace(NewRe1 , ReC1)$ x0 E* O# ]* y9 q9 J1 ^& [
- Book_Introduction3 = Book_Introduction2.replace(NewRe2 , ReC2)
3 Y: w5 R+ j8 E9 P! C: p2 a - Book_Introduction4 = Book_Introduction3.replace(NewRe3 , ReC3)
7 a4 D4 b) ]0 o) }: e/ t - Book_Introduction = re.sub(NewRe4 , ReC4 , Book_Introduction4)
{5 P" h: y6 c - Titleid = response.url.split(BookID + "/")[-1].split("/")[0] ; [) b3 M# w: @) a. U0 O/ H
- Book_img = response.save['img'], #小说图片' \+ V" r% E$ p# p( [3 k3 Y
-
) x* L; d! V7 S N - #insert into MySQL 小说入库
' m3 R4 z5 p# w - 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
7 Z( q# H9 ]- S2 _# I$ y% D$ B9 r - 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
% b0 `4 J3 W G! _2 g2 V; ^ - #post提交发布
. L0 G' w U) |2 U: G - 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2 {7 x4 M% _. @: _4 } T
- Datos = {
3 F2 m* Q4 t/ r D - "Cater_Name":response.save['Cater_Name'],0 Z, P" }7 O& d# v- Y7 O+ b( H9 t
- "Book_author":response.save['Book_author'],8 [2 @6 Q, F/ S& B! n
- "Book_Introduction":response.save['Book_Introduction'],8 l. O' C6 I( z. }7 d
- "Book_Synopsis":response.save['Book_Synopsis'],8 e% C. i1 X1 ~* @4 Y ^
- "Book_Palabras":response.save['Book_Palabras'],
) B% a: w* z( w \8 W% ?" p! O - "img":response.save['img'],
. d/ b9 d2 y& a1 A0 [4 y; T0 h% @/ ?( _ - }, X5 e' _$ a$ m* {1 w ?; T( T
- for each in response.doc('.articlebtn > a:nth-child(4)[href*="/xiaoshuo"]').items():4 G% O$ b! ~ k
- self.crawl(each.attr.href, callback=self.detail_page,save=Datos)
; X u0 N5 R+ ~7 x U# n - return {
5 G* a: g' D3 u6 \7 K6 k - "Cater_Name":Cater_Name,- j7 g8 I: Q6 ~( d0 F" {# Z
- "Bookname":Bookname,
% ]+ Y3 K. Q- l - "Book_author":Book_author,
2 q- a0 Y- s$ O3 \8 l" q& K( ` - "Book_Introduction":Book_Introduction,
9 X) C1 N- s M/ }3 A# ` - "Book_Synopsis":Book_Synopsis,5 s9 I7 D5 M }) B" x
- "Book_Palabras":Book_Palabras,
$ i' g& d/ F+ ^ - "Book_img":Book_img,
! t3 S; n: J$ |- k, ~ - "Bookurl": response.url,$ J" h! R5 x1 r% ?& Y6 P
- "Booktitle": Booktitle,1 U/ }1 b& x6 e3 H3 D4 C/ x! m; O
- "BookID": BookID,& r4 X1 {( A8 Q9 h
- "BookConte": BookConte,
$ I5 U1 [: T4 ^7 f7 G" }2 g6 G - "Titleid": Titleid,+ V" W8 P: R- F+ m: Z% ~" v
- "abover":abover,% t. o, ]5 a. c# `6 {; Y
- # "Book_Date" = str(datetime.datetime.now()),
' G* y% I$ B# Q4 F1 A: y7 N8 ~( o$ G3 ` - }( Q* W/ A1 C! L% h7 a
- def download(self, P_dir, imgDir, file_name, Book_img): ?' s2 q5 R0 y4 M. ~: ?$ `
- if not os.path.exists(imgDir):
9 U7 j1 ^. Q' \6 Q) x* Y - os.makedirs(imgDir)
% o: G: x0 q& O3 h: o' ?: \+ ? - file = imgDir + "/" + file_name
. z l# c; G8 \1 Z) n d' u, E R - # print file
( {. S3 o5 @$ t2 |9 W5 ]2 T* s - f = open(file, 'wb+')
0 r& K. c' Y9 ]! o8 @/ t - imag = requests.get(Book_img)
8 @1 S r9 p; | G: Z% J# ]4 V - f.write(imag.content): E! s" o3 }2 }: }, |# L, a$ O
- f.close()
4 r; s) ^7 P K; L) z - #保存图片前
; d8 ]1 L- h m# I$ c* G0 |9 m - def save_imgs(self,response):1 Q, B d0 }/ i) D( A6 F
- content = response.content8 U$ S6 Y+ s a0 b1 b
- file_name = response.save["file_name"]
- p4 p- E; m+ h' j - imgDir = response.save["imgDir"]
, y% x6 ]; J) i5 Y# d/ {+ j - file_path = imgDir + file_name# m- d1 ^' b2 N- e; C; `) ]$ V
- self.save_img(content,imgDir,file_path)1 Q* P4 R& S F- r
- #保存图片/ O. z/ J8 k. U. p
- def save_img(self,content,imgDir,path):: }; Q. L7 Y, n+ ~, k8 J7 f' I
- if not os.path.exists(imgDir): 2 S/ |/ N0 \! _* a' w
- os.makedirs(imgDir)$ x( p0 W7 v, @: K8 |) M
- f = open(path,"wb" )- _7 W3 a( _% j2 `
- f.write(content). j( a( B. r9 f# ~, G7 S5 ~8 ]# @6 @
- f.close()
' _7 C. f; G5 Q! z( d - #获取url后缀名# X; [/ V6 }( v1 C6 n9 m: \
- def getExtension(self,url):
* {* I- N$ L' P' W& O: B0 Z0 y - extension = url.split(".")[-1]4 u! {9 m8 U$ z, k' l
- return extension
: i, P! \4 ]8 G% o: ]4 ~: W9 a% I -
; w' b$ n; x; Q4 u. b - #获取图片名. g2 l7 G5 v' ]. N3 |' {) J% p
- def getname(self,url):0 U$ x# @& K* ?2 @% y9 k
- name=url.split("/")[-1].split(".")[0]
; Q1 v0 e" F$ Z3 T* l& B - return name
复制代码
7 j/ c" R# X0 M& b - Q9 C, O! k& {8 u( M. |& t! H
|