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!
# L# Z: j$ c, m8 [5 P1 l4 @' |- #!/usr/bin/env python
1 b8 s% Z0 k5 P2 @# x - # -*- encoding: utf-8 -*-
% i+ K- ~: M% j% P* V. v0 p# r% Y - # Created on 2019-05-05 21:43:11
: O/ B0 a3 `* [ B+ W% y - # Project: XiaoShuo
# S ]6 y* X9 K- D9 t0 u, n -
% U3 n' Z/ j& d - from pyspider.libs.base_handler import *5 k% |! X) q" G7 P9 ~
- import pymysql
3 l( ]: }9 I7 x# d! {; L4 ~ - import random
* J$ [- J' g( ?' f6 M - import datetime8 Z4 C0 \% E; F' ~% y
- import urllib2,HTMLParser,re$ H9 O4 O& H; R% E t+ {
- import os
2 J0 Z" F% {; w# O* `: r - import sys
2 p' b+ ~: L3 I8 T# ^1 n- s% M4 c - import re, S* y+ l/ P: k5 s* q% ?0 m
- import codecs! l: ~; u/ M2 D8 Y* p, m: S
- import requests
2 M4 ^, B, o3 e2 K7 ^3 e; I - import json1 c# B1 i) R2 Z' I- l* A3 [- s
- - z, y7 v9 C1 X
- class Handler(BaseHandler):
, a. k; m, {" U5 V! T2 S - global Datos2 w1 l5 d1 ^( w, G8 f; K0 F- j6 J
- global P_dir
, |1 E" c9 X7 [+ D4 W2 p - P_dir = '/Tools/Debug/' #采集时候图片保持到本地的路径% N! v4 |" L0 n: d
- global Datos6 O! \0 A( o7 S; }' M5 A# [8 {5 c
- Datos = {}
0 U; |( h! @* `) p - headers= {
2 k* ?4 E3 _1 F3 z# g) o6 t m - 'Accept':'text/html,application/xhtml+xml,application/xml;q=0.9,image/webp,*/*;q=0.8',7 f. J+ i/ C; G! N. n( T; [
- 'Accept-Encoding':'gzip, deflate, sdch',
! Y: i: }& ]( \ - 'Accept-Language':'zh-CN,zh;q=0.8',2 B& [. d2 B6 e* h2 \" L
- 'Cache-Control':'max-age=0',& w K- a4 ^) H
- 'Connection':'keep-alive',
8 T) \ n7 L* }# 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'
; w, U8 y" F6 |8 D - }: u1 T, ?+ w$ O% M7 G1 i+ n
- crawl_config = {8 M" m ?1 G. O2 u6 g
- 'headers' : headers,: b3 O7 ?% j6 ]! `6 H8 X
- 'timeout' : 3006 X# l9 o* } R! p4 S6 b
- }5 ~" f4 }3 v$ u- F# d
- def add_BookFile(self,Bookname, BookIDs, img, Locaimg, Book_Dates):
7 V" {& z: ?$ C- L8 `4 u - db = pymysql.connect(host="localhost",user="数据库用户名",password="密码",db="数据库名",charset="utf8")
- e( m6 N" r0 ~; a8 t" q - try:
' G/ L6 m+ k9 E - cursor = db.cursor()
3 ?6 ]. _# @2 }/ F n. f; P+ L9 E- p - #注意此处字符串的占位符要加双引号"%s"3 F) X ?; P3 j9 I/ @" }: `( Y
- sql = 'insert into BookFile(Bookname, BookID, img, Locaimg, Book_Date) values ("%s","%s","%s","%s","%s")' % (Bookname, BookIDs, img, Locaimg, Book_Dates);
0 q: z2 F0 N* O- {9 \ - # print(sql)+ r& ] g2 r9 Y4 Y
- cursor.execute(sql)
$ ?) T3 Z1 L9 |# ]+ O - " F$ a. _ d2 O5 P4 O
- #qid = cursor.lastrowid
W) R5 z- Z- z9 ~) ?8 B2 ^ - #print(qid)
- U: h$ S4 ~( w5 w( ] -
( A' U5 k- Q2 v/ n - db.commit()
: H$ Q4 n) @7 c, ?' S' @ - except Exception as err:. b& V2 A, g% |! {# z* |
- print("Error %s for execute sql: %s" % (err, sql))
( d8 e. x5 f5 h) ~1 H - db.rollback() D; M4 {3 ~$ s. | @, Q5 _
- def add_comment(self,Bookname, Booktitle, BookID, Titleid, Book_Date):
9 i3 Y1 H, `& Z, R' J - db = pymysql.connect(host="localhost",user="数据库用户名",password="密码",db="数据库名",charset="utf8")# m9 c" D3 {* f3 P4 c
- try:6 O9 U4 s5 A( [: H
- cursor = db.cursor() c% i& E2 R! ^3 [5 h! q8 C! w
- #注意此处字符串的占位符要加双引号"%s"
! k' O( k6 U5 d- j - sql = 'insert into BookTitle(Bookname, Booktitle, BookID, Titleid, Book_Date) values ("%s","%s","%s","%s","%s")' % (Bookname, Booktitle, BookID, Titleid, Book_Date);
/ D7 i6 w' R! i% M( U - # print(sql)2 K3 B& {/ l" R
- cursor.execute(sql): N* U# J; u1 N0 j( o& V6 [" A- |% O1 ~3 n
- " ]* {$ F3 X1 l
- #qid = cursor.lastrowid
8 C* `6 t! G7 R* S - #print(qid)# L/ }, m) m5 e. h
-
" w# g- t" M) m, y- V" L - db.commit(): c2 p" O* u2 u* I0 t
- except Exception as err:
) Q6 Z4 A& l- o$ i5 ^' |! I# i - print("Error %s for execute sql: %s" % (err, sql))/ U: l/ o; y! o0 O8 m
- db.rollback()
p5 F1 X0 C! V - def add_question(self,Bookname,Cater_Name,Book_author,Book_Introduction,Book_Synopsis,Book_Palabras,Bookurl,Booktitle,BookID,BookConte,Titleid,abover,Book_Date):
% f7 n! r, E0 }- w3 z0 F5 J; | - db = pymysql.connect(host="localhost",user="数据库用户名",password="密码",db="数据库名",charset="utf8")
; B% u7 D8 S" g% f - try:+ p& q' F4 t2 s! j& r4 E; K1 d
- cursor = db.cursor()
# k s4 i6 r) |4 K# M& m - #注意此处字符串的占位符要加双引号"%s"
, i' v( S. K) \ - 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);6 a% O- U1 o! u; X2 [
- print(sql)9 z9 {& F: [9 c; f
- cursor.execute(sql)
9 m& Q$ @! c2 w V, s - print(cursor.lastrowid)
- w7 V9 s7 o* W- n - db.commit()
& \9 C: m1 D0 a+ i5 e7 t z$ i: u - except Exception as err:
5 e2 v5 ~/ Q$ Z+ `* P: ^ - # except:
6 h5 x8 i; Q2 }# w: _ - # print('Failed')
; O7 M8 O& `; X& { - print("Error %s for execute sql: %s" % (err, sql))8 F' l4 x# I1 a+ b" i3 ~
- db.rollback()& v) N" a t/ I& o
- ( l# k' E5 @, @8 V4 J
- def add_locoy(self,Bookname,Cater_Name,Book_author,Book_Introduction,Book_Palabras,Book_img,Booktitle,BookConte,abover): 4 C+ f8 `0 w5 l, j/ ~/ C5 F: L
- reload(sys)
$ s5 \# Z' o( k/ M9 D - sys.setdefaultencoding("gbk")6 t* ]% @/ G( {; Q$ {
- locoy_url = 'http://www.******.net/locoy/?my=book' #697火车头发接口地址3 n9 T9 N1 J& l1 g+ l" M7 n0 @
- locoy_data = {! O& O; R$ w4 x
- 'my_u':'用户名', #后台用户名
6 \ }5 Y9 Z* ^! F, ^ - 'my_p':'密码', #后台密码( c# W3 z- y1 @5 G; X- \
- 'subject_669977_net':Bookname.encode('gbk', 'ignore'),3 h6 U) H/ _7 W7 R# C6 v1 A
- 'caid':Cater_Name.encode('gbk', 'ignore'),
" v* z- T% j3 m5 S& n - 'title_669977_net':Booktitle.encode('gbk', 'ignore'),8 o; `& K' q8 K* z, v
- 'article':BookConte.encode('gbk', 'ignore'),1 H8 P$ n, _9 j$ T3 n% e& I2 \
- 'author':Book_author.encode('gbk', 'ignore'),3 N J, |! h+ {* r8 p
- 'ready_1':Book_Palabras.encode('gbk', 'ignore'),
" H! I5 }& D' W4 M* M/ X5 r1 D - 'thumb':Book_img,
, ^, C( P; o& s* A - 'content':Book_Introduction.encode('gbk', 'ignore'),. ]) d/ H' w& D, S
- 'abover':abover.encode('gbk', 'ignore') , W0 }# J# E6 W" s( d- p4 g# h7 Q, [
- }
2 [; o" p8 [5 s' G - res = requests.post(locoy_url, data=locoy_data)
5 s) g! W2 z5 t9 S' p( Z3 ^) l - print res.text
4 J1 _7 [" ? O6 n. V- L - print res.content
, p! e7 e; b. C( j& s N - # print Dsd( Q& |4 G7 R7 [- {" T- c' y
- return res: P% {1 F4 f- A3 F' p
-
I8 }3 \5 K0 J+ W! V - def __init__(self):6 U8 z# g4 D d' I6 @
- self.base_url1 = 'https://www.****.cc/'
1 p& k$ m7 R! w$ z - self.base_url2 = '/'0 Z4 A- t: H2 w* ~$ e9 {7 i# G8 L
- self.CaterId = []+ \6 M. w% k6 v5 Q
- self.CaterIds = ['xuanhuan', 'wuxia', 'yanqing', 'lishi', 'kehuan', 'kongbu', 'nvsheng']
0 z$ T6 X$ {( x2 t: I8 `9 c; w - self.page_num = 1
' N. K' q* n) o7 h. l- \" [% X! f - self.total_num = 200
2 g$ W( Y1 I& s$ f - 9 c" I @7 _/ j; Y
- @every(minutes=8 * 60) Z& n& p7 ^3 _, J$ c
- def on_start(self):
: C& { |! L8 V+ M# M7 z0 \ - global Cater_Name
3 K" ~6 O. c% A% S( j. l - Cater_Name = []
" s6 u% K0 e& Z" {' I) _ - while self.page_num <= self.total_num:
" B" d7 U3 e2 ]& H - for self.CaterId in self.CaterIds:
" V- i4 K) d0 F) O- m - if self.CaterId == 'xuanhuan':- n3 _0 f# ^2 d* B% y9 I
- Cater_Name = '玄幻'( c0 ^# L2 [8 Q& D
- if self.CaterId == 'wuxia':
. q2 {" m! w" Y1 d8 C) d7 E/ e - Cater_Name = '武侠'
5 J1 H* n: t, q/ t1 R! w' S& O - if self.CaterId == 'lishi':% Z0 Y* H t, K0 v' h3 E# G1 i
- Cater_Name = '历史'
) x! G; k- x+ Q( Y( q' f0 X- b - if self.CaterId == 'yanqing':( `) E6 L$ y5 D
- Cater_Name = '都市'
( {. |0 }# i! z - if self.CaterId == 'nvsheng':% f7 O1 H4 ]& a# o: M9 E7 W: w- \
- Cater_Name = '都市'
+ @# K* U( }( }# b7 ]0 t' x1 e - if self.CaterId == 'kehuan':
1 R; t4 d$ b8 ]3 @ - Cater_Name = '科幻' / L, M! j9 v. F6 B% }! S7 y+ ]% A
- if self.CaterId == 'kongbu':
5 F% b `8 q2 W0 |3 n6 \ - Cater_Name = '游戏' . z8 Y2 f/ H0 e3 k
- print self.CaterId
$ k X( l# @& v - url = self.base_url1 + str(self.CaterId) + self.base_url2 + str(self.page_num) + "/"
7 z4 J; b/ o5 {+ L8 i, \& C: c! ^, Y - self.crawl(url, callback=self.list_Caterg,save=Cater_Name)" T; b! }0 E$ A- x
- self.page_num += 1 % H P2 a; D2 y/ \9 E
- , D% O6 {* L3 v5 A
- def list_Caterg(self, response):
1 G/ [! D% P0 b - Cater_Name = response.save
6 @" {' _# R) V$ o5 v - for each in response.doc('.pic-list a[href^="http"]').items():
9 w* b% f' r& I# e7 e, Z - self.crawl(each.attr.href, callback=self.list_Caterg_detail,save=Cater_Name)
; [) w9 M1 A5 ^2 c1 r - 9 p" l9 G) O# i! u: @% R; {
- def list_Caterg_detail(self, response):5 i0 y+ a2 g5 n7 G% F I
- Cater_Name = response.save! j& p4 B8 c1 \, N( M3 F- v' ]
- # print Cater_Name1 I- ^0 r" A, p
- Bookname = response.doc('h1').text()
+ k M% m9 d. @3 P4 ? - print Bookname7 Q& D) W B/ Q6 e# g
- Book_author = response.doc('.authorname > a').text()
& V: h6 V9 D2 G g I - # print Book_author
k2 K+ t5 [/ n/ v2 T - Book_Introduction = response.doc('.book-intro > div').text()6 ^4 O. |7 T; S4 G% D
- # print Book_Introduction# d, [7 s x$ ^/ i, S7 Y! ]$ I9 b( D
- Book_Synopsis = response.doc('b').eq(1).text()8 Y9 f4 {: N* d1 }9 Y2 r- V
- # print Book_Synopsis
+ V% X. z- j' E) a - Book_Palabras = response.doc('.booktitle p').text().split(' ')[1].split('|')[0]
. Q. u6 ?& B5 V/ n - # print Book_Palabras
& b" A9 `, D( |# H) g; c - BookIDs = response.url.split("xiaoshuo/")[-1].split("/")[0] #小说ID
4 J6 w& L; R/ `0 V - # print BookIDs
) h/ z% g' K8 ^: I7 G3 d( V) s q - Book_Dates = str(datetime.datetime.now()) - M9 ]- S* M6 ^5 D9 M- w
- for imgs in response.doc('.bigpic > img[src^="http"]').items():
: K- P7 a+ v/ `" `7 N5 K' d& e - img = imgs.attr.src6 X3 n! ?" Y. P6 n$ [$ ?& T
- print img2 X8 G6 \! [* O4 D
- #小说封面下载
6 Z" q8 e( ~, l - extension = self.getExtension(img)" ?7 ]8 X- ?6 D! o- N" U
- name = self.getname(img)
) {6 u- _/ s& q9 c0 C4 n: G - file_name = name + "." + extension" Z& N3 C5 E3 A% c k
- imgDir = P_dir + name/ ]/ c, l9 A3 S
- Locaimg = imgDir + "/" + file_name
; k' S* c8 [3 `- C! |) O' ^ - print Locaimg
& E( b$ e0 {4 L6 T* ` - if(self.download(P_dir, imgDir, file_name, img)): #这2行可注译,图片下载到本地& u; Y: b4 p6 s9 s- ?3 T
- print('attachment url is ' + img) #
! { F5 {/ c/ x- a - Datos = {
6 l$ e. s) e& W$ s; c - "Cater_Name":Cater_Name,7 ?# }4 X2 R& ^8 l/ u. i) v
- "Book_author":Book_author,3 L5 f) O' M% `1 `* T
- "Book_Introduction":Book_Introduction,
% T! x& t, m+ F8 ^6 r. e+ x) f - "Book_Synopsis":Book_Synopsis,
* u: D4 N0 W4 z2 G) Y! T* f( K - "Book_Palabras":Book_Palabras,
% e/ h8 [; k2 b0 {; ~: d - "img":img,
% e9 R8 g1 j8 b# x. I- g3 g8 Q - }# Q' |& ]: f5 T- H3 A
- 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4 P6 \ f+ X* M% M: _, e( C9 ?& [
- for each in response.doc('div[class="bookbtn-txt"] a[class="catalogbtn"]').items():
+ W% M/ x$ ~) P% u, E8 q - self.crawl(each.attr.href, callback=self.index_page,save=Datos)
J1 y3 V' x5 ?( m8 B# J -
& l- T6 b& o& B. ^& c - @config(age=8 * 60 * 60)
) d% E( E7 S8 w2 f+ g& P+ ~- @5 m0 V! e - def index_page(self, response):
* h9 d L; h6 {# j6 S - Datos = {
( r. p4 j# }2 ~- B - "Cater_Name":response.save['Cater_Name'],. ?; h, r. X1 d2 O8 ]5 s: A
- "Book_author":response.save['Book_author'],
4 Y9 F8 `2 e! s+ Y - "Book_Introduction":response.save['Book_Introduction'],, P) v0 H9 j" Y) e
- "Book_Synopsis":response.save['Book_Synopsis'],
0 _4 h8 r4 y$ Q. F2 T- o7 @" @' k' x - "Book_Palabras":response.save['Book_Palabras'],6 a) {+ N0 d6 `( R
- "img":response.save['img'],
5 J" T4 x h6 X2 H, I$ Y - }
4 b2 A# P9 Y/ ?5 i& K4 s) m - for each in response.doc('.chapter-list li:first-child a[href^="http"]').items():
) L2 t4 K- g3 c - # for each in response.doc('.chapter-list a[href^="http"]').items(): 7 y( ^4 U1 X" x) }' v3 n
- self.crawl(each.attr.href, callback=self.detail_page,save=Datos)9 _* y5 [2 ~: }2 k+ Y. |
- @config(priority=2)6 k( V/ y# s. G; S" @
- @catch_status_code_error I4 m0 k+ Q9 [+ T$ ^* K
- def detail_page(self, response):
, B3 R" N7 X, c6 e - NewRe1 = u'哈书'# P7 D$ Y4 `: ~- I: a5 O; z6 a
- NewRe2 = u'huhjsd.CC'
, h/ v! Q- Q& B' g6 t - NewRe3 = r'^\\n\\n'0 I# {+ P: [ B% _! |
- NewRe5 = u'小说网'
0 V/ }% w2 k" V- U0 s: n - NewRe6 = u'fgdfgf'8 o. [3 [% O: L0 D' L! K
- NewRe7 = u'fgfgf'
6 F! | N) S& o' ]& d3 Y - NewRe8 = u'ffhgf'% ?3 V. i4 C9 d) H* a
- NewRe4 = r'[\f\t\v+\.\{\(\)\}\!\/_,$%^*(+"\')]+|[+——()?【】“”!,。?、~@#¥%……&*()]+'( _# W, g- b# j& ]* Z
- ReC1 = u'静思'- V" q# N, W, H4 C/ f
- ReC2 = u'aghgf.com'3 A# _. n( d4 }; x9 c" Z
- ReC3 = u'aghgfh.com'
. ^1 f) `2 ?+ n" }$ w* Z - ReC4 = u''! R4 {( c/ e: l
- ReC5 = u'文学网'. Y! h% V. v! f
- ReC6 = r'<BR>'
$ \# x! F3 Z4 R - Bookname = response.doc('.readlocation a').eq(2).text() #小说名称9 B* `( H7 \3 L9 D. m
- print Bookname
. e4 }( E7 O6 @' O0 I - Cater_Name = response.save['Cater_Name'] # 小说分类
- |& A+ |1 B; [; w$ D - Book_author = response.save['Book_author'] #小说作者
, U6 H+ ^+ d- ]* e' t; |' o4 C - Book_Introduction1 = response.save['Book_Introduction'] #小说简介& }* R# K. i# n8 i, i% v9 d9 ?
- Book_Synopsis = response.save['Book_Synopsis'] #最近更新
7 _" O9 U6 C+ s( D9 _( S% J0 T9 ^/ d - Book_Palabras = response.save['Book_Palabras'] #小说字数' i0 @+ d# q7 K
- Bookurl = response.url #小说网址9 m, c9 I2 y$ L4 b5 w0 `
- Booktitle = response.doc('.article-title').text() #章节名称# ^, V/ F2 L* j/ y m5 r8 X: L5 m9 ^" Q3 ?6 l
- BookID = response.doc('.readset-r span').text() #小说ID. o+ E" T8 u7 }6 M# X5 k
- BookConte1 = response.doc('.article-con').text() #小说章节内容
3 J# U' Q' p, ?% p$ ~ g - 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)9 M: z* n% E3 w+ s8 a' m
- Book_Date = str(datetime.datetime.now()) # 采集时间# `& y H& o& c6 T. f' A/ ?
- BookConte2 = BookConte1.replace(NewRe1 , ReC1)
& v0 I3 J/ h4 x - BookConte3 = BookConte2.replace(NewRe2 , ReC2)9 A8 C/ h5 Y) ?" A
- BookConte5 = BookConte3.replace(NewRe5 , ReC5)2 C4 B% b" z: Z1 E; ?. t
- BookConte6 = BookConte5.replace(NewRe6 , ReC2)
7 D# o4 R. V* y4 q, r, X4 X. J2 x - BookConte7 = BookConte6.replace(NewRe7 , ReC2)& I$ P1 w; O3 A& N' F0 L2 G
- BookConte8 = BookConte7.replace(NewRe3 , ReC6)5 f/ g& X+ q' \3 [1 [5 t+ d
- BookConte4 = re.sub(NewRe4 , ReC4 , BookConte8)' w; R1 `- z* E1 W. M
- BookConte = BookConte4.replace("\n\n","<br>"). L9 Z8 [1 i4 F
- print BookConte* h, H6 N$ A* H9 D2 F
- Book_Introduction2 = Book_Introduction1.replace(NewRe1 , ReC1)
; |* J+ T- j$ P; l9 x' l - Book_Introduction3 = Book_Introduction2.replace(NewRe2 , ReC2)) @" p X4 X* p7 `
- Book_Introduction4 = Book_Introduction3.replace(NewRe3 , ReC3), X" r8 A [, M2 ~/ ~4 t- t
- Book_Introduction = re.sub(NewRe4 , ReC4 , Book_Introduction4)
) X: A5 D+ Q. B$ o5 l0 X - Titleid = response.url.split(BookID + "/")[-1].split("/")[0] $ z* }- h1 s% u7 J- a
- Book_img = response.save['img'], #小说图片
% q/ G1 @1 ?1 d& S; a1 v -
2 E( l4 B. Y0 o* I! z* b - #insert into MySQL 小说入库8 r4 l$ J3 U" o+ {; F$ V
- 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
* B, L0 N: n% P8 u! X - 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. c1 E/ `% N. F6 y4 G& v
- #post提交发布
0 l' x6 u T+ U) r1 k& x - 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! H* x+ B$ E# g
- Datos = {
N$ z, a) }# S# z - "Cater_Name":response.save['Cater_Name'],
' N7 M, I; o$ |4 C - "Book_author":response.save['Book_author'],8 l8 Z6 A6 l, S8 ?8 ^
- "Book_Introduction":response.save['Book_Introduction'],4 S* D& v% I% z9 p1 p. ]$ Z( P
- "Book_Synopsis":response.save['Book_Synopsis'],# s8 a0 l0 l- P. L! t
- "Book_Palabras":response.save['Book_Palabras'],
; T# c% \5 H: o: X - "img":response.save['img'],5 a* i7 v- l9 ^* l" o5 m& ]
- }, V/ o9 t2 [3 t
- for each in response.doc('.articlebtn > a:nth-child(4)[href*="/xiaoshuo"]').items():
5 J0 [ Y8 n0 u5 E - self.crawl(each.attr.href, callback=self.detail_page,save=Datos)
}1 `- \# W2 X9 G: P; f - return {
3 g, p- ^! V$ i; a4 k1 W - "Cater_Name":Cater_Name,
+ W, O* w0 q: G* t0 o- s - "Bookname":Bookname,
/ e5 Y2 Y4 m4 \ - "Book_author":Book_author,8 |$ L0 g5 r6 r" D9 o4 g! ]- C6 Y
- "Book_Introduction":Book_Introduction,# @# H* k% ?/ _
- "Book_Synopsis":Book_Synopsis,) L7 W; _; M) I
- "Book_Palabras":Book_Palabras,6 ]1 ?0 U& w/ c9 }4 ^
- "Book_img":Book_img,5 H' d/ O" W b" ?0 M9 v% U
- "Bookurl": response.url,* x% g1 X+ Y; s4 c
- "Booktitle": Booktitle,5 u Z6 ^6 w+ Y( ?# _. g
- "BookID": BookID,9 R/ g2 m: S- i( T
- "BookConte": BookConte,) s4 L b! N3 J
- "Titleid": Titleid,
# F& ^4 ~/ _; I8 c/ X1 I - "abover":abover,) }/ C: L, B& Z7 z6 B6 r' s
- # "Book_Date" = str(datetime.datetime.now()),; h8 Q+ z2 J Z4 D, v" J( N7 _
- }- a, @! J) P! j( d
- def download(self, P_dir, imgDir, file_name, Book_img):( a% } j4 Z e
- if not os.path.exists(imgDir): ; A. i4 C) D# T
- os.makedirs(imgDir)
! U! O$ K9 @+ B. T - file = imgDir + "/" + file_name
# L, b' o+ w: b) }) m: q - # print file9 {1 \) t; d' e Y$ \
- f = open(file, 'wb+')
; U9 o( ^5 r4 U* g - imag = requests.get(Book_img) 9 T9 {4 c) t0 H- Y/ W2 i
- f.write(imag.content)( K, ?# U- \* m: L+ u" V) o( b9 V3 t
- f.close()7 C P8 T5 {+ }0 _
- #保存图片前! `# m! E$ R. s. _7 b# K) v7 N5 {
- def save_imgs(self,response):4 ]0 ~5 C& w- y% C+ ^. Q7 o1 s
- content = response.content8 R( D, i0 K# N N' x( l( P* O+ B
- file_name = response.save["file_name"]8 O; n4 P. _0 g% k2 U4 b
- imgDir = response.save["imgDir"]
% `# K) q0 ?, u1 C$ y1 j: } - file_path = imgDir + file_name+ d6 ~: s) k; P9 h# `+ {
- self.save_img(content,imgDir,file_path)
* O+ u5 D$ ~" N6 P2 N+ y - #保存图片
! s9 ]2 N `+ |; y2 j3 p, z - def save_img(self,content,imgDir,path):
6 c- l2 a( A9 i2 ^- g+ ~+ A - if not os.path.exists(imgDir):
4 x4 W- Q8 G; Q0 b, L- q& K7 P - os.makedirs(imgDir) f' O2 k& o( X! k
- f = open(path,"wb" ). [- c9 J2 F% A* W/ v6 a
- f.write(content)
+ t( R8 f% s8 s' p, I - f.close()
% c$ l) B* M8 z% s - #获取url后缀名
, ~ | q, b3 p! X - def getExtension(self,url):
/ A' i. H1 C) b5 T | - extension = url.split(".")[-1]) q! ]# h4 j- U
- return extension $ M% k' @$ C$ k7 n, ?
-
" |, N% ^- C7 {' l; P3 ^ - #获取图片名( _, Y" ]" |! J7 K5 X
- def getname(self,url):
) |$ R7 \; x! a - name=url.split("/")[-1].split(".")[0]( J* ?- E* D1 [- E
- return name
复制代码
@& n- g" B( i8 Z/ n! r ' m. {1 j% I3 L3 ?* H1 j+ D
|