使用正则表达式抽取新闻/BBS网页发表时间(修改版)
2012-03-15 10:50
369 查看
import java.text.ParseException;
import java.text.SimpleDateFormat;
import java.util.Date;
import java.util.regex.Matcher;
import java.util.regex.Pattern;
/**
* 分析时间戳
*
* @author xum
*
*/
public class PublishTimeExtract {
private static final String TIME_REGEX = "((:|>|\\s)?20[0-9]{2}(-|/|\\.|\\u5e74)\\d{1,2}(-|/|\\.|\\u6708)\\d{1,2}(\\u65e5)?\\s?\\d{1,2}(:|\\u65f6)\\d{2}((:|\\u5206)\\d{2})?|(:|>|\\s)?[0-9]{2}\\u5e74\\d{1,2}\\u6708\\d{1,2}(\\u65e5)?\\s?\\d{1,2}(:|\\u65f6)\\d{2}((:|\\u5206)\\d{2})?)";
private static final String TIME_REGEX_1 = "((:|>|\\s)?20[0-9]{2}(-|/|\\.|\\u5e74)\\d{1,2}(-|/|\\.|\\u6708)\\d{1,2}(\\u65e5)?(\\s| |<)+|(:|>|\\s)?[0-9]{2}\\u5e74\\d{1,2}\\u6708\\d{1,2}(\\u65e5)?(\\s| |<)+)";
private static Pattern pattern = Pattern.compile(TIME_REGEX);
private static Pattern pattern1 = Pattern.compile(TIME_REGEX_1);
private static SimpleDateFormat sdf = new SimpleDateFormat(
"yyyy-MM-dd HH:mm:ss");
private static final String BBS_URL = "(http://bbs/\\..*|http://www\\.tianya\\.cn/[a-zA-Z]*forum/content/.*)";
/**
* @param content
* @param url
* @return
*/
public static String extractDate(String content, String url) {
String con = content
.replaceAll("<!--.*?-->", "");
Matcher m = pattern.matcher(con);
Date now = new Date();
//
if (url.matches(BBS_URL)) {
String dateStr = null;
Date date = null;
while (m.find()) {
dateStr = m.group();
if (dateStr == null)
continue;
dateStr = dateStr.trim().replaceAll(">", "");
if (dateStr.startsWith(":")) {
dateStr = dateStr.replaceFirst(":", "");
}
dateStr = dateStr.replaceAll("\\.|/|\\u5e74|\\u6708", "-")
.replaceAll("\\u65e5", " ");
dateStr = dateStr.replaceAll("\\u65f6|\\u5206", ":");
Date tempDate;
try {
tempDate = sdf.parse(changeString2Date(dateStr));
if (tempDate.after(now)) {
continue;
}
} catch (ParseException e) {
continue;
}
if (date == null) {
date = tempDate;
} else if (tempDate.after(date)) {
date = tempDate;
}
}
if (date != null) {
return date.getTime() + "";
}
} else {
String dateStr = null;
if (m.find()) {
dateStr = m.group();
}
if (dateStr != null) {
dateStr = dateStr.trim().replaceAll(">", "");
if (dateStr.startsWith(":")) {
dateStr = dateStr.replaceFirst(":", "");
}
dateStr = dateStr.replaceAll("\\.|/|\\u5e74|\\u6708", "-")
.replaceAll("\\u65e5", " ");
;
dateStr = dateStr.replaceAll("\\u65f6|\\u5206", ":");
try {
return (sdf.parse(changeString2Date(dateStr)).getTime())
+ "";
} catch (ParseException e) {
System.out
.println("=======================>parsedate error: "
+ e.getMessage());
return (new Date()).getTime() + "";
}
}
}
m = pattern1.matcher(con);
if (url.matches(BBS_URL)) {
String dateStr = null;
Date date = null;
while (m.find()) {
dateStr = m.group();
if (dateStr == null)
continue;
dateStr = dateStr.trim().replaceAll(">", "");
if (dateStr.startsWith(":")) {
dateStr = dateStr.replaceFirst(":", "");
}
dateStr = dateStr.replaceAll("\\.|/|\\u5e74|\\u6708", "-")
.replaceAll("\\u65e5", " ");
dateStr = dateStr.replaceAll("( |<)", "");
Date tempDate;
try {
tempDate = sdf.parse(changeString2Date(dateStr));
if (tempDate.after(now)) {
continue;
}
} catch (ParseException e) {
continue;
}
if (date == null) {
date = tempDate;
} else if (tempDate.after(date)) {
date = tempDate;
}
}
if (date != null) {
return date.getTime() + "";
}
} else {
String dateStr = null;
if (m.find()) {
dateStr = m.group();
}
if (dateStr != null) {
dateStr = dateStr.trim().replaceAll(">", "");
if (dateStr.startsWith(":")) {
dateStr = dateStr.replaceFirst(":", "");
}
dateStr = dateStr.replaceAll("\\.|/|\\u5e74|\\u6708", "-")
.replaceAll("\\u65e5", " ");
;
dateStr = dateStr.replaceAll("( |<)", "");
try {
return (sdf.parse(changeString2Date(dateStr)).getTime())
+ "";
} catch (ParseException e) {
System.out
.println("=======================>parsedate error: "
+ e.getMessage());
return (new Date()).getTime() + "";
}
}
}
return (new Date()).getTime() + "";
}
private static String changeString2Date(String dateStr) {
System.out.println("=======================>Pre Date: " + dateStr);
String date = "";
String[] tempDateStr = dateStr.trim().replaceAll("\\s+", "@")
.replaceAll("(-|:)", "@").split("@");
if (tempDateStr.length > 0) {
String year = tempDateStr[0];
if (year.length() == 2) {
year = "20" + year;
}
date = date + year + "-";
}
if (tempDateStr.length > 1) {
date = date
+ (tempDateStr[1].length() > 1 ? tempDateStr[1] : "0"
+ tempDateStr[1]) + "-";
}
if (tempDateStr.length > 2) {
date = date
+ (tempDateStr[2].length() > 1 ? tempDateStr[2] : "0"
+ tempDateStr[2]) + " ";
}
if (tempDateStr.length > 3) {
date = date
+ (tempDateStr[3].length() > 1 ? tempDateStr[3] : "0"
+ tempDateStr[3]) + ":";
} else {
date = date + "00:";
}
if (tempDateStr.length > 4) {
date = date
+ (tempDateStr[4].length() > 1 ? tempDateStr[4] : "0"
+ tempDateStr[4]) + ":";
} else {
date = date + "00:";
}
if (tempDateStr.length > 5) {
date = date
+ (tempDateStr[5].length() > 1 ? tempDateStr[5] : "0"
+ tempDateStr[5]);
} else {
date = date + "00";
}
System.out.println("=======================> Date: " + date);
return date;
}
public static void main(String[] args) {
extractDate(
" <td align=\"center\">http://www.rednet.cn <!--时间开始-->2012/3/2 9:14:36<!--时间结束--> ",
"");
}
import java.text.SimpleDateFormat;
import java.util.Date;
import java.util.regex.Matcher;
import java.util.regex.Pattern;
/**
* 分析时间戳
*
* @author xum
*
*/
public class PublishTimeExtract {
private static final String TIME_REGEX = "((:|>|\\s)?20[0-9]{2}(-|/|\\.|\\u5e74)\\d{1,2}(-|/|\\.|\\u6708)\\d{1,2}(\\u65e5)?\\s?\\d{1,2}(:|\\u65f6)\\d{2}((:|\\u5206)\\d{2})?|(:|>|\\s)?[0-9]{2}\\u5e74\\d{1,2}\\u6708\\d{1,2}(\\u65e5)?\\s?\\d{1,2}(:|\\u65f6)\\d{2}((:|\\u5206)\\d{2})?)";
private static final String TIME_REGEX_1 = "((:|>|\\s)?20[0-9]{2}(-|/|\\.|\\u5e74)\\d{1,2}(-|/|\\.|\\u6708)\\d{1,2}(\\u65e5)?(\\s| |<)+|(:|>|\\s)?[0-9]{2}\\u5e74\\d{1,2}\\u6708\\d{1,2}(\\u65e5)?(\\s| |<)+)";
private static Pattern pattern = Pattern.compile(TIME_REGEX);
private static Pattern pattern1 = Pattern.compile(TIME_REGEX_1);
private static SimpleDateFormat sdf = new SimpleDateFormat(
"yyyy-MM-dd HH:mm:ss");
private static final String BBS_URL = "(http://bbs/\\..*|http://www\\.tianya\\.cn/[a-zA-Z]*forum/content/.*)";
/**
* @param content
* @param url
* @return
*/
public static String extractDate(String content, String url) {
String con = content
.replaceAll("<!--.*?-->", "");
Matcher m = pattern.matcher(con);
Date now = new Date();
//
if (url.matches(BBS_URL)) {
String dateStr = null;
Date date = null;
while (m.find()) {
dateStr = m.group();
if (dateStr == null)
continue;
dateStr = dateStr.trim().replaceAll(">", "");
if (dateStr.startsWith(":")) {
dateStr = dateStr.replaceFirst(":", "");
}
dateStr = dateStr.replaceAll("\\.|/|\\u5e74|\\u6708", "-")
.replaceAll("\\u65e5", " ");
dateStr = dateStr.replaceAll("\\u65f6|\\u5206", ":");
Date tempDate;
try {
tempDate = sdf.parse(changeString2Date(dateStr));
if (tempDate.after(now)) {
continue;
}
} catch (ParseException e) {
continue;
}
if (date == null) {
date = tempDate;
} else if (tempDate.after(date)) {
date = tempDate;
}
}
if (date != null) {
return date.getTime() + "";
}
} else {
String dateStr = null;
if (m.find()) {
dateStr = m.group();
}
if (dateStr != null) {
dateStr = dateStr.trim().replaceAll(">", "");
if (dateStr.startsWith(":")) {
dateStr = dateStr.replaceFirst(":", "");
}
dateStr = dateStr.replaceAll("\\.|/|\\u5e74|\\u6708", "-")
.replaceAll("\\u65e5", " ");
;
dateStr = dateStr.replaceAll("\\u65f6|\\u5206", ":");
try {
return (sdf.parse(changeString2Date(dateStr)).getTime())
+ "";
} catch (ParseException e) {
System.out
.println("=======================>parsedate error: "
+ e.getMessage());
return (new Date()).getTime() + "";
}
}
}
m = pattern1.matcher(con);
if (url.matches(BBS_URL)) {
String dateStr = null;
Date date = null;
while (m.find()) {
dateStr = m.group();
if (dateStr == null)
continue;
dateStr = dateStr.trim().replaceAll(">", "");
if (dateStr.startsWith(":")) {
dateStr = dateStr.replaceFirst(":", "");
}
dateStr = dateStr.replaceAll("\\.|/|\\u5e74|\\u6708", "-")
.replaceAll("\\u65e5", " ");
dateStr = dateStr.replaceAll("( |<)", "");
Date tempDate;
try {
tempDate = sdf.parse(changeString2Date(dateStr));
if (tempDate.after(now)) {
continue;
}
} catch (ParseException e) {
continue;
}
if (date == null) {
date = tempDate;
} else if (tempDate.after(date)) {
date = tempDate;
}
}
if (date != null) {
return date.getTime() + "";
}
} else {
String dateStr = null;
if (m.find()) {
dateStr = m.group();
}
if (dateStr != null) {
dateStr = dateStr.trim().replaceAll(">", "");
if (dateStr.startsWith(":")) {
dateStr = dateStr.replaceFirst(":", "");
}
dateStr = dateStr.replaceAll("\\.|/|\\u5e74|\\u6708", "-")
.replaceAll("\\u65e5", " ");
;
dateStr = dateStr.replaceAll("( |<)", "");
try {
return (sdf.parse(changeString2Date(dateStr)).getTime())
+ "";
} catch (ParseException e) {
System.out
.println("=======================>parsedate error: "
+ e.getMessage());
return (new Date()).getTime() + "";
}
}
}
return (new Date()).getTime() + "";
}
private static String changeString2Date(String dateStr) {
System.out.println("=======================>Pre Date: " + dateStr);
String date = "";
String[] tempDateStr = dateStr.trim().replaceAll("\\s+", "@")
.replaceAll("(-|:)", "@").split("@");
if (tempDateStr.length > 0) {
String year = tempDateStr[0];
if (year.length() == 2) {
year = "20" + year;
}
date = date + year + "-";
}
if (tempDateStr.length > 1) {
date = date
+ (tempDateStr[1].length() > 1 ? tempDateStr[1] : "0"
+ tempDateStr[1]) + "-";
}
if (tempDateStr.length > 2) {
date = date
+ (tempDateStr[2].length() > 1 ? tempDateStr[2] : "0"
+ tempDateStr[2]) + " ";
}
if (tempDateStr.length > 3) {
date = date
+ (tempDateStr[3].length() > 1 ? tempDateStr[3] : "0"
+ tempDateStr[3]) + ":";
} else {
date = date + "00:";
}
if (tempDateStr.length > 4) {
date = date
+ (tempDateStr[4].length() > 1 ? tempDateStr[4] : "0"
+ tempDateStr[4]) + ":";
} else {
date = date + "00:";
}
if (tempDateStr.length > 5) {
date = date
+ (tempDateStr[5].length() > 1 ? tempDateStr[5] : "0"
+ tempDateStr[5]);
} else {
date = date + "00";
}
System.out.println("=======================> Date: " + date);
return date;
}
public static void main(String[] args) {
extractDate(
" <td align=\"center\">http://www.rednet.cn <!--时间开始-->2012/3/2 9:14:36<!--时间结束--> ",
"");
}
相关文章推荐
- 使用正则表达式抽取新闻/BBS网页发表时间
- Java使用正则表达式及字符串操作,抽取网页信息
- [转]使用正则表达式优化网页代码
- 使用正则表达式实现模式图片新闻(加强).ASP
- 使用正则表达式验证输入的时间是否正确
- [C#][固定格式网页解析]使用正则表达式处理网页的初步体会
- java正则表达式的使用1 通过正则表达式抓取网页中的email
- [C#][固定格式网页解析]使用正则表达式处理网页的初步体会
- Linux企业级项目实践之网络爬虫(14)——使用正则表达式抽取HTML正文和URL
- 正则表达式 抽取网页中的e-mail地址
- C#中使用正则表达式验证电话号码、手机号、身份证号、数字、邮编、时间(仅年月日)、邮箱、小数的正则表达式
- 使用正则表达式优雅的解决SpringMVC时间转换问题
- c# 正则表达式对网页进行有效内容抽取 | 何苦
- 使用正则表达式优化网页代码
- Python使用正则表达式获取网页中所需要的信息
- php使用curl和正则表达式抓取网页数据示例
- 使用正则表达式匹配一个网页中的所有超级链接
- [C#][固定格式网页解析]使用正则表达式处理网页的初步体会
- 利用正则表达式抽取网页信息