服务器之家:专注于服务器技术及软件下载分享
分类导航

PHP教程|ASP.NET教程|Java教程|ASP教程|编程技术|正则表达式|C/C++|IOS|C#|Swift|Android|VB|R语言|JavaScript|易语言|vb.net|

服务器之家 - 编程语言 - Java教程 - Java实现excel大数据量导入

Java实现excel大数据量导入

2020-08-25 10:29王绍桦 Java教程

这篇文章主要为大家详细介绍了Java实现excel大数据量导入,文中示例代码介绍的非常详细,具有一定的参考价值,感兴趣的小伙伴们可以参考一下

本文实例为大家分享了Java实现excel大数据量导入的具体代码,供大家参考,具体内容如下

情景分析:

通常我们通过poi读取excel文件时,若在用户模式下,由于数据量较大、Sheet较多,很容易出现内存溢出的情况

用户模式读取excel的典型代码如下:

?
1
2
FileInputStream file = new FileInputStream("c:\\test.xlsx");
Workbook wb=new XSSFWorkbook(file);

而03版(xls)excel文件每个sheet最大长度为 65536行,07版(xlsx)excel文件每个sheet最大长度为 1048576行,因此我们主要解决07版(xlsx)excel大数据量导入出现内存溢出的问题

由于读取cvs文件效率高且占用内存少,所以我们采用xlsx转换为cvs的方式来解决

读取xlsx格式数据:

?
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
import java.io.File;
import java.io.FileOutputStream;
import java.io.IOException;
import java.io.InputStream;
import java.io.OutputStream;
import java.io.PrintStream;
import java.text.SimpleDateFormat;
import java.util.ArrayList;
import java.util.Date;
import java.util.HashMap;
import java.util.List;
import java.util.Map;
 
import javax.xml.parsers.ParserConfigurationException;
import javax.xml.parsers.SAXParser;
import javax.xml.parsers.SAXParserFactory;
 
import org.apache.poi.hssf.usermodel.HSSFDateUtil;
import org.apache.poi.openxml4j.exceptions.OpenXML4JException;
import org.apache.poi.openxml4j.opc.OPCPackage;
import org.apache.poi.openxml4j.opc.PackageAccess;
import org.apache.poi.ss.usermodel.BuiltinFormats;
import org.apache.poi.ss.usermodel.DataFormatter;
import org.apache.poi.xssf.eventusermodel.ReadOnlySharedStringsTable;
import org.apache.poi.xssf.eventusermodel.XSSFReader;
import org.apache.poi.xssf.model.StylesTable;
import org.apache.poi.xssf.usermodel.XSSFCellStyle;
import org.apache.poi.xssf.usermodel.XSSFRichTextString;
import org.xml.sax.Attributes;
import org.xml.sax.InputSource;
import org.xml.sax.SAXException;
import org.xml.sax.XMLReader;
import org.xml.sax.helpers.DefaultHandler;
 
public class XLSX2CSV {
 
 enum xssfDataType {
 BOOL, ERROR, FORMULA, INLINESTR, SSTINDEX, NUMBER,
 }
 
 /**
 * 使用xssf_sax_API处理Excel,请参考: http://poi.apache.org/spreadsheet/how-to.html#xssf_sax_api
 */
 class MyXSSFSheetHandler extends DefaultHandler {
 
 private StylesTable stylesTable;
 
 private ReadOnlySharedStringsTable sharedStringsTable;
 
 private final PrintStream output;
 
 private final int minColumnCount;
 
 private boolean vIsOpen;
 
 private xssfDataType nextDataType;
 
 private short formatIndex;
 private String formatString;
 private final DataFormatter formatter;
 
 private int thisColumn = -1;
 private int lastColumnNumber = -1;
 
 private StringBuffer value;
 private String[] record;
 private List<String[]> rows = new ArrayList<String[]>();
 private boolean isCellNull = false;
 
 public MyXSSFSheetHandler(StylesTable styles,ReadOnlySharedStringsTable strings, int cols, PrintStream target) {
 this.stylesTable = styles;
 this.sharedStringsTable = strings;
 this.minColumnCount = cols;
 this.output = target;
 this.value = new StringBuffer();
 this.nextDataType = xssfDataType.NUMBER;
 this.formatter = new DataFormatter();
 record = new String[this.minColumnCount];
 rows.clear();// 每次读取都清空行集合
 }
 
 public void startElement(String uri, String localName, String name,Attributes attributes) throws SAXException {
 if ("inlineStr".equals(name) || "v".equals(name)) {
 vIsOpen = true;
 // Clear contents cache
 value.setLength(0);
 }
 // c => cell
 else if ("c".equals(name)) {
 // Get the cell reference
 String r = attributes.getValue("r");
 int firstDigit = -1;
 for (int c = 0; c < r.length(); ++c) {
  if (Character.isDigit(r.charAt(c))) {
  firstDigit = c;
  break;
  }
 }
 thisColumn = nameToColumn(r.substring(0, firstDigit));
 
 // Set up defaults.
 this.nextDataType = xssfDataType.NUMBER;
 this.formatIndex = -1;
 this.formatString = null;
 String cellType = attributes.getValue("t");
 String cellStyleStr = attributes.getValue("s");
 if ("b".equals(cellType))
  nextDataType = xssfDataType.BOOL;
 else if ("e".equals(cellType))
  nextDataType = xssfDataType.ERROR;
 else if ("inlineStr".equals(cellType))
  nextDataType = xssfDataType.INLINESTR;
 else if ("s".equals(cellType))
  nextDataType = xssfDataType.SSTINDEX;
 else if ("str".equals(cellType))
  nextDataType = xssfDataType.FORMULA;
 else if (cellStyleStr != null) {
  // It's a number, but almost certainly one
  // with a special style or format
  int styleIndex = Integer.parseInt(cellStyleStr);
  XSSFCellStyle style = stylesTable.getStyleAt(styleIndex);
  this.formatIndex = style.getDataFormat();
  this.formatString = style.getDataFormatString();
  if (this.formatString == null)
  this.formatString = BuiltinFormats
  .getBuiltinFormat(this.formatIndex);
 }
 }
 
 }
 
 public void endElement(String uri, String localName, String name)throws SAXException {
 
 String thisStr = null;
 
 // v => contents of a cell
 if ("v".equals(name)) {
 // Process the value contents as required.
 // Do now, as characters() may be called more than once
 switch (nextDataType) {
 
 case BOOL:
  char first = value.charAt(0);
  thisStr = first == '0' ? "FALSE" : "TRUE";
  break;
 
 case ERROR:
  thisStr = "\"ERROR:" + value.toString() + '"';
  break;
 
 case FORMULA:
  // A formula could result in a string value,
  // so always add double-quote characters.
  thisStr = '"' + value.toString() + '"';
  break;
 
 case INLINESTR:
  // TODO: have seen an example of this, so it's untested.
  XSSFRichTextString rtsi = new XSSFRichTextString(value.toString());
  thisStr = '"' + rtsi.toString() + '"';
  break;
 
 case SSTINDEX:
  String sstIndex = value.toString();
  try {
  int idx = Integer.parseInt(sstIndex);
  XSSFRichTextString rtss = new XSSFRichTextString(sharedStringsTable.getEntryAt(idx));
  thisStr = rtss.toString();
  } catch (NumberFormatException ex) {
  output.println("Failed to parse SST index '" + sstIndex
  + "': " + ex.toString());
  }
  break;
 
 case NUMBER:
  String n = value.toString();
  // 判断是否是日期格式
  if (HSSFDateUtil.isADateFormat(this.formatIndex, n)) {
  Double d = Double.parseDouble(n);
  Date date=HSSFDateUtil.getJavaDate(d);
  thisStr=formateDateToString(date);
  } else if (this.formatString != null)
  thisStr = formatter.formatRawCellContents(
  Double.parseDouble(n), this.formatIndex,
  this.formatString);
  else
  thisStr = n;
  break;
 
 default:
  thisStr = "(TODO: Unexpected type: " + nextDataType + ")";
  break;
 }
 
 // Output after we've seen the string contents
 // Emit commas for any fields that were missing on this row
 if (lastColumnNumber == -1) {
  lastColumnNumber = 0;
 }
 //判断单元格的值是否为空
 if (thisStr == null || "".equals(isCellNull)) {
  isCellNull = true;// 设置单元格是否为空值
 }
 record[thisColumn] = thisStr;
 // Update column
 if (thisColumn > -1)
  lastColumnNumber = thisColumn;
 
 } else if ("row".equals(name)) {
 // Print out any missing commas if needed
 if (minColumns > 0) {
  // Columns are 0 based
  if (lastColumnNumber == -1) {
  lastColumnNumber = 0;
  }
  if (isCellNull == false && record[0] != null
  && record[1] != null)// 判断是否空行
  {
  rows.add(record.clone());
  isCellNull = false;
  for (int i = 0; i < record.length; i++) {
  record[i] = null;
  }
  }
 }
 lastColumnNumber = -1;
 }
 
 }
 
 public List<String[]> getRows() {
 return rows;
 }
 
 public void setRows(List<String[]> rows) {
 this.rows = rows;
 }
 
 /**
 * Captures characters only if a suitable element is open. Originally
 * was just "v"; extended for inlineStr also.
 */
 public void characters(char[] ch, int start, int length)throws SAXException {
 if (vIsOpen)
 value.append(ch, start, length);
 }
 
 /**
 * Converts an Excel column name like "C" to a zero-based index.
 * @param name
 * @return Index corresponding to the specified name
 */
 private int nameToColumn(String name) {
 int column = -1;
 for (int i = 0; i < name.length(); ++i) {
 int c = name.charAt(i);
 column = (column + 1) * 26 + c - 'A';
 }
 return column;
 }
 
 private String formateDateToString(Date date) {
 SimpleDateFormat sdf = new SimpleDateFormat("yyyy-MM-dd HH:mm:ss");//格式化日期
 return sdf.format(date);
 
 }
 
 }
 
 private OPCPackage xlsxPackage;
 private int minColumns;
 private PrintStream output;
 
 /**
 * Creates a new XLSX -> CSV converter
 *
 * @param pkg
 *   The XLSX package to process
 * @param output
 *   The PrintStream to output the CSV to
 * @param minColumns
 *   The minimum number of columns to output, or -1 for no minimum
 */
 public XLSX2CSV(OPCPackage pkg, PrintStream output,int minColumns) {
 this.xlsxPackage = pkg;
 this.output = output;
 this.minColumns = minColumns;
 }
 
 /**
 * Parses and shows the content of one sheet using the specified styles and
 * shared-strings tables.
 * @param styles
 * @param strings
 * @param sheetInputStream
 */
 public List<String[]> processSheet(StylesTable styles,
 ReadOnlySharedStringsTable strings, InputStream sheetInputStream)
 throws IOException, ParserConfigurationException, SAXException {
 InputSource sheetSource = new InputSource(sheetInputStream);
 SAXParserFactory saxFactory = SAXParserFactory.newInstance();
 SAXParser saxParser = saxFactory.newSAXParser();
 XMLReader sheetParser = saxParser.getXMLReader();
 MyXSSFSheetHandler handler = new MyXSSFSheetHandler(styles, strings,this.minColumns, this.output);
 sheetParser.setContentHandler(handler);
 sheetParser.parse(sheetSource);
 return handler.getRows();
 }
 
 /**
 * 初始化这个处理程序
 * @throws IOException
 * @throws OpenXML4JException
 * @throws ParserConfigurationException
 * @throws SAXException
 */
 public List<String[]> process() throws IOException, OpenXML4JException,ParserConfigurationException, SAXException {
 ReadOnlySharedStringsTable strings = new ReadOnlySharedStringsTable(this.xlsxPackage);
 XSSFReader xssfReader = new XSSFReader(this.xlsxPackage);
 List<String[]> list = null;
 StylesTable styles = xssfReader.getStylesTable();
 XSSFReader.SheetIterator iter = (XSSFReader.SheetIterator) xssfReader.getSheetsData();
 int index = 0;
 while (iter.hasNext()) {
 InputStream stream = iter.next();
 index++;
 //默认取第一个工作薄
 if(index == 1){
 list = processSheet(styles, strings, stream);
 stream.close();
 }
 }
 return list;
 }
 
 /**
 * 读取Excel
 *
 * @param path
 *   文件路径
 * @param sheetName
 *   sheet名称
 * @param minColumns
 *   列总数
 * @return
 * @throws SAXException
 * @throws ParserConfigurationException
 * @throws OpenXML4JException
 * @throws IOException
 */
 private static List<String[]> readerExcel(String path,int minColumns) throws IOException, OpenXML4JException,ParserConfigurationException, SAXException {
 OPCPackage p = OPCPackage.open(path, PackageAccess.READ);
 XLSX2CSV xlsx2csv = new XLSX2CSV(p, System.out, minColumns);
 List<String[]> list = xlsx2csv.process();
 p.close();
 return list;
 }
 
 /**
 * 读取Excel
 *
 * @param file
 *   File
 * @param sheetName
 *   sheet名称
 * @param minColumns
 *   列总数
 * @return
 * @throws SAXException
 * @throws ParserConfigurationException
 * @throws OpenXML4JException
 * @throws IOException
 */
 private static List<String[]> readerExcelInputStream(File file,int minColumns) throws IOException, OpenXML4JException, ParserConfigurationException, SAXException {
 OPCPackage p = OPCPackage.openOrCreate(file);
 XLSX2CSV xlsx2csv = new XLSX2CSV(p, System.out, minColumns);
 List<String[]> list = xlsx2csv.process();
 p.close();
 return list;
 }
 
 //获取表头及数据内容,可在业务层调用
 public static Map<String,Object> getTableDate(InputStream in) throws IOException, OpenXML4JException, ParserConfigurationException, SAXException{
 File out = new File("C:\\date.xlsx");//默认在C盘生成date.xlsx文件
 OutputStream output = null;
 try {
 output = new FileOutputStream(out);
 byte[] buf = new byte[1024];
 int bytesRead;
 while ((bytesRead = in.read(buf)) > 0) {
 output.write(buf, 0, bytesRead);
 }
 } finally {
 in.close();
 output.close();
 }
 Map<String,Object> map = new HashMap<String,Object>();
 List<String> headresult = new ArrayList<String>();//表头集合
 List<List<String>> dataresult = new ArrayList<List<String>>();//表数据集合
 List<String[]> list = readerExcelInputStream(out,30);//默认读取30列
 for(int i=0;i<list.size();i++){
 if(i == 0){
 for(String cell : list.get(i)){
  if(cell != null){ 
  headresult.add(cell);
  }
 }
 }else{
 List<String> dataList = new ArrayList<String>();
 for(String cell : list.get(i)){
  dataList.add(cell == null ? "" : cell);
 }
 dataresult.add(dataList);
 }
 }
 map.put("headresult", headresult);
 map.put("dataresult", dataresult);
 return map;
 }
 
 public static void main(String[] args) throws Exception {
 File file = new File("C:/Users/Administrator/Desktop/测试.xlsx");
 System.out.println("开始读取...");
 List<String[]> list = XLSX2CSV.readerExcelInputStream(file, 30);//默认读取30列
 System.out.println("数据量:"+list.size());
 for(int i=0;i<list.size();i++){
 if(i > 0){
 for (String cell : list.get(i)) {
  System.out.print(cell + ",");
 }
 System.out.println();
 }
 }
 }
 
}

说明:

代码中的加载excel的核心方法:

?
1
2
3
OPCPackage pkg = OPCPackage.open(path);//文件路径
OPCPackage pkg = OPCPackage.openOrCreate(file);//文件
OPCPackage pkg = OPCPackage.open(InputStream);//文件流

文件流方式对内存依赖极大,所以实际应用时,如果只能获取文件流的话,可以先将文件通过流拷贝到本地再解析代码中的

若上传文件大小超过限制,可在配置文件中设置,SpringBoot2项目application.properties中设置:

?
1
2
spring.servlet.multipart.maxFileSize=50MB
spring.servlet.multipart.maxRequestSize=100MB

实际业务数据入库的话也可采用异步任务@Async的方式来提高入库效率:

?
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
import java.util.Map;
 
import org.springframework.beans.factory.annotation.Autowired;
import org.springframework.scheduling.annotation.Async;
import org.springframework.stereotype.Service;
 
import cn.com.app.dao.JkconfigDao;
 
@Service
public class JkrzService {
 
 @Autowired
 private JkconfigDao jkconfigDao;
 
 @Async("myTaskAsyncPool")
 public void transJkrz(Map<String, Object> m) {
 jkconfigDao.insert(m);
 }
}

线程池配置:

?
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
import java.util.concurrent.Executor;
import java.util.concurrent.ThreadPoolExecutor;
 
import org.springframework.beans.factory.annotation.Autowired;
import org.springframework.context.annotation.Bean;
import org.springframework.context.annotation.Configuration;
import org.springframework.scheduling.annotation.EnableAsync;
import org.springframework.scheduling.concurrent.ThreadPoolTaskExecutor;
 
@Configuration
@EnableAsync
public class AsyncTaskConfig {
 
 @Autowired
 private TaskThreadPoolConfig config;
 
 // ThredPoolTaskExcutor的处理流程
 // 当池子大小小于corePoolSize,就新建线程,并处理请求
 // 当池子大小等于corePoolSize,把请求放入workQueue中,池子里的空闲线程就去workQueue中取任务并处理
 // 当workQueue放不下任务时,就新建线程入池,并处理请求,如果池子大小撑到了maximumPoolSize,就用RejectedExecutionHandler来做拒绝处理
 // 当池子的线程数大于corePoolSize时,多余的线程会等待keepAliveTime长时间,如果无请求可处理就自行销毁
 // 当threadNamePrefix设置为true,则核心线程也会超时关闭
 @Bean
 public Executor myTaskAsyncPool() {
  ThreadPoolTaskExecutor executor = new ThreadPoolTaskExecutor();
   executor.setCorePoolSize(config.getCorePoolSize());
   executor.setMaxPoolSize(config.getMaxPoolSize());
   executor.setQueueCapacity(config.getQueueCapacity());
   executor.setKeepAliveSeconds(config.getKeepAliveSeconds());
   executor.setAllowCoreThreadTimeOut(true);
   executor.setThreadNamePrefix("MyExecutor-");
 
   // rejection-policy:当pool已经达到max size的时候,如何处理新任务
   // CALLER_RUNS:不在新线程中执行任务,而是由调用者所在的线程来执行
   executor.setRejectedExecutionHandler(new ThreadPoolExecutor.CallerRunsPolicy());
   executor.initialize();
  return executor;
 }
 
}

参数配置:

?
1
2
3
4
5
6
7
8
#核心线程数,当线程数小于核心线程数时,即使有线程空闲,线程池也会优先创建新线程,设置allowCoreThreadTimeout=true(默认false)时,核心线程会超时退出
spring.task.pool.corePoolSize=20
#最大线程数,当线程数大于等于corePoolSize,且任务队列已满时,线程池会创建新线程来处理任务
spring.task.pool.maxPoolSize=60
#线程空闲时间,当线程空闲时间达到keepAliveSeconds(秒)时,线程会退出,直到线程数量等于corePoolSize,如果allowCoreThreadTimeout=true,则会直到线程数量等于0
spring.task.pool.keepAliveSeconds=1
#任务队列容量,当核心线程数达到最大时,新任务会放在队列中排队等待执行
spring.task.pool.queueCapacity=400

读取xls格式数据:

?
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
import org.apache.poi.hssf.eventusermodel.*;
import org.apache.poi.hssf.eventusermodel.dummyrecord.LastCellOfRowDummyRecord;
import org.apache.poi.hssf.eventusermodel.dummyrecord.MissingCellDummyRecord;
import org.apache.poi.hssf.model.HSSFFormulaParser;
import org.apache.poi.hssf.record.*;
import org.apache.poi.hssf.usermodel.HSSFDataFormatter;
import org.apache.poi.hssf.usermodel.HSSFWorkbook;
import org.apache.poi.poifs.filesystem.POIFSFileSystem;
 
import java.io.FileInputStream;
import java.io.InputStream;
import java.util.ArrayList;
import java.util.List;
 
/**
 * @desc 用于解决.xls2003版本大数据量问题
 **/
public class ExcelXlsReader implements HSSFListener {
 
 //存储所有数据
 private List<List<String>> dataList = new ArrayList<List<String>>();
 
 private int minColums = -1;
 
 private POIFSFileSystem fs;
 
 /**
  * 总行数
  */
 private int totalRows=0;
 
 /**
  * 上一行row的序号
  */
 private int lastRowNumber;
 
 /**
  * 上一单元格的序号
  */
 private int lastColumnNumber;
 
 /**
  * 是否输出formula,还是它对应的值
  */
 private boolean outputFormulaValues = true;
 
 /**
  * 用于转换formulas
  */
 private EventWorkbookBuilder.SheetRecordCollectingListener workbookBuildingListener;
 
 //excel2003工作簿
 private HSSFWorkbook stubWorkbook;
 
 private SSTRecord sstRecord;
 
 private FormatTrackingHSSFListener formatListener;
 
 private final HSSFDataFormatter formatter = new HSSFDataFormatter();
 
 /**
  * 文件的绝对路径
  */
 private String filePath = "";
 
 //表索引
 private int sheetIndex = 0;
 
 private BoundSheetRecord[] orderedBSRs;
 
 @SuppressWarnings("unchecked")
 private ArrayList boundSheetRecords = new ArrayList();
 
 private int nextRow;
 
 private int nextColumn;
 
 private boolean outputNextStringRecord;
 
 //当前行
 private int curRow = 0;
 
 //存储一行记录所有单元格的容器
 private List<String> cellList = new ArrayList<String>();
 
 /**
  * 判断整行是否为空行的标记
  */
 private boolean flag = false;
 
 @SuppressWarnings("unused")
 private String sheetName;
 
 /**
  * 遍历excel下所有的sheet
  *
  * @param fileName
  * @throws Exception
  */
 public int process(String fileName) throws Exception {
  filePath = fileName;
  this.fs = new POIFSFileSystem(new FileInputStream(fileName));
  MissingRecordAwareHSSFListener listener = new MissingRecordAwareHSSFListener(this);
  formatListener = new FormatTrackingHSSFListener(listener);
  HSSFEventFactory factory = new HSSFEventFactory();
  HSSFRequest request = new HSSFRequest();
  if (outputFormulaValues) {
   request.addListenerForAllRecords(formatListener);
  } else {
   workbookBuildingListener = new EventWorkbookBuilder.SheetRecordCollectingListener(formatListener);
   request.addListenerForAllRecords(workbookBuildingListener);
  }
  factory.processWorkbookEvents(request, fs);
 
  return totalRows; //返回该excel文件的总行数,不包括首列和空行
 }
 
 public List<List<String>> process(InputStream in) throws Exception {
  //filePath = fileName;
  this.fs = new POIFSFileSystem(in);
  MissingRecordAwareHSSFListener listener = new MissingRecordAwareHSSFListener(this);
  formatListener = new FormatTrackingHSSFListener(listener);
  HSSFEventFactory factory = new HSSFEventFactory();
  HSSFRequest request = new HSSFRequest();
  if (outputFormulaValues) {
   request.addListenerForAllRecords(formatListener);
  } else {
   workbookBuildingListener = new EventWorkbookBuilder.SheetRecordCollectingListener(formatListener);
   request.addListenerForAllRecords(workbookBuildingListener);
  }
  factory.processWorkbookEvents(request, fs);
 
  return dataList; //返回该excel文件的总行数,不包括首列和空行
 }
 
 /**
  * HSSFListener 监听方法,处理Record
  * 处理每个单元格
  * @param record
  */
 @SuppressWarnings("unchecked")
 public void processRecord(Record record) {
  int thisRow = -1;
  int thisColumn = -1;
  String thisStr = null;
  String value = null;
  switch (record.getSid()) {
   case BoundSheetRecord.sid:
    boundSheetRecords.add(record);
    break;
   case BOFRecord.sid: //开始处理每个sheet
    BOFRecord br = (BOFRecord) record;
    if (br.getType() == BOFRecord.TYPE_WORKSHEET) {
     //如果有需要,则建立子工作簿
     if (workbookBuildingListener != null && stubWorkbook == null) {
      stubWorkbook = workbookBuildingListener.getStubHSSFWorkbook();
     }
 
     if (orderedBSRs == null) {
      orderedBSRs = BoundSheetRecord.orderByBofPosition(boundSheetRecords);
     }
     sheetName = orderedBSRs[sheetIndex].getSheetname();
     sheetIndex++;
    }
    break;
   case SSTRecord.sid:
    sstRecord = (SSTRecord) record;
    break;
   case BlankRecord.sid: //单元格为空白
    BlankRecord brec = (BlankRecord) record;
    thisRow = brec.getRow();
    thisColumn = brec.getColumn();
    thisStr = "";
    cellList.add(thisColumn, thisStr);
    break;
   case BoolErrRecord.sid: //单元格为布尔类型
    BoolErrRecord berec = (BoolErrRecord) record;
    thisRow = berec.getRow();
    thisColumn = berec.getColumn();
    thisStr = berec.getBooleanValue() + "";
    cellList.add(thisColumn, thisStr);
    checkRowIsNull(thisStr); //如果里面某个单元格含有值,则标识该行不为空行
    break;
   case FormulaRecord.sid://单元格为公式类型
    FormulaRecord frec = (FormulaRecord) record;
    thisRow = frec.getRow();
    thisColumn = frec.getColumn();
    if (outputFormulaValues) {
     if (Double.isNaN(frec.getValue())) {
      outputNextStringRecord = true;
      nextRow = frec.getRow();
      nextColumn = frec.getColumn();
     } else {
      thisStr = '"' + HSSFFormulaParser.toFormulaString(stubWorkbook, frec.getParsedExpression()) + '"';
     }
    } else {
     thisStr = '"' + HSSFFormulaParser.toFormulaString(stubWorkbook, frec.getParsedExpression()) + '"';
    }
    cellList.add(thisColumn, thisStr);
    checkRowIsNull(thisStr); //如果里面某个单元格含有值,则标识该行不为空行
    break;
   case StringRecord.sid: //单元格中公式的字符串
    if (outputNextStringRecord) {
     StringRecord srec = (StringRecord) record;
     thisStr = srec.getString();
     thisRow = nextRow;
     thisColumn = nextColumn;
     outputNextStringRecord = false;
    }
    break;
   case LabelRecord.sid:
    LabelRecord lrec = (LabelRecord) record;
    curRow = thisRow = lrec.getRow();
    thisColumn = lrec.getColumn();
    value = lrec.getValue().trim();
    value = value.equals("") ? "" : value;
    cellList.add(thisColumn, value);
    checkRowIsNull(value); //如果里面某个单元格含有值,则标识该行不为空行
    break;
   case LabelSSTRecord.sid: //单元格为字符串类型
    LabelSSTRecord lsrec = (LabelSSTRecord) record;
    curRow = thisRow = lsrec.getRow();
    thisColumn = lsrec.getColumn();
    if (sstRecord == null) {
     cellList.add(thisColumn, "");
    } else {
     value = sstRecord.getString(lsrec.getSSTIndex()).toString().trim();
     value = value.equals("") ? "" : value;
     cellList.add(thisColumn, value);
     checkRowIsNull(value); //如果里面某个单元格含有值,则标识该行不为空行
    }
    break;
   case NumberRecord.sid: //单元格为数字类型
    NumberRecord numrec = (NumberRecord) record;
    curRow = thisRow = numrec.getRow();
    thisColumn = numrec.getColumn();
 
    //第一种方式
    //value = formatListener.formatNumberDateCell(numrec).trim();//这个被写死,采用的m/d/yy h:mm格式,不符合要求
 
    //第二种方式,参照formatNumberDateCell里面的实现方法编写
    Double valueDouble=((NumberRecord)numrec).getValue();
    String formatString=formatListener.getFormatString(numrec);
    if (formatString.contains("m/d/yy")){
     formatString="yyyy-MM-dd hh:mm:ss";
    }
    int formatIndex=formatListener.getFormatIndex(numrec);
    value=formatter.formatRawCellContents(valueDouble, formatIndex, formatString).trim();
 
    value = value.equals("") ? "" : value;
    //向容器加入列值
    cellList.add(thisColumn, value);
    checkRowIsNull(value); //如果里面某个单元格含有值,则标识该行不为空行
    break;
   default:
    break;
  }
 
  //遇到新行的操作
  if (thisRow != -1 && thisRow != lastRowNumber) {
   lastColumnNumber = -1;
  }
 
  //空值的操作
  if (record instanceof MissingCellDummyRecord) {
   MissingCellDummyRecord mc = (MissingCellDummyRecord) record;
   curRow = thisRow = mc.getRow();
   thisColumn = mc.getColumn();
   cellList.add(thisColumn, "");
  }
 
  //更新行和列的值
  if (thisRow > -1)
   lastRowNumber = thisRow;
  if (thisColumn > -1)
   lastColumnNumber = thisColumn;
 
  //行结束时的操作
  if (record instanceof LastCellOfRowDummyRecord) {
   if (minColums > 0) {
    //列值重新置空
    if (lastColumnNumber == -1) {
     lastColumnNumber = 0;
    }
   }
   lastColumnNumber = -1;
 
   if (flag&&curRow!=0) { //该行不为空行且该行不是第一行,发送(第一行为列名,不需要)
    //ExcelReaderUtil.sendRows(filePath, sheetName, sheetIndex, curRow + 1, cellList); //每行结束时,调用sendRows()方法
    totalRows++;
    //添加到数据集合中
    dataList.add(cellList);
   }
   //清空容器
   cellList = new ArrayList<String>();
   //cellList.clear();
   flag=false;
  }
 }
 
 /**
  * 如果里面某个单元格含有值,则标识该行不为空行
  * @param value
  */
 public void checkRowIsNull(String value){
  if (value != null && !"".equals(value)) {
   flag = true;
  }
 }
}

读取xlsx格式数据(自动获取表头长度方式):

?
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
import org.apache.poi.openxml4j.opc.OPCPackage;
import org.apache.poi.ss.usermodel.BuiltinFormats;
import org.apache.poi.ss.usermodel.DataFormatter;
import org.apache.poi.xssf.eventusermodel.XSSFReader;
import org.apache.poi.xssf.model.SharedStringsTable;
import org.apache.poi.xssf.model.StylesTable;
import org.apache.poi.xssf.usermodel.XSSFCellStyle;
import org.apache.poi.xssf.usermodel.XSSFRichTextString;
import org.xml.sax.Attributes;
import org.xml.sax.InputSource;
import org.xml.sax.SAXException;
import org.xml.sax.XMLReader;
import org.xml.sax.helpers.DefaultHandler;
import org.xml.sax.helpers.XMLReaderFactory;
 
import java.io.InputStream;
import java.util.ArrayList;
import java.util.List;
 
/**
 * @desc POI读取excel有两种模式,一种是用户模式,一种是事件驱动模式
 * 采用SAX事件驱动模式解决XLSX文件,可以有效解决用户模式内存溢出的问题,
 * 该模式是POI官方推荐的读取大数据的模式,
 * 在用户模式下,数据量较大,Sheet较多,或者是有很多无用的空行的情况下,容易出现内存溢出
 * <p>
 * 用于解决.xlsx2007版本大数据量问题
 **/
public class ExcelXlsxReader extends DefaultHandler {
 
 //存储所有数据
 private List<List<String>> dataList = new ArrayList<List<String>>();
 
 /**
  * 单元格中的数据可能的数据类型
  */
 enum CellDataType {
  BOOL, ERROR, FORMULA, INLINESTR, SSTINDEX, NUMBER, DATE, NULL
 }
 
 /**
  * 共享字符串表
  */
 private SharedStringsTable sst;
 
 /**
  * 上一次的索引值
  */
 private String lastIndex;
 
 /**
  * 文件的绝对路径
  */
 private String filePath = "";
 
 /**
  * 工作表索引
  */
 private int sheetIndex = 0;
 
 /**
  * sheet名
  */
 private String sheetName = "";
 
 /**
  * 总行数
  */
 private int totalRows=0;
 
 /**
  * 一行内cell集合
  */
 private List<String> cellList = new ArrayList<String>();
 
 /**
  * 判断整行是否为空行的标记
  */
 private boolean flag = false;
 
 /**
  * 当前行
  */
 private int curRow = 1;
 
 /**
  * 当前列
  */
 private int curCol = 0;
 
 /**
  * T元素标识
  */
 private boolean isTElement;
 
 /**
  * 判断上一单元格是否为文本空单元格
  */
 private boolean startElementFlag = true;
 private boolean endElementFlag = false;
 private boolean charactersFlag = false;
 
 /**
  * 异常信息,如果为空则表示没有异常
  */
 private String exceptionMessage;
 
 /**
  * 单元格数据类型,默认为字符串类型
  */
 private CellDataType nextDataType = CellDataType.SSTINDEX;
 
 private final DataFormatter formatter = new DataFormatter();
 
 /**
  * 单元格日期格式的索引
  */
 private short formatIndex;
 
 /**
  * 日期格式字符串
  */
 private String formatString;
 
 //定义前一个元素和当前元素的位置,用来计算其中空的单元格数量,如A6和A8等
 private String prePreRef = "A", preRef = null, ref = null;
 
 //定义该文档一行最大的单元格数,用来补全一行最后可能缺失的单元格
 private String maxRef = null;
 
 /**
  * 单元格
  */
 private StylesTable stylesTable;
 
 /**
  * 遍历工作簿中所有的电子表格
  * 并缓存在mySheetList中
  *
  * @param filename
  * @throws Exception
  */
 public int process(String filename) throws Exception {
  filePath = filename;
  OPCPackage pkg = OPCPackage.open(filename);
  XSSFReader xssfReader = new XSSFReader(pkg);
  stylesTable = xssfReader.getStylesTable();
  SharedStringsTable sst = xssfReader.getSharedStringsTable();
  XMLReader parser = XMLReaderFactory.createXMLReader("org.apache.xerces.parsers.SAXParser");
  this.sst = sst;
  parser.setContentHandler(this);
  XSSFReader.SheetIterator sheets = (XSSFReader.SheetIterator) xssfReader.getSheetsData();
  while (sheets.hasNext()) { //遍历sheet
   curRow = 1; //标记初始行为第一行
   sheetIndex++;
   InputStream sheet = sheets.next(); //sheets.next()和sheets.getSheetName()不能换位置,否则sheetName报错
   sheetName = sheets.getSheetName();
   InputSource sheetSource = new InputSource(sheet);
   parser.parse(sheetSource); //解析excel的每条记录,在这个过程中startElement()、characters()、endElement()这三个函数会依次执行
   sheet.close();
  }
  return totalRows; //返回该excel文件的总行数,不包括首列和空行
 }
 
 public List<List<String>> process(InputStream in) throws Exception {
  //filePath = filename;
  OPCPackage pkg = OPCPackage.open(in);
  XSSFReader xssfReader = new XSSFReader(pkg);
  stylesTable = xssfReader.getStylesTable();
  SharedStringsTable sst = xssfReader.getSharedStringsTable();
  //XMLReader parser = XMLReaderFactory.createXMLReader("org.apache.xerces.parsers.SAXParser");
  XMLReader parser = XMLReaderFactory.createXMLReader();
  this.sst = sst;
  parser.setContentHandler(this);
  XSSFReader.SheetIterator sheets = (XSSFReader.SheetIterator) xssfReader.getSheetsData();
  while (sheets.hasNext()) { //遍历sheet
   curRow = 1; //标记初始行为第一行
   sheetIndex++;
   InputStream sheet = sheets.next(); //sheets.next()和sheets.getSheetName()不能换位置,否则sheetName报错
   sheetName = sheets.getSheetName();
   InputSource sheetSource = new InputSource(sheet);
   parser.parse(sheetSource); //解析excel的每条记录,在这个过程中startElement()、characters()、endElement()这三个函数会依次执行
   sheet.close();
  }
  return dataList;
 }
 
 /**
  * 第一个执行
  *
  * @param uri
  * @param localName
  * @param name
  * @param attributes
  * @throws SAXException
  */
 @Override
 public void startElement(String uri, String localName, String name, Attributes attributes) throws SAXException {
  //c => 单元格
  if ("c".equals(name)) {
 
   //前一个单元格的位置
   if (preRef == null) {
    preRef = attributes.getValue("r");
 
   } else {
    //中部文本空单元格标识 ‘endElementFlag' 判断前一次是否为文本空字符串,true则表明不是文本空字符串,false表明是文本空字符串跳过把空字符串的位置赋予preRef
    if (endElementFlag){
     preRef = ref;
    }
   }
 
   //当前单元格的位置
   ref = attributes.getValue("r");
   //首部文本空单元格标识 ‘startElementFlag' 判断前一次,即首部是否为文本空字符串,true则表明不是文本空字符串,false表明是文本空字符串, 且已知当前格,即第二格带“B”标志,则ref赋予preRef
   if (!startElementFlag && !flag){ //上一个单元格为文本空单元格,执行下面的,使ref=preRef;flag为true表明该单元格之前有数据值,即该单元格不是首部空单元格,则跳过
    // 这里只有上一个单元格为文本空单元格,且之前的几个单元格都没有值才会执行
    preRef = ref;
   }
 
   //设定单元格类型
   this.setNextDataType(attributes);
   endElementFlag = false;
   charactersFlag = false;
   startElementFlag = false;
  }
 
  //当元素为t时
  if ("t".equals(name)) {
   isTElement = true;
  } else {
   isTElement = false;
  }
 
  //置空
  lastIndex = "";
 }
 
 
 
 /**
  * 第二个执行
  * 得到单元格对应的索引值或是内容值
  * 如果单元格类型是字符串、INLINESTR、数字、日期,lastIndex则是索引值
  * 如果单元格类型是布尔值、错误、公式,lastIndex则是内容值
  * @param ch
  * @param start
  * @param length
  * @throws SAXException
  */
 @Override
 public void characters(char[] ch, int start, int length) throws SAXException {
  startElementFlag = true;
  charactersFlag = true;
  lastIndex += new String(ch, start, length);
 }
 
 /**
  * 第三个执行
  *
  * @param uri
  * @param localName
  * @param name
  * @throws SAXException
  */
 @Override
 public void endElement(String uri, String localName, String name) throws SAXException {
  //t元素也包含字符串
  if (isTElement) {
   //将单元格内容加入rowlist中,在这之前先去掉字符串前后的空白符
   String value = lastIndex.trim();
   cellList.add(curCol, value);
   endElementFlag = true;
   curCol++;
   isTElement = false;
   //如果里面某个单元格含有值,则标识该行不为空行
   if (value != null && !"".equals(value)) {
    flag = true;
   }
  } else if ("v".equals(name)) {
   //v => 单元格的值,如果单元格是字符串,则v标签的值为该字符串在SST中的索引
   String value = this.getDataValue(lastIndex.trim(), "");//根据索引值获取对应的单元格值
 
   //补全单元格之间的空单元格
   if (!ref.equals(preRef)) {
    int len = countNullCell(ref, preRef);
    for (int i = 0; i < len; i++) {
     cellList.add(curCol, "");
     curCol++;
    }
   } else if (ref.equals(preRef) && !ref.startsWith("A")){ //ref等于preRef,且以B或者C...开头,表明首部为空格
    int len = countNullCell(ref, "A");
    for (int i = 0; i <= len; i++) {
     cellList.add(curCol, "");
     curCol++;
    }
   }
   cellList.add(curCol, value);
   curCol++;
   endElementFlag = true;
   //如果里面某个单元格含有值,则标识该行不为空行
   if (value != null && !"".equals(value)) {
    flag = true;
   }
  } else {
   //如果标签名称为row,这说明已到行尾,调用optRows()方法
   if ("row".equals(name)) {
    //默认第一行为表头,以该行单元格数目为最大数目
    if (curRow == 1) {
     maxRef = ref;
    }
    //补全一行尾部可能缺失的单元格
    if (maxRef != null) {
     int len = -1;
     //前一单元格,true则不是文本空字符串,false则是文本空字符串
     if (charactersFlag){
      len = countNullCell(maxRef, ref);
     }else {
      len = countNullCell(maxRef, preRef);
     }
     for (int i = 0; i <= len; i++) {
      cellList.add(curCol, "");
      curCol++;
     }
    }
 
    if (flag&&curRow!=1){ //该行不为空行且该行不是第一行,则发送(第一行为列名,不需要)
     //ExcelReaderUtil.sendRows(filePath, sheetName, sheetIndex, curRow, cellList);
     totalRows++;
     //添加到数据集合中
     dataList.add(cellList);
    }
 
    //清空容器
    cellList = new ArrayList<String>();
    //cellList.clear();
    curRow++;
    curCol = 0;
    preRef = null;
    prePreRef = null;
    ref = null;
    flag=false;
   }
  }
 }
 
 /**
  * 处理数据类型
  *
  * @param attributes
  */
 public void setNextDataType(Attributes attributes) {
  nextDataType = CellDataType.NUMBER; //cellType为空,则表示该单元格类型为数字
  formatIndex = -1;
  formatString = null;
  String cellType = attributes.getValue("t"); //单元格类型
  String cellStyleStr = attributes.getValue("s"); //
  String columnData = attributes.getValue("r"); //获取单元格的位置,如A1,B1
 
  if ("b".equals(cellType)) { //处理布尔值
   nextDataType = CellDataType.BOOL;
  } else if ("e".equals(cellType)) { //处理错误
   nextDataType = CellDataType.ERROR;
  } else if ("inlineStr".equals(cellType)) {
   nextDataType = CellDataType.INLINESTR;
  } else if ("s".equals(cellType)) { //处理字符串
   nextDataType = CellDataType.SSTINDEX;
  } else if ("str".equals(cellType)) {
   nextDataType = CellDataType.FORMULA;
  }
 
  if (cellStyleStr != null) { //处理日期
   int styleIndex = Integer.parseInt(cellStyleStr);
   XSSFCellStyle style = stylesTable.getStyleAt(styleIndex);
   formatIndex = style.getDataFormat();
   formatString = style.getDataFormatString();
   if (formatString.contains("m/d/yyyy") || formatString.contains("yyyy/mm/dd")|| formatString.contains("yyyy/m/d") ) {
    nextDataType = CellDataType.DATE;
    formatString = "yyyy-MM-dd hh:mm:ss";
   }
 
   if (formatString == null) {
    nextDataType = CellDataType.NULL;
    formatString = BuiltinFormats.getBuiltinFormat(formatIndex);
   }
  }
 }
 
 /**
  * 对解析出来的数据进行类型处理
  * @param value 单元格的值,
  *    value代表解析:BOOL的为0或1, ERROR的为内容值,FORMULA的为内容值,INLINESTR的为索引值需转换为内容值,
  *    SSTINDEX的为索引值需转换为内容值, NUMBER为内容值,DATE为内容值
  * @param thisStr 一个空字符串
  * @return
  */
 @SuppressWarnings("deprecation")
 public String getDataValue(String value, String thisStr) {
  switch (nextDataType) {
   // 这几个的顺序不能随便交换,交换了很可能会导致数据错误
   case BOOL: //布尔值
    char first = value.charAt(0);
    thisStr = first == '0' ? "FALSE" : "TRUE";
    break;
   case ERROR: //错误
    thisStr = "\"ERROR:" + value.toString() + '"';
    break;
   case FORMULA: //公式
    thisStr = '"' + value.toString() + '"';
    break;
   case INLINESTR:
    XSSFRichTextString rtsi = new XSSFRichTextString(value.toString());
    thisStr = rtsi.toString();
    rtsi = null;
    break;
   case SSTINDEX: //字符串
    String sstIndex = value.toString();
    try {
     int idx = Integer.parseInt(sstIndex);
     XSSFRichTextString rtss = new XSSFRichTextString(sst.getEntryAt(idx));//根据idx索引值获取内容值
     thisStr = rtss.toString();
     //System.out.println(thisStr);
     //有些字符串是文本格式的,但内容却是日期
 
     rtss = null;
    } catch (NumberFormatException ex) {
     thisStr = value.toString();
    }
    break;
   case NUMBER: //数字
    if (formatString != null) {
     thisStr = formatter.formatRawCellContents(Double.parseDouble(value), formatIndex, formatString).trim();
    } else {
     thisStr = value;
    }
    thisStr = thisStr.replace("_", "").trim();
    break;
   case DATE: //日期
    thisStr = formatter.formatRawCellContents(Double.parseDouble(value), formatIndex, formatString);
    // 对日期字符串作特殊处理,去掉T
    thisStr = thisStr.replace("T", " ");
    break;
   default:
    thisStr = " ";
    break;
  }
  return thisStr;
 }
 
 public int countNullCell(String ref, String preRef) {
  //excel2007最大行数是1048576,最大列数是16384,最后一列列名是XFD
  String xfd = ref.replaceAll("\\d+", "");
  String xfd_1 = preRef.replaceAll("\\d+", "");
 
  xfd = fillChar(xfd, 3, '@', true);
  xfd_1 = fillChar(xfd_1, 3, '@', true);
 
  char[] letter = xfd.toCharArray();
  char[] letter_1 = xfd_1.toCharArray();
  int res = (letter[0] - letter_1[0]) * 26 * 26 + (letter[1] - letter_1[1]) * 26 + (letter[2] - letter_1[2]);
  return res - 1;
 }
 
 public String fillChar(String str, int len, char let, boolean isPre) {
  int len_1 = str.length();
  if (len_1 < len) {
   if (isPre) {
    for (int i = 0; i < (len - len_1); i++) {
     str = let + str;
    }
   } else {
    for (int i = 0; i < (len - len_1); i++) {
     str = str + let;
    }
   }
  }
  return str;
 }
 
 /**
  * @return the exceptionMessage
  */
 public String getExceptionMessage() {
  return exceptionMessage;
 }
}

读取工具类:

?
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
import java.io.InputStream;
import java.util.List;
 
public class ExcelReaderUtil {
 
 //excel2003扩展名
 public static final String EXCEL03_EXTENSION = ".xls";
 //excel2007扩展名
 public static final String EXCEL07_EXTENSION = ".xlsx";
 
 //读取xls格式
 public static List<List<String>> readExcelXls(InputStream in) throws Exception {
  ExcelXlsReader excelXls=new ExcelXlsReader();
  List<List<String>> dataList =excelXls.process(in);
  return dataList;
 }
 
 //读取xlsx格式
 public static List<List<String>> readExcelXlsx(InputStream in) throws Exception {
  ExcelXlsxReader excelXls=new ExcelXlsxReader();
  List<List<String>> dataList =excelXls.process(in);
  return dataList;
 }
 
 public static void readExcel(String fileName) throws Exception {
  int totalRows =0;
  if (fileName.endsWith(EXCEL03_EXTENSION)) { //处理excel2003文件
   ExcelXlsReader excelXls=new ExcelXlsReader();
   totalRows =excelXls.process(fileName);
  } else if (fileName.endsWith(EXCEL07_EXTENSION)) {//处理excel2007文件
   ExcelXlsxReader excelXlsxReader = new ExcelXlsxReader();
   totalRows = excelXlsxReader.process(fileName);
  } else {
   throw new Exception("文件格式错误,fileName的扩展名只能是xls或xlsx。");
  }
  System.out.println("发送的总行数:" + totalRows);
 }
 
 public static void main(String[] args) throws Exception {
  String path="C:\\test.xlsx";
  ExcelReaderUtil.readExcel(path);
 }
}

以上就是本文的全部内容,希望对大家的学习有所帮助,也希望大家多多支持服务器之家。

原文链接:https://blog.csdn.net/rexueqingchun/article/details/89306707

延伸 · 阅读

精彩推荐