Excel 动态拼接表头实现导出
public class Column {
//单元格内容
private String content;
//字段名称,用户导出表格时反射调用
private String fieldName;
//这个单元格的集合
private List<Column> listTpamscolumn = new ArrayList<Column>();
int totalRow;
int totalCol;
int row;//excel第几行
int col;//excel第几列
int rLen; //excel 跨多少行
int cLen;//excel跨多少列
private boolean HasChilren;//是否有子节点
private int tree_step;//树的级别 从0开始
private String id;
private String pid;
public Column() {
}
;
public Column(String content, String fieldName) {
this.content = content;
this.fieldName = fieldName;
}
public Column(String fieldName, String content, int tree_step) {
this.tree_step = tree_step;
this.fieldName = fieldName;
this.content = content;
}
public int getTotalRow() {
return totalRow;
}
public void setTotalRow(int totalRow) {
this.totalRow = totalRow;
}
public int getTotalCol() {
return totalCol;
}
public void setTotalCol(int totalCol) {
this.totalCol = totalCol;
}
public String getPid() {
return pid;
}
public void setPid(String pid) {
this.pid = pid;
}
public String getId() {
return id;
}
public void setId(String id) {
this.id = id;
}
public boolean isHasChilren() {
return HasChilren;
}
public void setHasChilren(boolean hasChilren) {
HasChilren = hasChilren;
}
public String getContent() {
return content;
}
public void setContent(String content) {
this.content = content;
}
public String getFieldName() {
return fieldName;
}
public void setFieldName(String fieldName) {
this.fieldName = fieldName;
}
public List<Column> getListTpamscolumn() {
return listTpamscolumn;
}
public void setListTpamscolumn(List<Column> listTpamscolumn) {
this.listTpamscolumn = listTpamscolumn;
}
public int getTree_step() {
return tree_step;
}
public void setTree_step(int tree_step) {
this.tree_step = tree_step;
}
public int getRow() {
return row;
}
public void setRow(int row) {
this.row = row;
}
public int getCol() {
return col;
}
public void setCol(int col) {
this.col = col;
}
public int getrLen() {
return rLen;
}
public void setrLen(int rLen) {
this.rLen = rLen;
}
public int getcLen() {
return cLen;
}
public void setcLen(int cLen) {
this.cLen = cLen;
}
}
/**
* excel处理工具
* 概念-> 表头数据:报表的表头
* 行内数据:表头以下的数据
* 功能:动态生成单级,多级Excel表头
* 备注:tree型结构数据的root节点的id必须为零(0)
*/
public class ExcelTool<T> {
private HSSFWorkbook workbook;//excel 对象
private String title; //表格标题
private int colWidth = 20; //单元格宽度
private int rowHeight = 20;//单元格行高度
private HSSFCellStyle styleHead; //表头样式
private HSSFCellStyle styleBody; //主体样式
private SimpleDateFormat sdf = new SimpleDateFormat("yyyy-MM-dd HH:mm:ss"); //日期格式化,默认yyyy-MM-dd HH:mm:ss
/**
* 无参数 初始化 对象
*/
public ExcelTool() {
this.title = "sheet1";
this.workbook = new HSSFWorkbook();
init(0);
}
/**
* 有参数 初始化 对象
*
* @param title
* @param colWidth
* @param rowHeight
* @param dateFormat
*/
public ExcelTool(String title, int colWidth, int rowHeight, String dateFormat) {
this.colWidth = colWidth;
this.rowHeight = rowHeight;
this.title = title;
this.workbook = new HSSFWorkbook();
this.sdf = new SimpleDateFormat(dateFormat);
init(0);
}
public ExcelTool(String title, int colWidth, int rowHeight) {
this.colWidth = colWidth;
this.rowHeight = rowHeight;
this.title = title;
this.workbook = new HSSFWorkbook();
init(0);
}
public ExcelTool(String title, int colWidth, int rowHeight, int flag) {
this.colWidth = colWidth;
this.rowHeight = rowHeight;
this.title = title;
this.workbook = new HSSFWorkbook();
init(flag);
}
public ExcelTool(String title) {
this.title = title;
this.workbook = new HSSFWorkbook();
init(0);
}
/**
* ExcelTool 属性 get、set 方法 开始
*/
public int getColWidth() {
return colWidth;
}
public void setColWidth(int colWidth) {
this.colWidth = colWidth;
}
public int getRowHeight() {
return rowHeight;
}
public void setRowHeight(int rowHeight) {
this.rowHeight = rowHeight;
}
public HSSFWorkbook getWorkbook() {
return this.workbook;
}
public void setWorkbook(HSSFWorkbook workbook) {
this.workbook = workbook;
}
public String getTitle() {
return title;
}
public void setTitle(String title) {
this.title = title;
}
public HSSFCellStyle getStyleHead() {
return styleHead;
}
public void setStyleHead(HSSFCellStyle styleHead) {
this.styleHead = styleHead;
}
public HSSFCellStyle getStyleBody() {
return styleBody;
}
public void setStyleBody(HSSFCellStyle styleBody) {
this.styleBody = styleBody;
}
/**
* ExcelTool 属性 get、set 方法 结束
*/
//内部统一调用的样式初始化
private void init(int styleFlag) {
this.styleHead = this.workbook.createCellStyle();
this.styleHead.setAlignment(HSSFCellStyle.ALIGN_CENTER);// 左右居中
this.styleHead.setVerticalAlignment(HSSFCellStyle.VERTICAL_CENTER);// 上下居中
this.styleHead.setRightBorderColor(HSSFColor.BLACK.index);
this.styleHead.setBottomBorderColor(HSSFColor.BLACK.index);
switch (styleFlag) {
case 1:
this.styleBody = this.workbook.createCellStyle();
this.styleBody.setAlignment(HSSFCellStyle.ALIGN_LEFT);// 左右居中ALIGN_CENTER
this.styleBody.setVerticalAlignment(HSSFCellStyle.VERTICAL_CENTER);// 上下居中
this.styleBody.setRightBorderColor(HSSFColor.BLACK.index);
this.styleBody.setBottomBorderColor(HSSFColor.BLACK.index);
this.styleBody.setBorderRight((short) 1);// 边框的大小
this.styleBody.setBorderBottom((short) 1);// 边框的大小
break;
default:
this.styleBody = this.workbook.createCellStyle();
this.styleBody.setAlignment(HSSFCellStyle.ALIGN_CENTER);// 左右居中ALIGN_CENTER
this.styleBody.setVerticalAlignment(HSSFCellStyle.VERTICAL_CENTER);// 上下居中
this.styleBody.setRightBorderColor(HSSFColor.BLACK.index);
this.styleBody.setBottomBorderColor(HSSFColor.BLACK.index);
this.styleBody.setBorderRight((short) 1);// 边框的大小
this.styleBody.setBorderBottom((short) 1);// 边框的大小
break;
}
}
/**
* 导出表格 无返回
*
* @param listParamColumn 表头数据
* @param dataList 行内数据
* @param FilePath 保存路径
* @param flag
* @param rowFlag
* @throws Exception
*/
public void exportExcel(List<Column> listParamColumn, List<T> dataList, String FilePath, boolean flag, boolean rowFlag) throws Exception {
splitDataToSheets(dataList, listParamColumn, flag, rowFlag);
save(this.workbook, FilePath);
}
/**
* 导出表格 无返回
*
* @param listParamColumn 表头数据
* @param dataList 行内数据
* @param materialColumn 物料基础信息合并单元格
* @param fileHelper 上传附件
* @param vendorColumn 供应商合并单元格
* @param vendorList 供应商数据
* @param awardVO 计算定选价格
* @param paramsCode 上传附件的场景值
* @throws Exception
*/
public FileInfoVo awardExportExcel(List<Column> listParamColumn, List<T> dataList, HashMap<String, String> paramsCode, int materialColumn, FileHelper fileHelper, int vendorColumn, List<QuoteAmountVO> vendorList, AwardVO awardVO) throws Exception {
int dataCount = dataList.size();
int maxColumn = 65535;
int pieces = dataCount / maxColumn;
for (int i = 1; i <= pieces; i++) {
HSSFSheet sheet = this.workbook.createSheet(this.title + i);
List<T> subList = dataList.subList((i - 1) * maxColumn, i * maxColumn);
writeSheet(sheet, subList, listParamColumn, true, true);
}
String sheetName = this.title;
if (pieces != 0) {
sheetName = this.title + (pieces + 1);
}
HSSFSheet sheet = this.workbook.createSheet(sheetName);
writeSheet(sheet, dataList.subList(pieces * maxColumn, dataCount), listParamColumn, true, true);
// 构建结尾数据
List<AwardExportDTO> exportDTOList = Lists.newArrayList();
exportDTOList.add(new AwardExportDTO().setCode("currencyName").setName("报价币种"));
exportDTOList.add(new AwardExportDTO().setCode("adjustExtent").setName("首末总报价调整幅度(含税)"));
exportDTOList.add(new AwardExportDTO().setCode("adjustAmount").setName("首末总报价调整额度(含税)"));
exportDTOList.add(new AwardExportDTO().setCode("adjustExtentExclTax").setName("首末总报价调整幅度(未税)"));
exportDTOList.add(new AwardExportDTO().setCode("adjustAmountExclTax").setName("首末总报价调整额度(未税)"));
exportDTOList.add(new AwardExportDTO().setCode("amount").setName("含税报价总计"));
exportDTOList.add(new AwardExportDTO().setCode("amountExclTax").setName("未税报价总计"));
exportDTOList.add(new AwardExportDTO().setCode("awardAmount").setName("含税定选总计"));
exportDTOList.add(new AwardExportDTO().setCode("awardAmountExclTax").setName("未税定选总计"));
List<AwardLineVO> awardLineVOList = awardVO.getAwardLineVOList();
HashMap<Long, BigDecimal> vendorAmountMap = new HashMap<>(16);
HashMap<Long, BigDecimal> vendorAmountExclMap = new HashMap<>(16);
awardLineVOList.forEach(item -> {
List<AwardLineVendorVO> awardLineVendorVOS = item.getAwardLineVendorVOS();
for (AwardLineVendorVO vendorVO : awardLineVendorVOS) {
if (Objects.equals(vendorVO.getSelectedStatus(), BooleanEnum.TRUE.getCode())) {
Long vendorId = vendorVO.getParticipatingVendorId();
BigDecimal amount = vendorVO.getAmount();
if (vendorAmountMap.containsKey(vendorId)) {
BigDecimal addAmount = vendorAmountMap.get(vendorId);
amount = amount.add(addAmount);
}
vendorAmountMap.put(vendorId, amount);
BigDecimal amountExcl = vendorVO.getAmountExclTax();
if (vendorAmountExclMap.containsKey(vendorId)) {
BigDecimal addAmount = vendorAmountExclMap.get(vendorId);
amountExcl = amountExcl.add(addAmount);
}
vendorAmountExclMap.put(vendorId, amountExcl);
}
}
});
for (int i = 0; i < exportDTOList.size(); i++) {
AwardExportDTO awardExportDTO = exportDTOList.get(i);
// 拼接 后面的合并表格
int lastRowNum = sheet.getLastRowNum() + 1;
HSSFRow a = sheet.createRow(lastRowNum);
HSSFCell ac = a.createCell(1);
ac.setCellValue(awardExportDTO.getName());
CellRangeAddress one = new CellRangeAddress(lastRowNum, lastRowNum, 1, materialColumn);
sheet.addMergedRegion(one);
for (int j = 0; j < vendorList.size(); j++) {
QuoteAmountVO amountVO = vendorList.get(j);
int firstCol = materialColumn + 1 + j * vendorColumn;
HSSFCell ab = a.createCell(firstCol);
String column = fillColumn(awardExportDTO.getCode(), amountVO);
if (Objects.equals(awardExportDTO.getCode(), "amount") || Objects.equals(awardExportDTO.getCode(), "amountExclTax")) {
BigDecimal columnDecimal = new BigDecimal(column).setScale(amountVO.getCurrencyAmountPrecision(), RoundingMode.HALF_UP);
column = amountVO.getCurrencySymbol() + columnDecimal;
}
if (Objects.equals(column, "0E-8") || Objects.equals(column, "null")) {
column = "--";
}
if (Objects.equals(awardExportDTO.getCode(), "adjustExtent") || Objects.equals(awardExportDTO.getCode(), "adjustExtentExclTax")) {
if (Objects.equals(column, "0E-8") || Objects.equals(column, "null") || Objects.equals(column, "--")) {
column = "0%";
} else {
column = column + "%";
}
}
Long vendorId = amountVO.getParticipatingVendorId();
if (Objects.equals(awardExportDTO.getCode(), "awardAmount") && vendorAmountMap.containsKey(vendorId)) {
BigDecimal columnDecimal = vendorAmountMap.get(vendorId).setScale(amountVO.getCurrencyAmountPrecision(), RoundingMode.HALF_UP);
column = amountVO.getCurrencySymbol() + columnDecimal;
}
if (Objects.equals(awardExportDTO.getCode(), "awardAmountExclTax") && vendorAmountExclMap.containsKey(vendorId)) {
BigDecimal columnDecimal = vendorAmountExclMap.get(vendorId).setScale(amountVO.getCurrencyAmountPrecision(), RoundingMode.HALF_UP);
column = amountVO.getCurrencySymbol() + columnDecimal;
}
ab.setCellValue(column);
CellRangeAddress two = new CellRangeAddress(lastRowNum, lastRowNum, firstCol, firstCol + vendorColumn - 1);
sheet.addMergedRegion(two);
}
}
ByteArrayOutputStream fileOutputStream = new ByteArrayOutputStream();
workbook.write(fileOutputStream);
String fileName = sheet.getSheetName() + "-" + DateUtil.format(new Date(), "yyyyMMdd-HHmmss") + ExcelTypeEnum.XLSX.getValue();
return fileHelper.uploadRemoteFile(awardVO.getTenantId(), paramsCode.get("moduleCode"), paramsCode.get("sceneCode"), null, fileName, fileOutputStream.toByteArray());
// save(this.workbook, FilePath);
}
private String fillColumn(String code, QuoteAmountVO amountVO) throws IllegalAccessException {
Class<?> targetClass = amountVO.getClass();
//获取类中所有字段集合
Field[] fields = targetClass.getDeclaredFields();
//遍历字段
for (Field field : fields) {
//取消安全检查,允许防伪private修饰的
field.setAccessible(true);
String name = field.getName();
//字段名和值
if (Objects.equals(code, name)) {
return String.valueOf(field.get(amountVO));
}
}
return "";
}
/**
* 返回workbook
*
* @param listTpamscolumn 表头数据
* @param datas 行内数据
* @param flag 是否写入行内数据
* @return
* @throws Exception
*/
public HSSFWorkbook exportWorkbook(List<Column> listTpamscolumn, List<T> datas, boolean flag) throws Exception {
splitDataToSheets(datas, listTpamscolumn, flag, false);
return this.workbook;
}
/**
* 导出表格 有返回值
*
* @param listTpamscolumn 表头数据
* @param datas 行内数据
* @param flag 只输出表头数据
* @param rowFlag
* @return
* @throws Exception
*/
public InputStream exportExcel(List<Column> listTpamscolumn, List<T> datas, boolean flag, boolean rowFlag) throws Exception {
splitDataToSheets(datas, listTpamscolumn, flag, rowFlag);
return save(this.workbook);
}
/**
* 导出Excel,适用于web导出excel
*
* @param sheet excel
* @param data 行内数据
* @param listParamsColumn 表头数据
* @param flag 只输出表头数据
* @param rowFlag 输出展示数据的结构(表头下面行的数据)
* @throws Exception
*/
private void writeSheet(HSSFSheet sheet, List<T> data, List<Column> listParamsColumn, boolean flag, boolean rowFlag) throws Exception {
sheet.setDefaultColumnWidth(colWidth);
sheet.setDefaultRowHeightInPoints(rowHeight);
sheet = createHead(sheet, listParamsColumn.get(0).getTotalRow(), listParamsColumn.get(0).getTotalCol());
createHead(listParamsColumn, sheet, 0);
//控制是否 bug修复:每次写入行数据时,总是漏第一个条数据 rowIndex 错误
if (flag) {
writeSheetContent(listParamsColumn, data, sheet, listParamsColumn.get(0).getTotalRow() + 1, rowFlag);
}
}
/**
* 拆分sheet,因为每个sheet不能超过65535,否则会报异常
*
* @param data 行内数据
* @param listParamsColumn 表头数据
* @param flag 只输出表头数据
* @param rowFlag 输出展示数据的结构(表头下面行的数据)
* @throws Exception
*/
private void splitDataToSheets(List<T> data, List<Column> listParamsColumn, boolean flag, boolean rowFlag) throws Exception {
int dataCount = data.size();
int maxColumn = 65535;
int pieces = dataCount / maxColumn;
for (int i = 1; i <= pieces; i++) {
HSSFSheet sheet = this.workbook.createSheet(this.title + i);
List<T> subList = data.subList((i - 1) * maxColumn, i * maxColumn);
writeSheet(sheet, subList, listParamsColumn, flag, rowFlag);
}
HSSFSheet sheet = this.workbook.createSheet(this.title + (pieces));
writeSheet(sheet, data.subList(pieces * maxColumn, dataCount), listParamsColumn, flag, rowFlag);
}
/**
* 把数据写入到单元格
*
* @param listParamsColumn 表头数据
* @param dataList 行内数据
* @param sheet 工作表(excel分页)
* @throws Exception void
*/
private void writeSheetContent(List<Column> listParamsColumn, List<T> dataList, HSSFSheet sheet, int rowIndex, boolean rowFlag) throws Exception {
HSSFRow row = null;
List<Column> listCol = new ArrayList<>();
rowFlag = false;
if (rowFlag) {//暂时没有用 后面扩展用
for (int i = 0, index = rowIndex; i < dataList.size(); i++, index++) {
row = sheet.createRow(index);//创建行
for (int j = 0; j < listParamsColumn.size(); j++) {
createColl(row, j, listParamsColumn.get(j).getFieldName(), dataList.get(i));
}
}
} else {
getColumnList(listParamsColumn, listCol);
for (int i = 0, index = rowIndex; i < dataList.size(); i++, index++) {
row = sheet.createRow(index);//创建行
for (int j = 0; j < listCol.size(); j++) {
Column c = listCol.get(j);
createCol(row, c, dataList.get(i));
}
}
}
}
/**
* 根据list 来创建单元格 暂时没有用
*
* @param row
* @param j
* @param finame
* @param t
*/
private void createColl(HSSFRow row, int j, String finame, T t) {
HSSFCell cell = row.createCell(j); //创建单元格
cell.setCellStyle(this.styleBody); //设置单元格样式
String text = "";
if (t instanceof List) {
List<Map> temp = (List<Map>) t;
if (j >= temp.size()) return;
text = String.valueOf(temp.get(j).get(finame) == null ? "" : temp.get(j).get(finame));
}
HSSFRichTextString richString = new HSSFRichTextString(text);
cell.setCellValue(richString);
}
/**
* 把column的columnList整理成一个list<column> 过滤表头的脏数据
*
* @param list 表头数据
* @param listCol 返回新的list
* @return List<column>
*/
private void getColumnList(List<Column> list, List<Column> listCol) {
for (int i = 0; i < list.size(); i++) {
if (list.get(i).getFieldName() != null) {
listCol.add(list.get(i));
}
List<Column> listChilren = list.get(i).getListTpamscolumn();
if (listChilren.size() > 0) {
getColumnList(listChilren, listCol);
}
}
}
/**
* 保存Excel到InputStream,此方法适合web导出excel
*
* @param workbook
* @return
*/
private InputStream save(HSSFWorkbook workbook) {
ByteArrayOutputStream bos = new ByteArrayOutputStream();
try {
workbook.write(bos);
InputStream bis = new ByteArrayInputStream(bos.toByteArray());
return bis;
} catch (Exception e) {
e.printStackTrace();
throw new RuntimeException(e);
}
}
/**
* 保存excel到本机指定的路径
*
* @param workbook
* @param filePath
* @throws IOException
*/
private void save(HSSFWorkbook workbook, String filePath) {
File file = new File(filePath);
if (!file.getParentFile().exists()) {
file.getParentFile().mkdirs();
}
FileOutputStream fOut = null;
try {
fOut = new FileOutputStream(file);
workbook.write(fOut);
fOut.flush();
} catch (Exception e) {
e.printStackTrace();
}
try {
if (null != fOut) fOut.close();
} catch (Exception e1) {
}
}
/**
* 创建行
*
* @param row Excel对应的行
* @param tpamscolumn 当前单元格属性
* @param v
* @param j
* @return
* @throws Exception
*/
public int createRowVal(HSSFRow row, Column tpamscolumn, T v, int j) throws Exception {
//遍历标题
if (tpamscolumn.getListTpamscolumn() != null && tpamscolumn.getListTpamscolumn().size() > 0) {
for (int i = 0; i < tpamscolumn.getListTpamscolumn().size(); i++) {
createRowVal(row, tpamscolumn.getListTpamscolumn().get(i), v, j);
}
} else {
createCol(row, tpamscolumn, v);
}
return j;
}
/**
* 创建单元格
*
* @param row Excel对应的行
* @param tpamscolumn 当前单元格对象
* @param v
* @throws Exception
*/
public void createCol(HSSFRow row, Column tpamscolumn, T v) throws Exception {
HSSFCell cell = row.createCell(tpamscolumn.getCol()); //创建单元格
cell.setCellStyle(this.styleBody); //设置单元格样式
final Object[] value = {null};
if (v instanceof Map) {
Map m = (Map) v;
m.forEach((k, val) -> {
if (k.equals(tpamscolumn.getFieldName()) && !tpamscolumn.isHasChilren()) {
value[0] = val;
}
});
} else {
Class<?> cls = v.getClass();// 拿到该类
Field[] fields = cls.getDeclaredFields();// 获取实体类的所有属性,返回Field数组
for (int i = 0; i < fields.length; i++) {
Field f = fields[i];
f.setAccessible(true); // 设置些属性是可以访问的
if (tpamscolumn.getFieldName().equals(f.getName()) && !tpamscolumn.isHasChilren())// && !tpamscolumn.isHasChilren()
value[0] = f.get(v);
if (value[0] instanceof Date)
value[0] = parseDate((Date) value[0]);
}
}
if (value[0] != null) {
HSSFRichTextString richString = new HSSFRichTextString(value[0].toString());
cell.setCellValue(richString);
}
}
/**
* 时间转换
*
* @param date
* @return String
*/
private String parseDate(Date date) {
String dateStr = "";
try {
dateStr = this.sdf.format(date);
} catch (Exception e) {
e.printStackTrace();
}
return dateStr;
}
/**
* 根据数据的行数和列数,在excel创建单元格cell
*
* @param sheetCo excel分页
* @param r excel 行数
* @param c excel 列数
* @return
*/
public HSSFSheet createHead(HSSFSheet sheetCo, int r, int c) {
for (int i = 0; i < r; i++) {
HSSFRow row = sheetCo.createRow(i);
for (int j = 0; j < c; j++) {
HSSFCell cell = row.createCell(j);
}
}
return sheetCo;
}
/**
* 使用递归 在excel写入表头数据 支持单级,多级表头的创建
*
* @param listTpamscolumn 表头数据
* @param sheetCo 哪个分页
* @param rowIndex 当前Excel的第几行
*/
public void createHead(List<Column> listTpamscolumn, HSSFSheet sheetCo, int rowIndex) {
HSSFRow row = sheetCo.getRow(rowIndex);
// if(row == null)row = sheetCo.createRow(rowIndex);
int len = listTpamscolumn.size();//当前行 有多少列
for (int i = 0; i < len; i++) {//i是headers的索引,n是Excel的索引 多级表头
Column tpamscolumn = listTpamscolumn.get(i);
//创建这一行的第几列单元格
int r = tpamscolumn.getRow();
int rLen = tpamscolumn.getrLen();
int c = tpamscolumn.getCol();
int cLen = tpamscolumn.getcLen();
int endR = r + rLen;
int endC = c + cLen;
if (endC > c) {
endC--;
}
HSSFCell cell = row.getCell(c);
// if( null == cell)cell = row.createCell(c);
HSSFRichTextString text = new HSSFRichTextString(tpamscolumn.getContent());
cell.setCellStyle(this.styleHead); //设置表头样式
cell.setCellValue(text);
// 合并单元格
CellRangeAddress cra = new CellRangeAddress(r, endR, c, endC);
sheetCo.addMergedRegion(cra);
// 使用RegionUtil类为合并后的单元格添加边框
RegionUtil.setBorderBottom(1, cra, sheetCo, this.workbook); // 下边框
RegionUtil.setBorderLeft(1, cra, sheetCo, this.workbook); // 左边框
RegionUtil.setBorderRight(1, cra, sheetCo, this.workbook); // 有边框
if (tpamscolumn.isHasChilren()) {
rowIndex = r + 1;
createHead(tpamscolumn.getListTpamscolumn(), sheetCo, rowIndex);
}
}
}
/**
* 转换成column对象
* 支持List<T>的数据结构:map String ,只能是单级的数据
*
* @param list 需要转换的数据
* @return
*/
public List<Column> columnTransformer(List<T> list) {
List<Column> lc = new ArrayList<>();
if (list.get(0) instanceof Map) {
final int[] i = {1};
for (Map<String, String> m : (List<Map<String, String>>) list) {
m.forEach((k, val) -> {
Column tpamscolumn = new Column();
tpamscolumn.setId(String.valueOf(i[0]));
tpamscolumn.setPid("0");
tpamscolumn.setContent(k);
tpamscolumn.setFieldName(val);
lc.add(tpamscolumn);
i[0]++;
});
}
} else {
int i = 1;
for (String s : (List<String>) list) {
Column tpamscolumn = new Column();
tpamscolumn.setId(String.valueOf(i));
tpamscolumn.setPid("0");
tpamscolumn.setContent(s);
tpamscolumn.setFieldName(null);
lc.add(tpamscolumn);
i++;
}
}
setParm(lc, "0");//处理一下
List<Column> s = TreeTool.buildByRecursive(lc, "0");
setColNum(lc, s, s);
return s;
}
/**
* 转换成column对象 返回tree数据结构
* 支持:List<map>、某个具体对象(entity)数据的转换
*
* @param list 需要转换的数据
* @param id 当前节点id 字段的名称 主键
* @param pid 父节点id 字段的名称
* @param content 填写表头单元格内容的 字段名称
* @param fielName 填写行内数据对的 字段名称
* @param rootid rootid的值
* @return
* @throws Exception
*/
public List<Column> columnTransformer(List<T> list, String id, String pid, String content, String fielName, String rootid) throws Exception {
// 结果
List<Column> lc = new ArrayList<>();
if (list.get(0) instanceof Map) {
for (Map m : (List<Map>) list) {
Column tpamscolumn = new Column();
m.forEach((k, val) -> {//java8 以上的遍历方式
if (id.equals(k))
tpamscolumn.setId(String.valueOf(val));
if (pid.equals(k)) tpamscolumn.setPid((String) val);
if (content.equals(k)) tpamscolumn.setContent((String) val);
if (fielName.equals(k) && fielName != null) tpamscolumn.setFieldName((String) val);
});
lc.add(tpamscolumn);
}
} else {
for (T t : list) {//反射
Column tpamscolumn = new Column();
Class cls = t.getClass();
Field[] fs = cls.getDeclaredFields();
for (int i = 0; i < fs.length; i++) {
Field f = fs[i];
f.setAccessible(true); // 设置些属性是可以访问的
if (id.equals(f.getName()) && f.get(t) != null)
tpamscolumn.setId(f.get(t).toString());
if (pid.equals(f.getName()) && f.get(t) != null)
tpamscolumn.setPid(f.get(t).toString());
// if (pid.equals(f.getName()) && ( f.get(t) == null || "".equals(f.get(t)))) tpamscolumn.setPid("0");
if (content.equals(f.getName()) && f.get(t) != null)
tpamscolumn.setContent(f.get(t).toString());
if (f.get(t) != null && fielName != null && fielName.equals(f.getName()))
tpamscolumn.setFieldName(f.get(t).toString());
}
lc.add(tpamscolumn);
}
}
setParm(lc, rootid);//处理一下
List<Column> s = TreeTool.buildByRecursive(lc, rootid);
setColNum(lc, s, s);
return s;
}
/**
* 设置基础的参数
*
* @param list
*/
public static void setParm(List<Column> list, String rootid) {
int row = 0;//excel第几行
int rLen = 0; //excel 跨多少行
// 总行数
int totalRow = TreeTool.getMaxStep(list);
// 总列数
int totalCol = TreeTool.getDownChilren(list, rootid);
for (int i = 0; i < list.size(); i++) {
Column poit = list.get(i);
// 往上遍历tree
int tree_step = TreeTool.getTreeStep(list, poit.getPid(), 0);
poit.setTree_step(tree_step);
// 设置第几行
poit.setRow(tree_step);
//判断是否有节点
boolean hasCh = TreeTool.hasChild(list, poit);
poit.setHasChilren(hasCh);
if (hasCh) {
poit.setrLen(0);//设置跨多少行
} else {
if (tree_step < totalRow) {
rLen = totalRow - tree_step;
}
poit.setrLen(rLen);
}
// boolean flag=false;//控制只有root 节点才有总的行数信息
// if(rootid == null && rootid == poit.getId() )flag = true;
// if(rootid != null && rootid.equals(poit.getId()))flag = true;
// if(flag){
//
// }
poit.setTotalRow(totalRow);
poit.setTotalCol(totalCol);
}
}
/**
* 设置基础的参数
*
* @param list 所有list数据,一条一条
* @param treeList 转成tree结构的list
*/
public static void setColNum(List<Column> list, List<Column> treeList, List<Column> flist) {
// int col = pcIndex;//excel第几列
// int cLen ;//xcel跨多少列
List<Column> new_list = new ArrayList<>();//新的遍历list
for (int i = 0; i < treeList.size(); i++) {
Column poit = treeList.get(i);
// String temp_id = TreeTool.getStepFid(list,poit.getId() ,1);
int col = TreeTool.getFCol(list, poit.getPid()).getCol();
int brotherCol = TreeTool.getBrotherChilNum(list, poit);
poit.setCol(col + brotherCol);
int cLen = TreeTool.getDownChilren(list, poit.getId());
if (cLen <= 1) cLen = 0;
// else cLen--;
poit.setcLen(cLen);//设置跨多少列
if (poit.getListTpamscolumn().size() > 0) {
new_list.addAll(poit.getListTpamscolumn());
}
}
if (new_list.size() > 0) {
setColNum(list, new_list, flist);
}
}
//========上部分是导出excel的使用(生成excel),下部分是解析excel,由于excel导入==================================================================================================================================
/**
* 根据HSSFCell类型设置数据
*
* @param cell 单元格
* @return
*/
public static String getCellFormatValue(Cell cell) {
String cellvalue = "";
if (cell != null) {
switch (cell.getCellType()) { // 判断当前Cell的Type
case HSSFCell.CELL_TYPE_NUMERIC: // 如果当前Cell的Type为NUMERIC
case HSSFCell.CELL_TYPE_FORMULA: {
// 判断当前的cell是否为Date
if (HSSFDateUtil.isCellDateFormatted(cell)) {
Date date = cell.getDateCellValue();
SimpleDateFormat sdf = new SimpleDateFormat("yyyy-MM-dd");
cellvalue = sdf.format(date);
} else { // 如果是纯数字
cellvalue = String.valueOf(cell.getNumericCellValue());
}
break;
}
case HSSFCell.CELL_TYPE_STRING: // 如果当前Cell的Type为STRIN
// 取得当前的Cell字符串
cellvalue = cell.getRichStringCellValue().getString();
break;
default: // 默认的Cell值
cellvalue = "";
}
} else {
cellvalue = "";
}
return cellvalue;
}
/**
* 描述:根据文件后缀,自适应上传文件的版本
*
* @param inStr,fileName
* @return
* @throws Exception
*/
public static Workbook getWorkbookType(InputStream inStr, String fileName) throws Exception {
Workbook wb = null;
String fileType = fileName.substring(fileName.lastIndexOf("."));
if (".xls".equals(fileType)) {
wb = new HSSFWorkbook(inStr); //2003-
} else if (".xlsx".equals(fileType)) {
wb = new XSSFWorkbook(inStr); //2007+
} else {
throw new Exception("导入格式错误");
}
return wb;
}
/**
* 获取单元格数据内容为字符串类型的数据
*
* @param cell Excel单元格
* @return String 单元格数据内容
*/
public static String getStringCellValue(Cell cell) {
String strCell = "";
if (cell == null) {
return "";
}
switch (cell.getCellType()) {
case Cell.CELL_TYPE_STRING:
strCell = cell.getStringCellValue().trim();
break;
case Cell.CELL_TYPE_NUMERIC:
strCell = String.valueOf(cell.getNumericCellValue()).trim();
break;
case Cell.CELL_TYPE_BOOLEAN:
strCell = String.valueOf(cell.getBooleanCellValue()).trim();
break;
case Cell.CELL_TYPE_BLANK:
strCell = "";
break;
default:
strCell = "";
break;
}
if (strCell.equals("") || strCell == null) {
return "";
}
return strCell;
}
/**
* 判断指定的单元格是否是合并单元格
*
* @param sheet
* @param row 行下标
* @param column 列下标
* @return
*/
public boolean isMergedRegion(Sheet sheet, int row, int column) {
int sheetMergeCount = sheet.getNumMergedRegions();
for (int i = 0; i < sheetMergeCount; i++) {
CellRangeAddress range = sheet.getMergedRegion(i);
int firstColumn = range.getFirstColumn();
int lastColumn = range.getLastColumn();
int firstRow = range.getFirstRow();
int lastRow = range.getLastRow();
if (row >= firstRow && row <= lastRow) {
if (column >= firstColumn && column <= lastColumn) {
return true;
}
}
}
return false;
}
/**
* 获取合并单元格的值
*
* @param sheet
* @param row 行下标
* @param column 列下标
* @return
*/
public String getMergedRegionValue(Sheet sheet, int row, int column) {
int sheetMergeCount = sheet.getNumMergedRegions();
for (int i = 0; i < sheetMergeCount; i++) {
CellRangeAddress ca = sheet.getMergedRegion(i);
int firstColumn = ca.getFirstColumn();
int lastColumn = ca.getLastColumn();
int firstRow = ca.getFirstRow();
int lastRow = ca.getLastRow();
if (row >= firstRow && row <= lastRow) {
if (column >= firstColumn && column <= lastColumn) {
Row fRow = sheet.getRow(firstRow);
Cell fCell = fRow.getCell(firstColumn);
return getStringCellValue(fCell);
}
}
}
return "";
}
/**
* 获取excel的值 返回的 List<List<String>>的数据结构
*
* @param fileUrl 文件路径
* @param sheetNum 工作表(第几分页[1,2,3.....])
* @return List<List < String>>
*/
public List<List<String>> getExcelValues(String fileUrl, int sheetNum) throws Exception {
List<List<String>> values = new ArrayList<List<String>>();
File file = new File(fileUrl);
InputStream is = new FileInputStream(file);
Workbook workbook = WorkbookFactory.create(is);
int sheetCount = sheetNum - 1; //workbook.getNumberOfSheets();//sheet 数量,可以只读取手动指定的sheet页
//int sheetCount1= workbook.getNumberOfSheets();
Sheet sheet = workbook.getSheetAt(sheetCount); //读取第几个工作表sheet
int rowNum = sheet.getLastRowNum();//有多少行
for (int i = 1; i <= rowNum; i++) {
Row row = sheet.getRow(i);//第i行
if (row == null) {//过滤空行
continue;
}
List<String> list = new ArrayList<>();
int colCount = sheet.getRow(0).getLastCellNum();//用表头去算有多少列,不然从下面的行计算列的话,空的就不算了
for (int j = 0; j < colCount; j++) {//第j列://+1是因为最后一列是空 也算进去
Cell cell = row.getCell(j);
String cellValue;
boolean isMerge = false;
if (cell != null) {
isMerge = isMergedRegion(sheet, i, cell.getColumnIndex());
}
//判断是否具有合并单元格
if (isMerge) {
cellValue = getMergedRegionValue(sheet, row.getRowNum(), cell.getColumnIndex());
} else {
cellValue = getStringCellValue(cell);
}
list.add(cellValue);
}
values.add(list);
}
return values;
}
/**
* 判断整行是否为空
*
* @param row excel得行对象
* @param maxRow 有效值得最大列数
*/
private static boolean CheckRowNull(Row row, int maxRow) {
int num = 0;
for (int j = 0; j < maxRow; j++) {
Cell cell = row.getCell(j);
if (cell == null || cell.equals("") || cell.getCellType() == HSSFCell.CELL_TYPE_BLANK) {
num++;
}
}
if (maxRow == num) return true;
return false;
}
/**
* 根据sheet数获取excel的值 返回List<List<Map<String,String>>>的数据结构
*
* @param fileUrl 文件路径
* @param sheetNum 工作表(第几分页[1,2,3.....])
* @return List<List < Map < String, String>>>
*/
public List<List<Map<String, String>>> getExcelMapVal(String fileUrl, int sheetNum) throws Exception {
List<List<Map<String, String>>> values = new ArrayList<List<Map<String, String>>>();
File file = new File(fileUrl);
InputStream is = new FileInputStream(file);
Workbook workbook = WorkbookFactory.create(is);
int sheetCount = sheetNum - 1; //workbook.getNumberOfSheets();//sheet 数量,可以只读取手动指定的sheet页
//int sheetCount1= workbook.getNumberOfSheets();
Sheet sheet = workbook.getSheetAt(sheetCount); //读取第几个工作表sheet
int rowNum = sheet.getLastRowNum();//有多少行
Row rowTitle = sheet.getRow(0);//第i行
int colCount = sheet.getRow(0).getLastCellNum();//用表头去算有多少列,不然从下面的行计算列的话,空的就不算了
for (int i = 1; i <= rowNum; i++) {
Row row = sheet.getRow(i);//第i行
if (row == null || CheckRowNull(row, colCount)) {//过滤空行
continue;
}
List<Map<String, String>> list = new ArrayList<Map<String, String>>();
for (int j = 0; j < colCount; j++) {//第j列://+1是因为最后一列是空 也算进去
Map<String, String> map = new HashMap<>();
Cell cell = row.getCell(j);
Cell cellTitle = rowTitle.getCell(j);
String cellValue;
String cellKey = getStringCellValue(cellTitle);
boolean isMerge = false;
if (cell != null) {
isMerge = isMergedRegion(sheet, i, cell.getColumnIndex());
}
//判断是否具有合并单元格
if (isMerge) {
cellValue = getMergedRegionValue(sheet, row.getRowNum(), cell.getColumnIndex());
} else {
cellValue = getStringCellValue(cell);
}
map.put(cellKey, cellValue);
list.add(map);
}
values.add(list);
}
return values;
}
/**
* 获取当前excel的工作表sheet总数
*
* @param fileUrl
* @return
* @throws Exception
*/
public int hasSheetCount(String fileUrl) throws Exception {
File file = new File(fileUrl);
InputStream is = new FileInputStream(file);
Workbook workbook = WorkbookFactory.create(is);
int sheetCount = workbook.getNumberOfSheets();
return sheetCount;
}
}
/**
* 表头的实体类: 在具体的项目里,可以是你从数据库里查询出来的数据
*/
@Data
public class TitleEntity {
public String t_id;
public String t_pid;
public String t_content;
public String t_fileName;
public TitleEntity() {
}
public TitleEntity(String t_id, String t_pid, String t_content, String t_fileName) {
this.t_id = t_id;
this.t_pid = t_pid;
this.t_content = t_content;
this.t_fileName = t_fileName;
}
}
import java.util.ArrayList;
import java.util.Collections;
import java.util.Iterator;
import java.util.List;
/**
* 处理tree结构的数据 工具类
*/
public class TreeTool {
/**
* 传入的id 必须存在list集合里
* 获取某节点的深度
*
* @param list
* @param id 根节点
* @param step
* @return
*/
public static int getTreeStep(List<Column> list, String id, int step) {
if ("".equals(id) || null == id) return step;
for (Column cc : list) {
if (id.equals(cc.getId())) {
int temp = step + 1;
return getTreeStep(list, cc.getPid(), temp);
}
}
return step;
}
/**
* 遍历所有数据 获取树最大的深度
*
* @param list
* @return
*/
public static int getMaxStep(List<Column> list) {
List<Integer> nums = new ArrayList<Integer>();
for (Column cc : list) {
nums.add(getTreeStep(list, cc.getId(), 0));
}
return Collections.max(nums);
}
/**
* 获取最底部子节点的个数 所有叶子节点个数
*
* @param list
* @param did
* @return
*/
public static int getDownChilren(List<Column> list, String did) {
int sum = 0;
for (Column cc : list) {
if (did.equals(cc.getPid())) {
sum++;
//判断该节点 是否有子节点
if (hasChild(list, cc)) {
sum += getDownChilren(list, cc.getId()) - 1;
}
}
}
return sum;
}
/**
* 获取父节点
*
* @param list 所有的list数据,一条一条
* @param did 当前节点id
* @return
*/
public static Column getFCol(List<Column> list, String did) {
for (Column cc : list) {
if (did != null && did.equals(cc.getId())) {
return cc;
}
if (did == null && did == cc.getId()) {
return cc;
}
}
return new Column() {{
setCol(0);
setRow(0);
}};
}
/**
* 获取兄弟节点个数 这个必须是有排序的
*
* @param list 所有的list数据,一条一条
* @param column 当前节点信息
* @return
*/
public static int getBrotherChilNum(List<Column> list, Column column) {
int sum = 0;
for (Column cc : list) {
if (column.getId().equals(cc.getId())) {
break;
}
if (!column.getPid().equals(cc.getPid())) {
continue;
}
int temp = getDownChilren(list, cc.getId());
if (temp == 0 || temp == 1)
sum++;
else
sum += temp;
}
return sum;
}
/**
* 根据某节点的第几层的父节点id
*
* @param list 所有的list数据,一条一条
* @param id 当前节点id
* @param step 第几层(深度 从零开始)
* @return
*/
public static String getStepFid(List<Column> list, String id, int step) {
String f_id = null;
for (Column cc : list) {
if (id.equals(cc.getId())) {
int cstep = getTreeStep(list, cc.getId(), 0);
if (step == cstep) {
return id;
}
int fstep = getTreeStep(list, cc.getPid(), 0);
if (step == fstep) {
f_id = cc.getPid();
break;
} else {
getStepFid(list, cc.getPid(), step);
}
}
}
return f_id;
}
/**
* 判断是否有子节点
*
* @param list 遍历的数据
* @param node 某个节点
* @return
*/
public static boolean hasChild(List<Column> list, Column node) {
return getChildList(list, node).size() > 0 ? true : false;
}
/**
* 得到子节点列表
*
* @param list 遍历的数据
* @param node 某个节点
* @return
*/
public static List<Column> getChildList(List<Column> list, Column node) {
List<Column> nodeList = new ArrayList<Column>();
Iterator<Column> it = list.iterator();
while (it.hasNext()) {
Column n = (Column) it.next();
if (n.getPid() != null && n.getPid().equals(node.getId())) {
nodeList.add(n);
}
}
return nodeList;
}
/**
* 使用递归方法建树
*
* @param treeNodes
* @return
*/
public static List<Column> buildByRecursive(List<Column> treeNodes, String rootID) {
List<Column> trees = new ArrayList<>();
boolean flag = false;
boolean sflag = false;
for (Column treeNode : treeNodes) {
if ((rootID == null && rootID == treeNode.getId())) {
flag = true;
}
if (rootID != null && rootID.equals(treeNode.getId())) {
flag = true;
}
if (flag) {
trees.add(findChildren(treeNode, treeNodes));
flag = false;
}
}
if (trees.size() <= 0) {
for (Column treeNode : treeNodes) {
if ((rootID == null && rootID == treeNode.getPid())) {
sflag = true;
}
if (rootID != null && rootID.equals(treeNode.getPid())) {
sflag = true;
}
if (sflag) {
trees.add(findChildren(treeNode, treeNodes));
sflag = false;
}
}
}
return trees;
}
/**
* 递归查找子节点
*
* @param treeNodes
* @return
*/
public static Column findChildren(Column treeNode, List<Column> treeNodes) {
for (Column it : treeNodes) {
if (treeNode.getId().equals(it.getPid())) {
if (treeNode.getListTpamscolumn() == null) {
treeNode.setListTpamscolumn(new ArrayList<Column>());
}
treeNode.getListTpamscolumn().add(findChildren(it, treeNodes));
}
}
return treeNode;
}
}
import java.util.ArrayList;
import java.util.HashMap;
import java.util.List;
import java.util.Map;
/**
* 测试类
*/
public class TextMain {
public static void main(String[] args) throws Exception {
// 单级的表头
// Map<String,String> map=new HashMap<String,String>();
// map.put("登录名","u_login_id");
// Map<String,String> map1=new HashMap<String,String>();
// map1.put("用户名","u_name");
// Map<String,String> map2=new HashMap<String,String>();
// map2.put("角色","u_role");
// Map<String,String> map3=new HashMap<String,String>();
// map3.put("部门","u_dep");//d_name
// Map<String,String> map4=new HashMap<String,String>();
// map4.put("用户类型","u_type");
// List<Map<String,String>> titleList=new ArrayList<>();
// titleList.add(map); titleList.add(map1); titleList.add(map2); titleList.add(map3); titleList.add(map4);
// //单级的 行内数据
// List<Map<String,String>> rowList=new ArrayList<>();
// for(int i=0;i<7;i++){
// Map m= new HashMap<String,String>();
// m.put("u_login_id","登录名"+i); m.put("u_name","张三"+i);
// m.put("u_role","角色"+i); m.put("u_dep","部门"+i);
// m.put("u_type","用户类型"+i);
// rowList.add(m);
// }
// ExcelTool excelTool = new ExcelTool("单级表头的表格",15,20);
// List<Column> titleData=excelTool.columnTransformer(titleList);
// excelTool.exportExcel(titleData,rowList,"D://outExcel.xls",true,false);
//List<Map>数据 多级表头,数据如下:
// 登录名 姓名 aa
// 角色 部门
// List<Map<String,String>> titleList=new ArrayList<>();
// Map<String,String> titleMap=new HashMap<String,String>();
// titleMap.put("id","11");titleMap.put("pid","0");titleMap.put("content","登录名");titleMap.put("fielName","u_login_id");
// Map<String,String> titleMap1=new HashMap<String,String>();
// titleMap1.put("id","1");titleMap1.put("pid","0");titleMap1.put("content","姓名");titleMap1.put("fielName","u_name");
// Map<String,String> titleMap2=new HashMap<String,String>();
// titleMap2.put("id","2");titleMap2.put("pid","0");titleMap2.put("content","角色加部门");titleMap2.put("fielName",null);
// Map<String,String> titleMap3=new HashMap<String,String>();
// titleMap3.put("id","3");titleMap3.put("pid","2");titleMap3.put("content","角色");titleMap3.put("fielName","u_role");
// Map<String,String> titleMap4=new HashMap<String,String>();
// titleMap4.put("id","4");titleMap4.put("pid","2");titleMap4.put("content","部门");titleMap4.put("fielName","u_dep");
// Map<String,String> titleMap5=new HashMap<String,String>();
// titleMap5.put("id","22");titleMap5.put("pid","0");titleMap5.put("content","角色加部门1");titleMap5.put("fielName",null);
// Map<String,String> titleMap6=new HashMap<String,String>();
// titleMap6.put("id","22_1");titleMap6.put("pid","22");titleMap6.put("content","角色1");titleMap6.put("fielName","u_role");
// Map<String,String> titleMap7=new HashMap<String,String>();
// titleMap7.put("id","22_2");titleMap7.put("pid","22");titleMap7.put("content","部门1");titleMap7.put("fielName","u_dep");
// titleList.add(titleMap); titleList.add(titleMap1); titleList.add(titleMap2); titleList.add(titleMap3); titleList.add(titleMap4);
// titleList.add(titleMap5); titleList.add(titleMap6); titleList.add(titleMap7);
// // 单级的 行内数据
// List<Map<String,String>> rowList=new ArrayList<>();
// for(int i=0;i<7;i++){
// Map m= new HashMap<String,String>();
// m.put("u_login_id","登录名"+i); m.put("u_name","张三"+i);
// m.put("u_role","角色"+i); m.put("u_dep","部门"+i);
// m.put("u_type","用户类型"+i);
// rowList.add(m);
// }
// ExcelTool excelTool = new ExcelTool("List<Map>数据 多级表头表格",20,20);
// List<Column> titleData=excelTool.columnTransformer(titleList,"id","pid","content","fielName","0");
// excelTool.exportExcel(titleData,rowList,"D://outExcel.xls",true,false);
//实体类(entity)数据 多级表头,数据如下:
// 登录名 姓名 aa
// 角色 部门
List<TitleEntity> titleList = new ArrayList<>();
TitleEntity titleEntity1 = new TitleEntity("1", "0", "序号", "num");
TitleEntity titleEntity2 = new TitleEntity("2", "0", "供应商1", null);
TitleEntity titleEntity5 = new TitleEntity("3", "0", "供应商2", null);
TitleEntity titleEntity3 = new TitleEntity("4", "2", "角色", "code");
TitleEntity titleEntity4 = new TitleEntity("5", "2", "部门", "name");
TitleEntity titleEntity6 = new TitleEntity("6", "3", "角色", "code");
TitleEntity titleEntity7 = new TitleEntity("7", "3", "部门", "name");
TitleEntity titleEntity8 = new TitleEntity("8", "0", "供应商3", null);
TitleEntity titleEntity9 = new TitleEntity("9", "8", "角色", "code");
TitleEntity titleEntity10 = new TitleEntity("10", "8", "部门", "name");
titleList.add(titleEntity1);
titleList.add(titleEntity2);
titleList.add(titleEntity3);
titleList.add(titleEntity4);
titleList.add(titleEntity5);
titleList.add(titleEntity6);
titleList.add(titleEntity7);
titleList.add(titleEntity8);
titleList.add(titleEntity9);
titleList.add(titleEntity10);
//单级的 行内数据
List<Map<String, String>> rowList = new ArrayList<>();
for (int i = 0; i < 7; i++) {
Map m = new HashMap<String, String>();
m.put("num", i);
m.put("code", i);
m.put("age", i);
m.put("name", "名称" + i);
rowList.add(m);
}
ExcelTool excelTool = new ExcelTool("多级表头表格", 20, 20);
List<Column> titleData = excelTool.columnTransformer(titleList, "t_id", "t_pid", "t_content", "t_fileName", "0");
excelTool.exportExcel(titleData, rowList, "/Users/mayuhan/Downloads/test.xls", true, true);
}
}