我遇到了同样的问题，它的根源是 PoiItemReader 中的 org.springframework.batch.item.excel.poi.PoiSheet 类。
问题发生在方法 public String[] getRow(final int rowNumber) 中，它获取 org.apache.poi.ss.usermodel.Row 对象并在检测到行中每一列的类型后将其转换为字符串数组。在这个方法中，我们有代码：
switch (cellType) {
case NUMERIC:
if (DateUtil.isCellDateFormatted(cell)) {
Date date = cell.getDateCellValue();
cells.add(String.valueOf(date.getTime()));
} else {
cells.add(String.valueOf(cell.getNumericCellValue()));
}
break;
case BOOLEAN:
cells.add(String.valueOf(cell.getBooleanCellValue()));
break;
case STRING:
case BLANK:
cells.add(cell.getStringCellValue());
break;
case ERROR:
cells.add(FormulaError.forInt(cell.getErrorCellValue()).getString());
break;
default:
throw new IllegalArgumentException("Cannot handle cells of type '" + cell.getCellTypeEnum() + "'");
}
其中标识为 NUMERIC 的单元格的处理是cells.add(String.valueOf(cell.getNumericCellValue()))。在这一行中，单元格值被转换为双精度（cell.getNumericCellValue()），而这个双精度被转换为字符串（String.valueOf()）。问题发生在 String.valueOf() 方法中，如果数字太大 (>=10000000) 或太小 (
作为cells.add(String.valueOf(cell.getNumericCellValue())) 行的替代方法，您可以使用
DataFormatter formatter = new DataFormatter();
cells.add(formatter.formatCellValue(cell));
这会将单元格的确切值作为字符串返回给您。但是，这也意味着您的十进制数字将取决于区域设置（您将收到来自为英国或印度配置的 Excel 中保存的文档中的字符串“2.5”和来自法国或巴西的字符串“2,5”）。
为了避免这种依赖，我们可以使用https://stackoverflow.com/a/25307973/9184574上提出的解决方案：
DecimalFormat df = new DecimalFormat("0", DecimalFormatSymbols.getInstance(Locale.ENGLISH));
df.setMaximumFractionDigits(340);
cells.add(df.format(cell.getNumericCellValue()));
这会将单元格转换为双精度，然后将其格式化为英文模式，无需科学记数法或在整数中添加“.0”。
我对 CustomPoiSheet（原始 PoiSheet 的小改编）的实现是：
class CustomPoiSheet implements Sheet {
protected final org.apache.poi.ss.usermodel.Sheet delegate;
private final int numberOfRows;
private final String name;
private FormulaEvaluator evaluator;
/**
* Constructor which takes the delegate sheet.
*
* @param delegate the apache POI sheet
*/
CustomPoiSheet(final org.apache.poi.ss.usermodel.Sheet delegate) {
super();
this.delegate = delegate;
this.numberOfRows = this.delegate.getLastRowNum() + 1;
this.name=this.delegate.getSheetName();
}
/**
* {@inheritDoc}
*/
@Override
public int getNumberOfRows() {
return this.numberOfRows;
}
/**
* {@inheritDoc}
*/
@Override
public String getName() {
return this.name;
}
/**
* {@inheritDoc}
*/
@Override
public String[] getRow(final int rowNumber) {
final Row row = this.delegate.getRow(rowNumber);
if (row == null) {
return null;
}
final List<String> cells = new LinkedList<>();
final int numberOfColumns = row.getLastCellNum();
for (int i = 0; i < numberOfColumns; i++) {
Cell cell = row.getCell(i);
CellType cellType = cell.getCellType();
if (cellType == CellType.FORMULA) {
FormulaEvaluator evaluator = getFormulaEvaluator();
if (evaluator == null) {
cells.add(cell.getCellFormula());
} else {
cellType = evaluator.evaluateFormulaCell(cell);
}
}
switch (cellType) {
case NUMERIC:
if (DateUtil.isCellDateFormatted(cell)) {
Date date = cell.getDateCellValue();
cells.add(String.valueOf(date.getTime()));
} else {
// Returns numeric value the closer possible to it's value and shown string, only formatting to english format
// It will result in an integer string (without decimal places) if the value is a integer, and will result
// on the double string without trailing zeros. It also suppress scientific notation
// Regards to https://stackoverflow.com/a/25307973/9184574
DecimalFormat df = new DecimalFormat("0", DecimalFormatSymbols.getInstance(Locale.ENGLISH));
df.setMaximumFractionDigits(340);
cells.add(df.format(cell.getNumericCellValue()));
//DataFormatter formatter = new DataFormatter();
//cells.add(formatter.formatCellValue(cell));
//cells.add(String.valueOf(cell.getNumericCellValue()));
}
break;
case BOOLEAN:
cells.add(String.valueOf(cell.getBooleanCellValue()));
break;
case STRING:
case BLANK:
cells.add(cell.getStringCellValue());
break;
case ERROR:
cells.add(FormulaError.forInt(cell.getErrorCellValue()).getString());
break;
default:
throw new IllegalArgumentException("Cannot handle cells of type '" + cell.getCellTypeEnum() + "'");
}
}
return cells.toArray(new String[0]);
}
private FormulaEvaluator getFormulaEvaluator() {
if (this.evaluator == null) {
this.evaluator = delegate.getWorkbook().getCreationHelper().createFormulaEvaluator();
}
return this.evaluator;
}
}
以及我调用 CustomPoiSheet 的 CustomPoiItemReader（对原始 PoiItemReader 的小改编）的实现：
public class CustomPoiItemReader<T> extends AbstractExcelItemReader<T> {
private Workbook workbook;
@Override
protected Sheet getSheet(final int sheet) {
return new CustomPoiSheet(this.workbook.getSheetAt(sheet));
}
public CustomPoiItemReader(){
super();
}
@Override
protected int getNumberOfSheets() {
return this.workbook.getNumberOfSheets();
}
@Override
protected void doClose() throws Exception {
super.doClose();
if (this.workbook != null) {
this.workbook.close();
}
this.workbook=null;
}
/**
* Open the underlying file using the {@code WorkbookFactory}. We keep track of the used {@code InputStream} so that
* it can be closed cleanly on the end of reading the file. This to be able to release the resources used by
* Apache POI.
*
* @param inputStream the {@code InputStream} pointing to the Excel file.
* @throws Exception is thrown for any errors.
*/
@Override
protected void openExcelFile(final InputStream inputStream) throws Exception {
this.workbook = WorkbookFactory.create(inputStream);
this.workbook.setMissingCellPolicy(Row.MissingCellPolicy.CREATE_NULL_AS_BLANK);
}
}