一个excel(20M)就能干趴你的poi，你信吗？

　　自从上一篇:一个普通类就能干趴你的springboot，你信吗？后，很巧的是这次又发现一个问题，所以有了这篇文章，还是想沿用上篇的”流水帐“的方式查找问题和解决问题。这篇文章主要是因为使用POI导入一个20M的excel发生了OOM(OutOfMemoryError)异常。说来也惭愧，工作了这么多年，还真没导入过这种大小的文件，并且还发生了内存溢出。如果你百度下基本上清一色的告诉你:POI导入excel文件有两种方式，第一种是用户模式，这种模式用起来很简单直观，可以类比xml的dom方式解析(这里只针对excel2007，因为2003本身就有最大条数限制并且目前基本用的很少，这里直接忽略)，第二种是event模式，这种通常是网上说的解决POI导入大excel的”万金油“方法,可以类比为xml的sax解析方式。呵呵，我这篇文章首先就是要干趴这种方法（JVM使用-Xms512m -Xmx512m）。不信你随便写一个导入接口，导入如下20M大小的execl看看:链接: https://pan.baidu.com/s/1DUrS8ctLPp7Z6imOc1aIUQ 提取码: hd79 。

　　首先，既然要导入大点的excel2007，那么我们应该稍微了解一下这种文件如何存储数据，我们百度上可以发现，2007其实就是一个压缩包，可以直接修改后缀成zip然后解压打开文件看看，如下

 

　　上图可以看到最大的两个文件就两个:sharedStrings.xml和sheet1.xml。其中sheet2.xml这个可以不关注，直接从excel删掉都没事，这里没删除主要是没多大关系，这个excel文件也是测试直接提供给我的。由于sheet2比较小，与这个文章说到的内存溢出并无关系，请不要胡思乱想，被分散了注意。

　　直接用大文本编辑工具打开上图两个大文件，可以发现sharedString.xml里内容其实就是excel中每个单元格里的字符串内容（数字类型除外）,sheet.xml就是每个sheet里的结构xml，了解到这里基本上就了解了本文章说到问题的基本知识，然后下面进入正题。

　　先使用百度中查到的提供的event方式导入excel,代码如下:

package com.example.utils;

import org.apache.poi.openxml4j.opc.OPCPackage;
import org.apache.poi.xssf.eventusermodel.ReadOnlySharedStringsTable;
import org.apache.poi.xssf.eventusermodel.XSSFReader;
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.File;
import java.io.InputStream;
import java.sql.SQLException;
import java.util.ArrayList;
import java.util.Iterator;
import java.util.List;

/**
 * 百度上直接copy过来的
 * XSSF and SAX (Event API)
 */
public abstract class BigDataParseExcelUtil extends DefaultHandler {
    private ReadOnlySharedStringsTable sst;
    private String lastContents;
    private boolean nextIsString;
    private int sheetIndex = -1;
    private List rowlist = new ArrayList();
    private int curRow = 0; // 当前行
    private int curCol = 0; // 当前列索引
    private int preCol = 0; // 上一列列索引
    private int titleRow = 0; // 标题行，一般情况下为0
    private int rowsize = 0; // 列数
    private List excelList = new ArrayList();  //excel全部转换为list

    // excel记录行操作方法，以sheet索引，行索引和行元素列表为参数，对sheet的一行元素进行操作，元素为String类型

    public abstract void optRows(int sheetIndex, int curRow,
                                 List rowlist, List excelList) throws SQLException, Exception;

    // 只遍历一个sheet，其中sheetId为要遍历的sheet索引，从1开始，1-3

    /**
     * @param filename
     * @param sheetId  sheetId为要遍历的sheet索引，从1开始，1-3
     * @throws Exception
     */
    public void processOneSheet(String filename, int sheetId) throws Exception {
        OPCPackage pkg = OPCPackage.open(filename);
        XSSFReader r = new XSSFReader(pkg);
        ReadOnlySharedStringsTable strings = new ReadOnlySharedStringsTable(pkg);
        XMLReader parser = fetchSheetParser(strings);
        // rId2 found by processing the Workbook
        // 根据 rId# 或 rSheet# 查找sheet
        InputStream sheet2 = r.getSheet("rId" + sheetId);
        sheetIndex++;
        InputSource sheetSource = new InputSource(sheet2);
        parser.parse(sheetSource);
        sheet2.close();
    }

    @Override
    public void characters(char[] ch, int start, int length)
        throws SAXException {
        // 得到单元格内容的值
        lastContents += new String(ch, start, length);
    }

    public void process(InputStream inputStream) throws Exception {
        OPCPackage pkg = OPCPackage.open(inputStream);
        XSSFReader r = new XSSFReader(pkg);
        ReadOnlySharedStringsTable strings = new ReadOnlySharedStringsTable(pkg);
        XMLReader parser = fetchSheetParser(strings);
        Iterator sheets = r.getSheetsData();
        while (sheets.hasNext()) {
            curRow = 0;
            sheetIndex++;
            InputStream sheet = sheets.next();
            InputSource sheetSource = new InputSource(sheet);
            parser.parse(sheetSource);
            sheet.close();
        }
    }

    /**
     * 遍历 excel 文件
     */
    public void process(File file) throws Exception {
        OPCPackage pkg = OPCPackage.open(file);
        XSSFReader r = new XSSFReader(pkg);
        ReadOnlySharedStringsTable strings = new ReadOnlySharedStringsTable(pkg);
        XMLReader parser = fetchSheetParser(strings);
        Iterator sheets = r.getSheetsData();
        while (sheets.hasNext()) {
            curRow = 0;
            sheetIndex++;
            InputStream sheet = sheets.next();
            InputSource sheetSource = new InputSource(sheet);
            parser.parse(sheetSource);
            sheet.close();
        }
    }

    public XMLReader fetchSheetParser(ReadOnlySharedStringsTable sst)
        throws SAXException {
        XMLReader parser = XMLReaderFactory.createXMLReader();
        // .createXMLReader("org.apache.xerces.parsers.SAXParser");
        this.sst = sst;
        parser.setContentHandler(this);
        return parser;
    }

    @Override
    public void startElement(String uri, String localName, String name,
                             Attributes attributes) throws SAXException {
        // c => 单元格
        if (name.equals("c")) {
            // 如果下一个元素是 SST 的索引，则将nextIsString标记为true
            String cellType = attributes.getValue("t");
            String rowStr = attributes.getValue("r");
            curCol = this.getRowIndex(rowStr);
            if (cellType != null && cellType.equals("s")) {
                nextIsString = true;
            } else {
                nextIsString = false;
            }
        }
        // 置空
        lastContents = "";
    }

    @Override
    public void endElement(String uri, String localName, String name)
        throws SAXException {
        // 根据SST的索引值的到单元格的真正要存储的字符串
        // 这时characters()方法可能会被调用多次
        if (nextIsString) {
            try {
                int idx = Integer.parseInt(lastContents);
                lastContents = new XSSFRichTextString(sst.getEntryAt(idx))
                    .toString();
            } catch (Exception e) {
            }
        }
        // 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
        // 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
        if (name.equals("v")) {
            String value = lastContents.trim();
            value = value.equals("") ? " " : value;
            int cols = curCol - preCol;
            if (cols > 1) {
                for (int i = 0; i < cols - 1; i++) {
                    rowlist.add(preCol, "");
                }
            }
            preCol = curCol;
            rowlist.add(curCol - 1, value);
        } else {
            // 如果标签名称为 row ，这说明已到行尾，调用 optRows() 方法
            if (name.equals("row")) {
                int tmpCols = rowlist.size();
                if (curRow > this.titleRow && tmpCols < this.rowsize) {
                    for (int i = 0; i < this.rowsize - tmpCols; i++) {
                        rowlist.add(rowlist.size(), "");
                    }
                }
                try {
                    optRows(sheetIndex, curRow, rowlist, excelList);
                } catch (SQLException e) {
                    e.printStackTrace();
                } catch (Exception e) {
                    // TODO Auto-generated catch block
                    e.printStackTrace();
                }
                if (curRow == this.titleRow) {
                    this.rowsize = rowlist.size();
                }
                rowlist.clear();
                curRow++;
                curCol = 0;
                preCol = 0;
            }
        }
    }

    // 得到列索引，每一列c元素的r属性构成为字母加数字的形式，字母组合为列索引，数字组合为行索引，
    // 如AB45,表示为第（A-A+1）*26+（B-A+1）*26列，45行
    public int getRowIndex(String rowStr) {
        rowStr = rowStr.replaceAll("[^A-Z]", "");
        byte[] rowAbc = rowStr.getBytes();
        int len = rowAbc.length;
        float num = 0;
        for (int i = 0; i < len; i++) {
            num += (rowAbc[i] - 'A' + 1) * Math.pow(26, len - i - 1);
        }
        return (int) num;
    }


}

package com.example.service;

import com.example.utils.BigDataParseExcelUtil;
import org.springframework.stereotype.Service;

import java.io.InputStream;
import java.sql.SQLException;
import java.util.List;

/**
 * @author: rongdi
 * @date:
 */
@Service
public class ExcelService {

    public void import1(InputStream inputStream) throws Exception {

        BigDataParseExcelUtil xlx = new BigDataParseExcelUtil() {
            @Override
            public void optRows(int sheetIndex, int curRow, List rowlist, List excelList)
                throws SQLException {
                System.out.println(rowlist);
            }
        };
        xlx.process(inputStream);
    }


}

package com.example.controller;

import com.example.service.ExcelService;
import org.springframework.beans.factory.annotation.Autowired;
import org.springframework.stereotype.Controller;
import org.springframework.web.bind.annotation.RequestMapping;
import org.springframework.web.bind.annotation.RequestParam;
import org.springframework.web.bind.annotation.ResponseBody;
import org.springframework.web.multipart.MultipartFile;

/**
 * @author: rongdi
 * @date:
 */
@Controller
public class ExcelController {

    @Autowired
    private ExcelService excelService;

    @RequestMapping("/excel/import1")
    @ResponseBody
    public String import1(@RequestParam("file") MultipartFile multipartFile) throws Exception {
        excelService.import1(multipartFile.getInputStream());
        return "ok";
    }

}

　　使用postman等工具，导入上面说的20M的文件22.xlsx，报错如下:

 　　那我们优化一下不使用inputStream，直接使用一个File传入看看

    public void import2(File file) throws Exception {
        BigDataParseExcelUtil xlx = new BigDataParseExcelUtil() {
            @Override
            public void optRows(int sheetIndex, int curRow, List rowlist, List excelList)
                throws SQLException {
                System.out.println(rowlist);
            }
        };
        xlx.process(file);
    }

    @RequestMapping("/excel/import2")
    @ResponseBody
    public String import2(@RequestParam("file") MultipartFile multipartFile) throws Exception {
        // 延迟解析比率
        ZipSecureFile.setMinInflateRatio(-1.0d);
        File tmp = Files.createTempFile("tmp-", ".xlsx").toFile();
        Files.copy(multipartFile.getInputStream(), Paths.get(tmp.getPath()), StandardCopyOption.REPLACE_EXISTING);
        excelService.import2(tmp);
        return "ok";
    }

　　使用postman上传文件运行效果如下：

 

 　　这时候就发现很尴尬了，难道是POI自己代码里就有bug，我们可以使用断点调试确认一下这个strings里到底是不是全部放了sharedStrings.xml里的内容。

 　　由上证明就是这个strings装多了导致内存溢出了，从这里可以看出网上说使用event事件解析excel的方案基本都是行不通的，哎，我也不懂为啥百度上都是这种答案，难道他们压根都没遇到过大数据导入吗?当然也有可能我冤枉了他们，因为sharedStrings.xml中存放的是每个单元格的字符串内容，这个存放是排重过的，如果虽然excel里单元格很多，但是大多都是整型或者大多都是重复的，那确实可以跳过这一步一路之后会一路畅通了，因为毕竟sax解析xml确实可以节省很多内存。

　　从上分析可以看到POI就两种方式导入：一种是用户方式写代码简单，基本按顺序数格子就好，但是类比dom方式解析xml，很耗内存。第二种事件方式，类比sax方式解析xml确实很省内存，但是POI提供的类里把解析出的大量字符串放入了集合中，还是会导致内存溢出。那么我们怎么解决这个问题，这里很常规的想法是到底这个strings是用来干啥的，怎么用的，如果可以保持strings相同逻辑功能的前提下，修改了ReadOnlySharedStringsTable这个类的逻辑，就可以解决这里的内存溢出了。那么我们可以直接搜索ReadOnlySharedStringsTable类里所有用到strings的方法上打上断点，特别是从strings里取值的方法上，然后调大jvm内存避免内存溢出的情况下断点调试如下

 　　我们是不是可以直接往strings里添加字符串和获取字符串的方法那里替换掉，不要使用strings这个集合存储所有字符串。但是既然excel里设计成使用一个sharedStrings.xml存放公共的字符串，而不是像csv格式那样，每次读一行取一行数据就好了。那么这个sharedStrings.xml中的数据总要解析出来，总要有个地方存储里面的数据，不然怎么结合sheet.xml的格式获取到每一行的数据呢？所以这里就很尴尬了，不能每次解析sharedStrings.xml时不保存每次需要获取strings的时候，再去解析一下这个xm吧,如果从本文章的xml上来看，要重复解析25W次，效率极其低。现在问题可以简化成我们需要把sharedStrings.xml解析出的所有字符串放在一个地方，还能方便解析，由于怕内存溢出，肯定不能放在内存中了。那么这里就有一些选择，比如解析出的字符串按加入strings集合的顺序放入数据库，文件，外部存储或者缓存(限制内存大小，多余写入文件)存储中。然后使用的时候按照索引位置idx去一一取出。本文章先使用临时文件来放这些数据，因为不想搞那么复杂，导入任务不管再多复杂的系统中，最终执行的都会是一个单节点，在单节点中先使用本机资源这种就近资源是最方便的。如下直接先复制源码，然后修改上述说的两个地方。

package com.example.utils;

import org.apache.poi.ooxml.util.SAXHelper;
import org.apache.poi.openxml4j.opc.OPCPackage;
import org.apache.poi.openxml4j.opc.PackagePart;
import org.apache.poi.ss.usermodel.RichTextString;
import org.apache.poi.util.Removal;
import org.apache.poi.xssf.model.SharedStrings;
import org.apache.poi.xssf.usermodel.XSSFRelation;
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 javax.xml.parsers.ParserConfigurationException;
import java.io.File;
import java.io.FileNotFoundException;
import java.io.FileOutputStream;
import java.io.FileReader;
import java.io.IOException;
import java.io.InputStream;
import java.io.LineNumberReader;
import java.io.PushbackInputStream;
import java.nio.file.Files;
import java.util.ArrayList;
import java.util.LinkedHashMap;
import java.util.List;
import java.util.Map;

import static org.apache.poi.xssf.usermodel.XSSFRelation.NS_SPREADSHEETML;

public class ReadOnlySharedStringsTable extends DefaultHandler implements SharedStrings {

    protected final boolean includePhoneticRuns;

    /**
     * An integer representing the total count of strings in the workbook. This count does not
     * include any numbers, it counts only the total of text strings in the workbook.
     */
    protected int count;

    /**
     * An integer representing the total count of unique strings in the Shared String Table.
     * A string is unique even if it is a copy of another string, but has different formatting applied
     * at the character level.
     */
    protected int uniqueCount;

    /**
     * The shared strings table.
     */
    private List strings;

    private File tmp = null;

    FileOutputStream fos = null;

    private int counts;

    private Map map = new LinkedHashMap();

    public ReadOnlySharedStringsTable(OPCPackage pkg)
            throws IOException, SAXException {
        this(pkg, true);
    }

    public ReadOnlySharedStringsTable(OPCPackage pkg, boolean includePhoneticRuns)
            throws IOException, SAXException {
        this.includePhoneticRuns = includePhoneticRuns;
        ArrayList parts =
                pkg.getPartsByContentType(XSSFRelation.SHARED_STRINGS.getContentType());

        // Some workbooks have no shared strings table.
        if (parts.size() > 0) {
            PackagePart sstPart = parts.get(0);
            readFrom(sstPart.getInputStream());
        }
    }

    /**
     * Like POIXMLDocumentPart constructor
     *
     * Calls {@link #ReadOnlySharedStringsTable(PackagePart, boolean)}, with a
     * value of true to include phonetic runs.
     *
     * @since POI 3.14-Beta1
     */
    public ReadOnlySharedStringsTable(PackagePart part) throws IOException, SAXException {
        this(part, true);
    }

    /**
     * Like POIXMLDocumentPart constructor
     *
     * @since POI 3.14-Beta3
     */
    public ReadOnlySharedStringsTable(PackagePart part, boolean includePhoneticRuns)
        throws IOException, SAXException {
        this.includePhoneticRuns = includePhoneticRuns;
        readFrom(part.getInputStream());
    }
    
    /**
     * Read this shared strings table from an XML file.
     *
     * @param is The input stream containing the XML document.
     * @throws IOException if an error occurs while reading.
     * @throws SAXException if parsing the XML data fails.
     */
    public void readFrom(InputStream is) throws IOException, SAXException {
        // test if the file is empty, otherwise parse it
        PushbackInputStream pis = new PushbackInputStream(is, 1);
        int emptyTest = pis.read();
        if (emptyTest > -1) {
            pis.unread(emptyTest);
            InputSource sheetSource = new InputSource(pis);
            try {
                XMLReader sheetParser = SAXHelper.newXMLReader();
                sheetParser.setContentHandler(this);
                sheetParser.parse(sheetSource);
            } catch(ParserConfigurationException e) {
                throw new RuntimeException("SAX parser appears to be broken - " + e.getMessage());
            }
        }
    }

    /**
     * Return an integer representing the total count of strings in the workbook. This count does not
     * include any numbers, it counts only the total of text strings in the workbook.
     *
     * @return the total count of strings in the workbook
     */
    @Override
    public int getCount() {
        return this.count;
    }

    /**
     * Returns an integer representing the total count of unique strings in the Shared String Table.
     * A string is unique even if it is a copy of another string, but has different formatting applied
     * at the character level.
     *
     * @return the total count of unique strings in the workbook
     */
    @Override
    public int getUniqueCount() {
        return this.uniqueCount;
    }

    /**
     * Return the string at a given index.
     * Formatting is ignored.
     *
     * @param idx index of item to return.
     * @return the item at the specified position in this Shared String table.
     * @deprecated use getItemAt instead
     */
    @Removal(version = "4.2")
    @Deprecated
    public String getEntryAt(int idx) {
        /**
         * 这里就是修改部分了，直接从按行存储的临时文件读取需要的字符串
         */
        String value = map.get(idx + 1);
        if(value == null) {

            return readString(idx,1000,this.uniqueCount);
        } else {
            return value;
        }

    }

    /**
     * 从指定位置读取size个字符串，这里是使用局部性原理，每次读取size个字符串，
     * 以免每次需要读取文件，性能极低
     * @return
     */
    private String readString(int idx,int size,int numbers) {
        map.clear();
        int currNumber = idx + 1;
        if (currNumber < 0 || currNumber > numbers) {
            return null;
        }
        try {
            FileReader in = new FileReader(tmp);
            LineNumberReader reader = new LineNumberReader(in);
            try {
                String line = "";
                for(int i = 1;i <= numbers;i ++) {
                    line = reader.readLine();
                    if(i >= currNumber && i < currNumber + size) {
                        map.put(i, line);
                    }
                }
            } finally {
                reader.close();
                in.close();
            }
        } catch (Exception e) {
            System.out.println(e.getMessage());
        }
        return map.get(idx + 1);
    }


    /**
     * Returns all the strings.
     * Formatting is ignored.
     *
     * @return a list with all the strings
     * @deprecated use getItemAt instead
     */
    @Removal(version = "4.2")
    @Deprecated
    public List getItems() {
        return strings;
    }

    @Override
    public RichTextString getItemAt(int idx) {
        return new XSSFRichTextString(getEntryAt(idx));
    }

    //// ContentHandler methods ////

    private StringBuilder characters;
    private boolean tIsOpen;
    private boolean inRPh;

    @Override
    public void startElement(String uri, String localName, String name,
                             Attributes attributes) throws SAXException {
        if (uri != null && ! uri.equals(NS_SPREADSHEETML)) {
            return;
        }

        if ("sst".equals(localName)) {
            String count = attributes.getValue("count");
            if(count != null) this.count = Integer.parseInt(count);
            String uniqueCount = attributes.getValue("uniqueCount");
            if(uniqueCount != null) this.uniqueCount = Integer.parseInt(uniqueCount);
            try {
                tmp = Files.createTempFile("tmp-", ".xlsx").toFile();
            } catch (IOException e) {
                e.printStackTrace();
            }
            //    this.strings = new ArrayList<>(this.uniqueCount);
            characters = new StringBuilder(64);
            try {
                fos = new FileOutputStream(tmp,true);
            } catch (FileNotFoundException e) {
                e.printStackTrace();
            }
        } else if ("si".equals(localName)) {
            characters.setLength(0);
        } else if ("t".equals(localName)) {
            tIsOpen = true;
        } else if ("rPh".equals(localName)) {
            inRPh = true;
            //append space...this assumes that rPh always comes after regular 
            if (includePhoneticRuns && characters.length() > 0) {
                characters.append(" ");
            }
        }
    }

    @Override
    public void endElement(String uri, String localName, String name) throws SAXException {
        if (uri != null && ! uri.equals(NS_SPREADSHEETML)) {
            return;
        }

        if ("si".equals(localName)) {
         //   strings.add(characters.toString().intern());
            try {
                /**
                 * 这里就是修改的一部分，这里直接把字符串按行存入临时文件
                 */
                counts ++;
                fos.write((characters.toString() + "\n").getBytes());
                if(counts == this.uniqueCount) {
                    fos.close();
                }
            } catch (IOException e) {
                e.printStackTrace();
            }
        } else if ("t".equals(localName)) {
            tIsOpen = false;
        } else if ("rPh".equals(localName)) {
            inRPh = false;
        }
    }

    /**
     * Captures characters only if a t(ext) element is open.
     */
    @Override
    public void characters(char[] ch, int start, int length) throws SAXException {
        if (tIsOpen) {
            if (inRPh && includePhoneticRuns) {
                characters.append(ch, start, length);
            } else if (! inRPh){
                characters.append(ch, start, length);
            }
        }
    }

}

　　然后在自己代码里把包换成自己的包，替换POI里该类的包，运行JVM堆情况如下毫无压力

　　自此内存溢出问题大功告成！针对使用POI导入大Excel遇到的问题总结如下：

　　1）网上给出的方案不管是用户模式还是事件模式，往往都不能支持大excel的导入

　　2）excel本质上是一堆excel的压缩包（这里只考虑2007忽略2003）改了个后缀名成xlsx

　　3）使用事件导入时应先将上传文件存入文件，再使用文件OPCPackage.open(file)，如果直接传入输入流，由于里面逻辑会将输入流的所有内容先存入ByteArrayOutputStream 中，这个输出流实际上是一个内存中的字节流，所以也会导致内存溢出。

　　4）用户模式不用考虑，事件模式会先将sharedString.xml这个大xml解析出来放入一个List中，不管通过什么方式都绕不开需要解析这个类，因为每个单元格的字符串都放在这个xml中，而要解析这个xml最常规的方法就是保存在内存使用list和map之内的容器存放我相信不会有人会想刚解析出一个xml还要存回文件中把，这里基本就绕不开ReadOnlySharedStringsTable或者SharedStringsTable,就算你侥幸绕开了，想自己解析，或许还是重复这两个类的悲剧，这就是另外一种内存溢出的根源。

　　回顾一下上述实现直接把sharedStrings.xml中的内容粗略的保存到文件中，然后再从文件中获取是属于很低劣的实现，只能说能满足不内存溢出，性能方面堪忧！下面直接借鉴easyexcel源码中用到的ReadCache来实现保存sharedStrings.xml中的内容

package com.example.advanceevent;

import com.example.utils.FileUtils;
import org.ehcache.Cache;
import org.ehcache.CacheManager;
import org.ehcache.config.CacheConfiguration;
import org.ehcache.config.builders.CacheConfigurationBuilder;
import org.ehcache.config.builders.CacheManagerBuilder;
import org.ehcache.config.builders.ResourcePoolsBuilder;
import org.ehcache.config.units.MemoryUnit;
import org.ehcache.core.Ehcache;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;

import java.io.File;
import java.util.HashMap;
import java.util.UUID;

/**
 * @author: rongdi
 * @date:
 */
public class ReadCache {

    private static final Logger LOGGER = LoggerFactory.getLogger(Ehcache.class);
    private int index = 0;
    private HashMap dataMap = new HashMap(1334);
    private static CacheManager fileCacheManager;
    private static CacheConfiguration fileCacheConfiguration;
    private static CacheManager activeCacheManager;
    private CacheConfiguration activeCacheConfiguration;
    private Cache fileCache;
    private Cache activeCache;
    private String cacheAlias;
    private int cacheMiss = 0;

    public ReadCache(int maxCacheActivateSize) {
        this.activeCacheConfiguration = CacheConfigurationBuilder.newCacheConfigurationBuilder(Integer.class, HashMap.class, ResourcePoolsBuilder.newResourcePoolsBuilder().heap((long)maxCacheActivateSize, MemoryUnit.MB)).withSizeOfMaxObjectGraph(1000000L).withSizeOfMaxObjectSize((long)maxCacheActivateSize, MemoryUnit.MB).build();
        init();
    }

    private void init() {
        this.cacheAlias = UUID.randomUUID().toString();
        this.fileCache = fileCacheManager.createCache(this.cacheAlias, fileCacheConfiguration);
        this.activeCache = activeCacheManager.createCache(this.cacheAlias, this.activeCacheConfiguration);
    }

    public void put(String value) {
        this.dataMap.put(this.index, value);
        if ((this.index + 1) % 1000 == 0) {
            this.fileCache.put(this.index / 1000, this.dataMap);
            this.dataMap = new HashMap(1334);
        }

        ++this.index;
        if (LOGGER.isDebugEnabled() && this.index % 1000000 == 0) {
            LOGGER.debug("Already put :{}", this.index);
        }

    }

    public String get(Integer key) {
        if (key != null && key >= 0) {
            int route = key / 1000;
            HashMap dataMap = (HashMap)this.activeCache.get(route);
            if (dataMap == null) {
                dataMap = (HashMap)this.fileCache.get(route);
                this.activeCache.put(route, dataMap);
                if (LOGGER.isDebugEnabled() && this.cacheMiss++ % 1000 == 0) {
                    LOGGER.debug("Cache misses count:{}", this.cacheMiss);
                }
            }

            return (String)dataMap.get(key);
        } else {
            return null;
        }
    }

    public void putFinished() {
        if (this.dataMap != null) {
            this.fileCache.put(this.index / 1000, this.dataMap);
        }
    }

    public void destroy() {
        fileCacheManager.removeCache(this.cacheAlias);
        activeCacheManager.removeCache(this.cacheAlias);
    }

    static {
        File cacheFile = FileUtils.createCacheTmpFile();
        fileCacheManager = CacheManagerBuilder.newCacheManagerBuilder().with(CacheManagerBuilder.persistence(cacheFile)).build(true);
        activeCacheManager = CacheManagerBuilder.newCacheManagerBuilder().build(true);
        fileCacheConfiguration = CacheConfigurationBuilder.newCacheConfigurationBuilder(Integer.class, HashMap.class, ResourcePoolsBuilder.newResourcePoolsBuilder().disk(10L, MemoryUnit.GB)).withSizeOfMaxObjectGraph(1000000L).withSizeOfMaxObjectSize(10L, MemoryUnit.GB).build();
    }

}

package com.example.advanceevent;

import org.apache.poi.ooxml.util.SAXHelper;
import org.apache.poi.openxml4j.opc.OPCPackage;
import org.apache.poi.openxml4j.opc.PackagePart;
import org.apache.poi.ss.usermodel.RichTextString;
import org.apache.poi.util.Removal;
import org.apache.poi.xssf.model.SharedStrings;
import org.apache.poi.xssf.usermodel.XSSFRelation;
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 javax.xml.parsers.ParserConfigurationException;
import java.io.IOException;
import java.io.InputStream;
import java.io.PushbackInputStream;
import java.util.ArrayList;
import java.util.List;

import static org.apache.poi.xssf.usermodel.XSSFRelation.NS_SPREADSHEETML;

public class ReadOnlySharedStringsTable extends DefaultHandler implements SharedStrings {

    protected final boolean includePhoneticRuns;

    /**
     * An integer representing the total count of strings in the workbook. This count does not
     * include any numbers, it counts only the total of text strings in the workbook.
     */
    protected int count;

    /**
     * An integer representing the total count of unique strings in the Shared String Table.
     * A string is unique even if it is a copy of another string, but has different formatting applied
     * at the character level.
     */
    protected int uniqueCount;

    /**
     * 缓存
     */
    ReadCache readCache = new ReadCache(100);

    private int counts;


    public ReadOnlySharedStringsTable(OPCPackage pkg)
            throws IOException, SAXException {
        this(pkg, true);
    }

    public ReadOnlySharedStringsTable(OPCPackage pkg, boolean includePhoneticRuns)
            throws IOException, SAXException {
        this.includePhoneticRuns = includePhoneticRuns;
        ArrayList parts =
                pkg.getPartsByContentType(XSSFRelation.SHARED_STRINGS.getContentType());

        // Some workbooks have no shared strings table.
        if (parts.size() > 0) {
            PackagePart sstPart = parts.get(0);
            readFrom(sstPart.getInputStream());
        }
    }

    /**
     * Like POIXMLDocumentPart constructor
     *
     * Calls {@link #ReadOnlySharedStringsTable(PackagePart, boolean)}, with a
     * value of true to include phonetic runs.
     *
     * @since POI 3.14-Beta1
     */
    public ReadOnlySharedStringsTable(PackagePart part) throws IOException, SAXException {
        this(part, true);
    }

    /**
     * Like POIXMLDocumentPart constructor
     *
     * @since POI 3.14-Beta3
     */
    public ReadOnlySharedStringsTable(PackagePart part, boolean includePhoneticRuns)
        throws IOException, SAXException {
        this.includePhoneticRuns = includePhoneticRuns;
        readFrom(part.getInputStream());
    }
    
    /**
     * Read this shared strings table from an XML file.
     *
     * @param is The input stream containing the XML document.
     * @throws IOException if an error occurs while reading.
     * @throws SAXException if parsing the XML data fails.
     */
    public void readFrom(InputStream is) throws IOException, SAXException {
        // test if the file is empty, otherwise parse it
        PushbackInputStream pis = new PushbackInputStream(is, 1);
        int emptyTest = pis.read();
        if (emptyTest > -1) {
            pis.unread(emptyTest);
            InputSource sheetSource = new InputSource(pis);
            try {
                XMLReader sheetParser = SAXHelper.newXMLReader();
                sheetParser.setContentHandler(this);
                sheetParser.parse(sheetSource);
            } catch(ParserConfigurationException e) {
                throw new RuntimeException("SAX parser appears to be broken - " + e.getMessage());
            }
        }
    }

    /**
     * Return an integer representing the total count of strings in the workbook. This count does not
     * include any numbers, it counts only the total of text strings in the workbook.
     *
     * @return the total count of strings in the workbook
     */
    @Override
    public int getCount() {
        return this.count;
    }

    /**
     * Returns an integer representing the total count of unique strings in the Shared String Table.
     * A string is unique even if it is a copy of another string, but has different formatting applied
     * at the character level.
     *
     * @return the total count of unique strings in the workbook
     */
    @Override
    public int getUniqueCount() {
        return this.uniqueCount;
    }

    /**
     * Return the string at a given index.
     * Formatting is ignored.
     *
     * @param idx index of item to return.
     * @return the item at the specified position in this Shared String table.
     * @deprecated use getItemAt instead
     */
    @Removal(version = "4.2")
    @Deprecated
    public String getEntryAt(int idx) {
        /**
         * 这里就是修改部分了，直接从按行存储的临时文件读取需要的字符串
         */
        return readCache.get(idx);

    }

    /**
     * Returns all the strings.
     * Formatting is ignored.
     *
     * @return a list with all the strings
     * @deprecated use getItemAt instead
     */
    @Removal(version = "4.2")
    @Deprecated
    public List getItems() {
        return null;
    }

    @Override
    public RichTextString getItemAt(int idx) {
        return new XSSFRichTextString(getEntryAt(idx));
    }

    //// ContentHandler methods ////

    private StringBuilder characters;
    private boolean tIsOpen;
    private boolean inRPh;

    @Override
    public void startElement(String uri, String localName, String name,
                             Attributes attributes) throws SAXException {
        if (uri != null && ! uri.equals(NS_SPREADSHEETML)) {
            return;
        }

        if ("sst".equals(localName)) {
            String count = attributes.getValue("count");
            if(count != null) this.count = Integer.parseInt(count);
            String uniqueCount = attributes.getValue("uniqueCount");
            if(uniqueCount != null) this.uniqueCount = Integer.parseInt(uniqueCount);
            //    this.strings = new ArrayList<>(this.uniqueCount);
            characters = new StringBuilder(64);
        } else if ("si".equals(localName)) {
            characters.setLength(0);
        } else if ("t".equals(localName)) {
            tIsOpen = true;
        } else if ("rPh".equals(localName)) {
            inRPh = true;
            //append space...this assumes that rPh always comes after regular 
            if (includePhoneticRuns && characters.length() > 0) {
                characters.append(" ");
            }
        }
    }

    @Override
    public void endElement(String uri, String localName, String name) throws SAXException {
        if (uri != null && ! uri.equals(NS_SPREADSHEETML)) {
            return;
        }

        if ("si".equals(localName)) {
         //   strings.add(characters.toString().intern());
            readCache.put(characters.toString());
            /**
             * 这里就是修改的一部分，这里直接把字符串按行存入临时文件
             */
            counts ++;
            if(counts == this.uniqueCount) {
                readCache.putFinished();
            }
        } else if ("t".equals(localName)) {
            tIsOpen = false;
        } else if ("rPh".equals(localName)) {
            inRPh = false;
        }
    }

    /**
     * Captures characters only if a t(ext) element is open.
     */
    @Override
    public void characters(char[] ch, int start, int length) throws SAXException {
        if (tIsOpen) {
            if (inRPh && includePhoneticRuns) {
                characters.append(ch, start, length);
            } else if (! inRPh){
                characters.append(ch, start, length);
            }
        }
    }

}

　　至此代码效率有了相当大的提高，而且内存溢出问题也得到解决。详细测试代码：https://github.com/rongdi/poi-example.git