pdfbox PDF解析票据[不同类型]

meta-one / 2024-10-22 / 原文

在做需求中出现完税凭证每个地区都出现不一样的格式这样解析出现了麻烦

话不多说直接上代码(有几种格式还没有解析成功(季度)) 通过解析文本获取不到值

 1     <dependency>
 2       <groupId>org.apache.pdfbox</groupId>
 3       <artifactId>pdfbox</artifactId>
 4       <version>2.0.24</version>
 5     </dependency>
 6    <dependency>
 7       <groupId>net.sf.cssbox</groupId>
 8       <artifactId>pdf2dom</artifactId>
 9       <version>2.0.1</version>
10     </dependency>

View Code

核心逻辑

import com.google.common.collect.Lists;
import org.apache.commons.lang3.StringUtils;
import org.apache.pdfbox.pdmodel.PDDocument;
import org.fit.pdfdom.PDFDomTree;
import org.fit.pdfdom.TextMetrics;
import org.w3c.dom.bootstrap.DOMImplementationRegistry;
import org.w3c.dom.ls.DOMImplementationLS;
import org.w3c.dom.ls.LSOutput;
import org.w3c.dom.ls.LSSerializer;

import java.io.File;
import java.io.FileInputStream;
import java.io.IOException;
import java.math.BigDecimal;
import java.util.ArrayList;
import java.util.List;
import java.util.regex.Matcher;
import java.util.regex.Pattern;

public class MyPdf extends PDFDomTree {

    private static List<String> keywordHeadList = Lists.newArrayList();
    private static List<String> keywordBodyList = Lists.newArrayList();

    private static List<String> keywordAllList = Lists.newArrayList();
    private static List<String> textNotList = Lists.newArrayList();

    private List<String> textAllList;

    public List<String> getTextAllList() {
        return textAllList;
    }

    public void setTextAllList(List<String> textAllList) {
        this.textAllList = textAllList;
    }

    public static List<String> getKeywordHeadList() {
        return keywordHeadList;
    }

    public static void setKeywordHeadList(List<String> keywordHeadList) {
        MyPdf.keywordHeadList = keywordHeadList;
    }

    public static List<String> getKeywordBodyList() {
        return keywordBodyList;
    }

    public static void setKeywordBodyList(List<String> keywordBodyList) {
        MyPdf.keywordBodyList = keywordBodyList;
    }

    static {

        keywordHeadList.add("税务机关");
        keywordHeadList.add("填发日期");
        keywordHeadList.add("纳税人识别号");
        keywordHeadList.add("纳税人名称");
        keywordBodyList.add("原凭证号");
        keywordBodyList.add("税种");
        keywordBodyList.add("品目名称");
        keywordBodyList.add("税款所属时期");
        keywordBodyList.add("税款所属期起");
        keywordBodyList.add("税款所属期止");
        keywordBodyList.add("入(退)库日期");
        keywordBodyList.add("入（退）库日期");
        keywordBodyList.add("实缴(退)金额");
        keywordBodyList.add("实缴(退)税额");
        keywordBodyList.add("实缴（退）税额");
        keywordBodyList.add("收款国库");
        keywordAllList.addAll(keywordHeadList);
        keywordAllList.addAll(keywordBodyList);
        textNotList.add("妥");
        textNotList.add("善");
        textNotList.add("保");
        textNotList.add("管");
        textNotList.add("手");
        textNotList.add("写");
        textNotList.add("无");
        textNotList.add("效");
    }

    public MyPdf() throws IOException {
        super();
    }

    protected void startNewPage() {
        super.startNewPage();
    }


    @Override
    protected void renderText(String data, TextMetrics metrics) {
        data = data.replace("：", "").replace(":", "").replace(",", "").replace("¥", "");
        for (String s : textNotList) {
            data = data.replace(s, "");
        }
        if (StringUtils.isNotBlank(data)) {
            textAllList.add(data);
            curpage.appendChild(createTextElement(data, metrics.getWidth()));
        }
    }

    public List<String> parsePdf(PDDocument doc) throws Exception {
        textAllList = Lists.newArrayList();
        DOMImplementationRegistry registry = DOMImplementationRegistry.newInstance();
        DOMImplementationLS impl = (DOMImplementationLS) registry.getDOMImplementation("LS");
        LSSerializer writer = impl.createLSSerializer();
        LSOutput output = impl.createLSOutput();
        writer.getDomConfig().setParameter("format-pretty-print", true);
        createDOM(doc);
        return textAllList;
    }

    public List<String> parsePdf(File file) throws Exception {
        PDDocument document = PDDocument.load(file);
        List<String> strings = parsePdf(document);
        document.close();
        return strings;

    }

    public List<String> parsePdf(FileInputStream fileInputStream) throws Exception {
        PDDocument document = PDDocument.load(fileInputStream);
        List<String> strings = parsePdf(document);
        document.close();
        return strings;
    }

    public static TaxInvoicePdfDto parsePdfText(List<String> textList) {
        try {
            TaxInvoicePdfDto taxInvoicePdfDto = new TaxInvoicePdfDto();
            List<TaxInvoicePdfDetail> detailList = Lists.newArrayList();
            for (String keyword : keywordHeadList) {
                LoopTextVo loopTextVo = new LoopTextVo(textList, keyword);
                LoopTextVo s = loopTextHeadList(loopTextVo);
                if (keyword.equals("税务机关")) {
                    taxInvoicePdfDto.setNatTaxAuth(s.getValue());
                }
                if (keyword.equals("纳税人名称")) {
                    taxInvoicePdfDto.setTaxpayerName(s.getValue());
                }
                if (keyword.equals("纳税人识别号")) {
                    taxInvoicePdfDto.setTaxpayerNumber(s.getValue());
                }
            }
            List<String> bodyHeaderList = Lists.newArrayList();
            int loopCount = 0;
            for (String keyword : keywordBodyList) {
                LoopTextVo loopTextVo = new LoopTextVo(textList, keyword);
                loopTextVo = loopTextBodyList(loopTextVo);
                bodyHeaderList.addAll(loopTextVo.getBodyHeaderList());
                loopCount = Math.max(loopTextVo.getLoopCount(), loopCount);
            }
            String s = bodyHeaderList.get(bodyHeaderList.size() - 1);
            LoopTextVo loopTextVo = new LoopTextVo(textList, s);
            loopTextVo = loopTextBodyList(loopTextVo);
            List<String> bodyDataList = textList.subList(loopTextVo.getLoopCount() + 1, loopCount);

            String join = StringUtils.join(bodyDataList, " ");
            String datePattern = " 至 ";
            String datePattern1 = "\\d{4}-\\d{2}-\\d{2}至\\d{4}-\\d{2}- \\d{2}";
            Pattern pattern = Pattern.compile(datePattern);
            Pattern pattern1 = Pattern.compile(datePattern1);
            Matcher matcher = pattern.matcher(join);
            Matcher matcher1 = pattern1.matcher(join);
            int index = 0;
            if (matcher.find()) {
                index += 2;
            }
            String[] s2 = join.split(" ");
            ArrayList<String> strings = Lists.newArrayList(s2);
            if (matcher1.find()) {
                join = join.replaceAll("至(\\d{4}-\\d{2}-) (\\d{2})", "至$1$2");
                s2 = join.split(" ");
                strings = Lists.newArrayList(s2);
                strings.remove(s2.length - 1);
            }

            List<List<String>> partition = Lists.partition(strings, bodyHeaderList.size() + index);
            for (List<String> stringList : partition) {
                if(stringList.size() < bodyHeaderList.size()){
                    continue;
                }
                TaxInvoicePdfDetail taxInvoicePdfDetail = new TaxInvoicePdfDetail();
                int index1 = 0;
                boolean flag = false;
                for (int i = 0; i < bodyHeaderList.size(); i++) {
                     if(stringList.size() -2 == bodyHeaderList.size()) {
                        if (bodyHeaderList.get(i).equals("税种")) {
                            taxInvoicePdfDetail.setTaxCategory(stringList.get(i));
                        }
                        if (bodyHeaderList.get(i).equals("品目名称")) {
                            taxInvoicePdfDetail.setTaxItems(stringList.get(i));
                        }
                        if (Lists.newArrayList("税款所属时期").contains(bodyHeaderList.get(i))) {
                            flag = true;
                            String s1 = stringList.get(i) + stringList.get(i + 1) + stringList.get(index1 + 2);
                            taxInvoicePdfDetail.setLegalizeBelongDateFrom(s1.split("至")[0]);
                            taxInvoicePdfDetail.setLegalizeBelongDateTo(s1.split("至")[1]);
                        }
                        if (Lists.newArrayList("税款所属时期", "税款所属期起").contains(bodyHeaderList.get(i))) {
                            taxInvoicePdfDetail.setLegalizeBelongDateFrom(stringList.get(i));
                        }
                        if (Lists.newArrayList("税款所属时期", "税款所属期止").contains(bodyHeaderList.get(i))) {
                            taxInvoicePdfDetail.setLegalizeBelongDateTo(stringList.get(i));
                        }
                        if (Lists.newArrayList("入(退)库日期", "入（退）库日期").contains(bodyHeaderList.get(i))) {
                            index1 = flag ? i + 2 : i;
                            taxInvoicePdfDetail.setPaymentDate(stringList.get(index1));
                        }
                        if (Lists.newArrayList("实缴(退)金额", "实缴(退)税额", "实缴（退）税额").contains(bodyHeaderList.get(i))) {
                            index1 = flag ? i + 2 : i;
                            taxInvoicePdfDetail.setTotalAmount(new BigDecimal(stringList.get(index1)));
                        }
                    }else{
                        if (bodyHeaderList.get(i).equals("税种")) {
                            taxInvoicePdfDetail.setTaxCategory(stringList.get(i));
                        }
                        if (bodyHeaderList.get(i).equals("品目名称")) {
                            taxInvoicePdfDetail.setTaxItems(stringList.get(i));
                        }
                        if (Lists.newArrayList("税款所属时期").contains(bodyHeaderList.get(i))) {
                            taxInvoicePdfDetail.setLegalizeBelongDateFrom(stringList.get(i).split("至")[0]);
                            taxInvoicePdfDetail.setLegalizeBelongDateTo(stringList.get(i).split("至")[1]);
                        }
                        if (Lists.newArrayList("税款所属期起").contains(bodyHeaderList.get(i))) {
                            taxInvoicePdfDetail.setLegalizeBelongDateFrom(stringList.get(i));
                        }
                        if (Lists.newArrayList("税款所属期止").contains(bodyHeaderList.get(i))) {
                            taxInvoicePdfDetail.setLegalizeBelongDateTo(stringList.get(i));
                        }
                        if (Lists.newArrayList("入(退)库日期", "入（退）库日期").contains(bodyHeaderList.get(i))) {
                            taxInvoicePdfDetail.setPaymentDate(stringList.get(i));
                        }
                        if (Lists.newArrayList("实缴(退)金额", "实缴(退)税额", "实缴（退）税额").contains(bodyHeaderList.get(i))) {
                            taxInvoicePdfDetail.setTotalAmount(new BigDecimal(stringList.get(i)));
                        }
                    }
                }
                detailList.add(taxInvoicePdfDetail);
            }
            taxInvoicePdfDto.setDetailList(detailList);
            return taxInvoicePdfDto;
        } catch (Exception e) {
           e.printStackTrace();
            return null;
        }
    }

    public static LoopTextVo loopTextHeadList(LoopTextVo loopTextVo) {
        int loopCount = loopTextVo.getLoopCount();
        List<String> textList = loopTextVo.getTextList();
        String value = loopTextVo.getValue();
        String header = loopTextVo.getHeader();
        String keyword = loopTextVo.getKeyword();
        boolean flag = loopTextVo.isFlag();
        if (loopCount == textList.size()) {
            return loopTextVo;
        }
        String s = textList.get(loopCount);
        if (s.equals("备注")) {
            return loopTextVo;
        }
        if (flag) {
            for (String keywordHead : keywordAllList) {
                if (keywordHead.startsWith(s)) {
                    return loopTextVo;
                }
            }
            loopTextVo.setValue(value + s);
        }
        if (keyword.startsWith(s) || (StringUtils.isNotBlank(header) && keyword.startsWith(header))) {
            header = header + s;
            if (StringUtils.isNotBlank(value)) {
                loopTextVo.setValue(value);
                return loopTextVo;
            }
            boolean equals = keyword.equals(header);
            loopTextVo.setFlag(equals);
            loopTextVo.setLoopCount(loopCount + 1);
            loopTextVo.setHeader(equals ? "" : header);
            return loopTextHeadList(loopTextVo);
        } else if (!flag) {
            loopTextVo.setHeader("");
        }
        if (s.contains(keyword)) {
            loopTextVo.setValue(s.replace(keyword, ""));
            return loopTextVo;
        }
        loopTextVo.setLoopCount(loopCount + 1);
        return loopTextHeadList(loopTextVo);
    }

    public static LoopTextVo loopTextBodyList(LoopTextVo loopTextVo) {
        int loopCount = loopTextVo.getLoopCount();
        List<String> textList = loopTextVo.getTextList();
        String header = loopTextVo.getHeader();
        String keyword = loopTextVo.getKeyword();
        List<String> bodyHeaderList = loopTextVo.getBodyHeaderList();

        if (loopCount == textList.size()) {
            return loopTextVo;
        }
        String s = textList.get(loopCount);
        if (s.startsWith("以上情况") || s.startsWith("金额合计")) {
            if(s.startsWith("金额合计") && textList.size()/loopCount > 2){
                loopTextVo.setLoopCount(loopCount + 1);
                return loopTextBodyList(loopTextVo);
            }
            textList.subList(0, loopCount);
            loopTextVo.setTextList(textList);
            return loopTextVo;
        }
        if (keyword.startsWith(s) || (StringUtils.isNotBlank(header) && keyword.startsWith(header))) {
            header = header + s;
            if (keyword.equals(header)) {
                bodyHeaderList.add(header);
                return loopTextVo;
            }
            loopTextVo.setLoopCount(loopCount + 1);
            loopTextVo.setHeader(header);
            return loopTextBodyList(loopTextVo);
        } else {
            loopTextVo.setHeader("");
        }
        if (s.contains(keyword)) {
            loopTextVo.setValue(s.replace(keyword, ""));
            return loopTextVo;
        }
        loopTextVo.setLoopCount(loopCount + 1);
        return loopTextBodyList(loopTextVo);
    }
}

View Code

VO:

@Data
public class LoopTextVo {

    LoopTextVo(List<String> textList,String keyword){
        this.textList = textList;
        this.keyword = keyword;
    }
    private List<String> textList;

    private List<String> bodyHeaderList = new ArrayList<>();
    private int LoopCount;
    private String keyword = "";
    private String header = "";
    private String value = "";
    private boolean flag = false;
}

View Code

封装对象:

@Data
public class TaxInvoicePdfDto {

    private String natTaxAuth;

    private String taxpayerNumber;

    private String taxpayerName;

    private List<TaxInvoicePdfDetail> detailList;

}

====两个类 分开写 这里就直接扔进去了
@Data
public class TaxInvoicePdfDetail {

    /**
     * 税种
     */
    private String taxCategory;

    /**
     * 品 目 名 称
     */
    private String taxItems;
    /**
     * 税款所属起
     */
    private String legalizeBelongDateFrom;
    /**
     * 税款所属止
     */
    private String legalizeBelongDateTo;

    /**
     * 入(退)库日期
     */
    private String paymentDate;
    /**
     * 实缴(退)金额
     */
    private BigDecimal totalAmount;
}

View Code

测试类:

 public static void main(String[] args) {

        try {
            File[] files = new File("D:\\pdf\\").listFiles(pathname -> {
                if (pathname.isDirectory()) {
                    return false;
                }
                return true;
            });
            for (File file : files) {
                System.out.println(file.getName());
                MyPdf pdfDomTree = new MyPdf();
                List<String> textList = pdfDomTree.parsePdf(file);
                TaxInvoicePdfDto taxInvoicePdfDto = pdfDomTree.parsePdfText(textList);
                System.out.println(taxInvoicePdfDto);
            }
        } catch (Exception e) {
            e.printStackTrace();
        }
    }