Use PDFBox 3 as default text converter; fallback to 1.8 in case of er…

…rors Issue: #4449
portfolio-performance · Jan 6, 2025 · 78a48b0 · 78a48b0
1 parent b898083
commit 78a48b0
Show file tree

Hide file tree

Showing 3 changed files with 45 additions and 33 deletions.
diff --git a/...abuchen.portfolio.ui/src/name/abuchen/portfolio/ui/handlers/CreateTextFromPDFHandler.java b/...abuchen.portfolio.ui/src/name/abuchen/portfolio/ui/handlers/CreateTextFromPDFHandler.java
@@ -49,8 +49,7 @@ public void execute(@Named(IServiceConstants.ACTIVE_PART) MPart part,
 
             StringBuilder textBuilder = new StringBuilder();
             textBuilder.append("```").append("\n");
-            textBuilder.append("PDFBox Version: ")
-                            .append(inputFile.getPDFBoxVersion().toString()) //
+            textBuilder.append("PDFBox Version: ").append(inputFile.getPDFBoxVersion()) //
                             .append("\n");
             textBuilder.append("Portfolio Performance Version: ")
                             .append(PortfolioPlugin.getDefault().getBundle().getVersion().toString()) //

diff --git a/name.abuchen.portfolio/src/name/abuchen/portfolio/datatransfer/pdf/PDFImportAssistant.java b/name.abuchen.portfolio/src/name/abuchen/portfolio/datatransfer/pdf/PDFImportAssistant.java
@@ -12,6 +12,7 @@
 import org.eclipse.core.runtime.IProgressMonitor;
 
 import name.abuchen.portfolio.Messages;
+import name.abuchen.portfolio.PortfolioLog;
 import name.abuchen.portfolio.datatransfer.Extractor;
 import name.abuchen.portfolio.datatransfer.Extractor.Item;
 import name.abuchen.portfolio.datatransfer.SecurityCache;
@@ -164,6 +165,27 @@ public Map<Extractor, List<Item>> run(IProgressMonitor monitor, Map<File, List<E
                     }
                 }
 
+                if (!extracted)
+                {
+                    inputFile.convertLegacyPDFtoText();
+                    for (Extractor extractor : extractors)
+                    {
+                        List<Item> items = extractor.extract(securityCache, inputFile, warnings);
+
+                        if (!items.isEmpty())
+                        {
+                            extracted = true;
+                            itemsByExtractor.computeIfAbsent(extractor, e -> new ArrayList<Item>()).addAll(items);
+                            break;
+                        }
+                    }
+
+                    if (extracted)
+                    {
+                        PortfolioLog.info("PDF successfully imported with PDFBox 1.8.x " + inputFile.getName()); //$NON-NLS-1$
+                    }
+                }
+
                 if (!extracted)
                 {
                     Predicate<? super Exception> isNotUnsupportedOperation = e -> !(e instanceof UnsupportedOperationException);

diff --git a/name.abuchen.portfolio/src/name/abuchen/portfolio/datatransfer/pdf/PDFInputFile.java b/name.abuchen.portfolio/src/name/abuchen/portfolio/datatransfer/pdf/PDFInputFile.java
@@ -7,17 +7,14 @@
 import java.util.List;
 import java.util.Scanner;
 
-import org.apache.pdfbox.exceptions.CryptographyException;
-import org.apache.pdfbox.pdmodel.PDDocument;
-import org.apache.pdfbox.util.PDFTextStripper;
-import org.osgi.framework.FrameworkUtil;
-import org.osgi.framework.Version;
-
 import name.abuchen.portfolio.datatransfer.Extractor;
+import name.abuchen.portfolio.pdfbox1.PDFBox1Adapter;
+import name.abuchen.portfolio.pdfbox3.PDFBox3Adapter;
 
 public class PDFInputFile extends Extractor.InputFile
 {
     private String text;
+    private String version;
 
     public PDFInputFile(File file)
     {
@@ -27,8 +24,7 @@ public PDFInputFile(File file)
     /* protected */ PDFInputFile(File file, String extractedText)
     {
         this(file);
-        this.text = extractedText;
-        this.text = withoutHorizontalWhitespace(extractedText);
+        this.text = sanitize(extractedText);
     }
 
     public static List<Extractor.InputFile> loadTestCase(Class<?> testCase, String... filenames)
@@ -62,37 +58,32 @@ public String getText()
         return text;
     }
 
-    public Version getPDFBoxVersion()
+    public String getPDFBoxVersion()
     {
-        return FrameworkUtil.getBundle(PDDocument.class).getVersion();
+        return version;
     }
 
     public void convertPDFtoText() throws IOException
     {
-        try (PDDocument document = PDDocument.load(getFile()))
-        {
-            boolean isProtected = document.isEncrypted();
-            if (isProtected)
-            {
-                document.decrypt(""); //$NON-NLS-1$
-                document.setAllSecurityToBeRemoved(true);
-            }
-
-            PDFTextStripper textStripper = new PDFTextStripper();
-            textStripper.setSortByPosition(true);
-            text = textStripper.getText(document);
-
-            text = withoutHorizontalWhitespace(text);
-        }
-        catch (CryptographyException e)
-        {
-            throw new IOException(e);
-        }
+        var adapter = new PDFBox3Adapter();
+
+        text = sanitize(adapter.convertToText(getFile()));
+        version = adapter.getPDFBoxVersion();
+    }
+
+    public void convertLegacyPDFtoText() throws IOException
+    {
+        var adapter = new PDFBox1Adapter();
+
+        text = sanitize(adapter.convertToText(getFile()));
+        version = adapter.getPDFBoxVersion();
     }
 
-    private String withoutHorizontalWhitespace(String s)
+    @SuppressWarnings("nls")
+    private String sanitize(String s)
     {
         // replace horizontal whitespace characters by normal whitespace
-        return s.replaceAll("\\h", " "); //$NON-NLS-1$ //$NON-NLS-2$
+        // without carriage returns
+        return s.replaceAll("\\h", " ").replace("\r", "");
     }
 }