Work in progress! This is guide will be improved over time. If you believe there is a missing topic, open an issue or help us with a contribution to improve the guide.
This guide describes the updates in Apache PDFBox 3.0 release. Use the information provided to upgrade your PDFBox 2.x applications to PDFBox 3.0. It provides information about the new, deprecated and unsupported features in this release.
PDFBox 3.0 requires at least Java 8. Testing has been done up to Java 19.
Apache Xmpbox no longer depends on javax.xml.bind.jaxb-api. All test classes were updated to use JUnit 5.
All libraries on which PDFBox depends are updated to their latest stable versions:
For test support the libraries are updated to
This section explains the fundamental differences between PDFBox 3.0 and 2.x releases.
All deprecated APIs and components from PDFBox 2.x have been removed in PDFBox 3.0. Deprecated APIs in
PDPageContentStream have been kept but you are encouraged to replace them with the non deprecated calls
as they are treated to be of internal use only.
All basic classes used for io-operations were moved to a separate module for a shared usage.
<dependency>
<groupId>org.apache.pdfbox</groupId>
<artifactId>pdfbox-io</artifactId>
</dependency>
The whole code was overhauled including the following changes:
PDFBox offers the following implementations of the interface org.apache.pdfbox.io.RandomAccessRead to be used as source to read a pdf:
RandomAccessReadBuffer stores all the data in memory. It is backed by the given byte array or ByteBuffer. Using the constructor with an InputStream copies the data to the buffer. Internally the data is stored in a chunk of ByteBuffers with a default chunk size of 4KB.
RandomAccessReadBufferedFile is backed by the given file. It has an in-memory cache using pages with a size of 4KB. The cache follows the FIFO principle. If the the maximum of 1000 pages is reached the first added page is replaced with new data.
RandomAccessReadMemoryMappedFile uses the memory mapping feature of java. The whole file is mapped to memory and the maximum allowed file size is Integer.MAX_VALUE.
There is a known issue with locked files after closing the memory mapped file on windows. PDFBox implements its own unmapper as a workaround.
Implementing your own reader
If there is any need to implement a different reader one has to implement the interface org.apache.pdfbox.io.RandomAccessRead. It shall be done thread safe to avoid issues in multithreaded environments.
PDFBox offers the following implementation of the interface org.apache.pdfbox.io.RandomAccess to be used to write and read data.
RandomAccessReadWriteBuffer extends the class RandomAccessReadBuffer and stores the all the data in memory as well. The implementation adds the ability to write data to the buffer which is automatically expanded by a new chunk.
PDFBox 3.0.x no longer uses a separate cache when reading a pdf, but still does for write operations. It introduces the interface org.apache.pdfbox.io.RandomAccessStreamCache to define a cache factory in a more flexible way.
Provided implementations
RandomAccessStreamCacheImpl is a simple default implementation using RandomAccessReadWriteBuffer as buffer.
The well known class ScratchFile is another implementation for a cache factory. It can be configured to use memory only, temp file only or a fix of both.
org.apache.pdfbox.io.MemoryUsageSetting
The MemoryUsageSetting parameter within the loadPDF methods was replaced by a parameter using the new functional interface StreamCacheCreateFunction to encapsulate the caching details within the IO package. IOUtils provides two variants of a possible cache for convenience. The memory only one uses RandomAccessStreamCache and the temporary file only uses ScratchFile as cache buffer factory. The newly introduced loader uses a memory only cache as default if the caller doesn't provide any cache.
Implementing your own stream cache
If there is any need to implement a different cache one has to implement the interface org.apache.pdfbox.io.RandomAccessStreamCache. It shall be done thread safe to avoid issues in multithreaded environments.
The new class org.apache.pdfbox.Loader is used for loading a PDF. It offers several methods to load a pdf using different kind of sources. All load methods have been removed from org.apache.pdfbox.pdmodel.PDDocument. The same is true for loading a FDF document.
The most flexible way is to use an instance of RandomAccessRead such as the following sample:
try (PDDocument document = Loader.loadPDF(new RandomAccessReadBufferedFile("yourfile.pdf")))
{
for (PDPage page : document.getPages())
{
....
}
}
org.apache.pdfbox.Loader provides two other kind of load methods for your convenience.
If a byte array is provided as source PDFBox uses org.apache.pdfbox.io.RandomAccessReadBuffer to hold the data. The byte buffer is backed by the given byte array.
If a file is provided as source PDFBox uses org.apache.pdfbox.io.RandomAccessReadBufferedFile to wrap the source data using the in-memory cache as described above.
When saving a PDF this will now be done in compressed mode by default. To override that (e.g. if you want to create a PDF/A-1b document) use PDDocument.save with CompressParameters.NO_COMPRESSION.
The input file must not be used as output for saving operations. It will corrupt the file and throw an exception as parts of the file are read the first time when saving it.
PDFBox now loads a PDF Document incrementally reducing the initial memory footprint (on demand parsing). This will also reduce the memory needed to consume a PDF if only certain parts of the PDF are accessed. Note that, due to the nature of PDF, uses such as iterating over all pages, accessing annotations, signing a PDF etc. might still load all parts of the PDF overtime which might consume a significant amount of memory.
Do not try to access parts of the document after the PDDocument object has been closed, because this may lead to incorrect results, as shown in PDFBOX-5720.
The introduction of the new io classes has a positive impact on the memory usage. Especially the re-usage of the source for reading parts of it instead of using intermediate streams reduces the memory footprint significantly.
There were a lot of changes and optimizations which have a more or less huge impact on the memory consumption.
The static instances of PDType1Font for the standard 14 fonts were removed as the underlying COSDictionary isn't supposed to be immutable which led to several issues.
A new constructor for PDType1Font was introduced to create a standard 14 font. The new Enum Standard14Fonts.FontName is the one and only parameter and defines the
name of the standard 14 font for which the instance of PDType1Font is created for. That instance isn't a singleton anymore and has to be recreated if necessary or cached
by the user if suitable.
The int triple overloads of the setStrokingColor and setNonStrokingColor methods of PDAbstractContentStream, with inputs representing RGB colors defined in the 0-255
range, have been removed. While usages passing in int triples will compile (thanks to implicit casting of the int values to float), an IllegalArgumentException can be
thrown at runtime as the float overloads of these methods accept only values in the range 0-1.
To retain RGB colors defines as 0-255 integer triples, construct a java.awt.Color instance and use the relevant overload. Alternatively, convert values to the 0-1
range and define using float triples instead.
Instead of using the PDAnnotationTextMarkup, PDAnnotationSquareCircle or the PDAnnotationMarkup classes when creating certain annotations, use their subclasses PDAnnotationCaret, PDAnnotationFreeText, PDAnnotationInk, PDAnnotationPolygon, PDAnnotationPolyline, PDAnnotationSound, PDAnnotationCircle, PDAnnotationSquare, PDAnnotationHighlight, PDAnnotationSquiggly, PDAnnotationStrikeout and PDAnnotationUnderline.
The convenience class org.apache.pdfbox.util.Charsets has been removed. Please use the java class java.nio.charset.StandardCharsets instead.
Instead of using the COSDocument.getObjects() to get all objects, use COSDocument.getXrefTable().keySet()
to get all the object keys, and then call COSDocument.getObjectFromPool() with these keys.
When accessing AcroForms using PDDocumentCatalog.getAcroForm() a number of fix ups are applied aligning PDFBox with most of the default behaviour
of Adobe Reader. If you'd like to bypass this use PDDocumentCatalog.getAcroForm(null).
The fix ups include
You can lookup the details in the org.apache.pdfbox.pdmodel.fixup package of the source distribution and also create your own fix up(s).
The command line interface for the PDFBox App has been rewritten. As a result
-h or --help to display usage information-V or --version to display the version informationThe following features were added to the PDFDebugger: