PD4ML PDF Optimizer

PD4ML PDF Optimizer rewrites an existing PDF file so that it contains only what the document still uses. A PDF that has been edited by incremental updates keeps objects nothing points to any more, plus one cross-reference section and trailer per update; the same font or image can also end up embedded twice. The optimizer drops the unreferenced objects, merges byte-identical duplicates and writes the result as one fresh file with a single cross-reference section and trailer. You call it from Java (com.pd4ml.pdf.optimizer) or run it from the command line.

What you can do with it

• Drop unreferenced objects: only what is reachable from the trailer's /Root and /Info is copied to the output, with compact object numbering.
• Collapse incremental-update history: the output is a full rewrite with one cross-reference section and trailer, not another incremental update.
• Compact output: from v4.2.0, a PDF 1.5 or later document is written with its dictionaries and arrays packed into compressed object streams and a compressed cross-reference stream, as PD4ML writes its own PDFs; a PDF 1.4 document (PDF/A-1 among them) and an encrypted output use the classic cross-reference table.
• Merge duplicate content: byte-identical objects (for example the same font, image or color space embedded twice) become one shared copy. On by default; page, annotation and form field dictionaries are never merged. Turn it off with deduplicateContent(false) or --no-dedup.
• Open encrypted input: pass the user or owner password. The output is plain unless you request encryption.
• Encrypt the output: AES-256 by default, or RC4-128 / AES-128 with your own permission bits.
• Linearize the output ("Fast Web View"): a viewer that loads the file over a network shows the first page, and then any other page, before the whole file has arrived. Turn it on with linearize(true) or --linearize (from v4.2.0; not together with output encryption).
• See what changed: PdfOptimizeResult reports input and output byte counts, object counts, removed objects and how many of them were merged duplicates.

The optimizer does not recompress streams: stream data is copied as it is.

Quick start

import com.pd4ml.pdf.optimizer.PdfOptimizeException;
import com.pd4ml.pdf.optimizer.PdfOptimizeResult;
import com.pd4ml.pdf.optimizer.PdfOptimizer;

import java.io.File;

public class OptimizeQuickStart {
    public static void main(String[] args) throws PdfOptimizeException {
        PdfOptimizeResult result = PdfOptimizer.optimize(
                new File("bloated.pdf"), new PdfOptimizer.Options(), new File("cleaned.pdf"));
        System.out.println(result.getInputByteCount() + " -> " + result.getOutputByteCount() + " bytes, "
                + result.getRemovedObjectCount() + " object(s) removed, "
                + result.getDuplicateObjectCount() + " of them duplicate content");
    }
}

The same from the command line (--in and --out are required; add --password=<pw> for an encrypted input):

java -cp pd4ml.jar com.pd4ml.pdf.optimizer.cli.Pd4OptimizeCli --in=bloated.pdf --out=cleaned.pdf

When to use it

PD4ML's HTML-to-PDF converter already writes its own output compactly, with nothing to enable (see Compact PDF output). The optimizer is for PDF files that already exist, whoever produced them: documents created elsewhere or before, edited by incremental updates, or carrying the same resource twice. A file without such leftovers shrinks little, because the optimizer removes objects and does not re-encode stream data.

In this section

• PD4ML PDF Optimizer API - Programmer's Manual: how optimizing works, the PdfOptimizer methods and Options, duplicate-content rules, error messages, encryption, linearized output and threading.
• Examples: seven runnable example classes from the pd4ml-examples repository, from a basic cleanup to error handling and deduplication.
• pd4optimizecli Reference: command-line flags, exit codes, examples and troubleshooting for Pd4OptimizeCli.
• Javadoc: API reference for the com.pd4ml.pdf.optimizer package.