Skip to content
PDF Oxide

Getting Started with PDF Oxide (Scala)

PDF Oxide is the fastest PDF library for the JVM with built-in text extraction — 0.8ms mean, 100% pass rate on 3,830 PDFs. The Scala 3 binding is a thin, idiomatic facade over the mature Java binding: it adds zero native code and layers Scala extension methods that turn java.util.Optional[T] into Option[T] and java.util.List[T] into Seq[T]. The AutoCloseable handles work directly with scala.util.Using.

Installation

Add the dependency to your build.sbt:

libraryDependencies += "fyi.oxide" % "pdf-oxide" % "0.3.69"

The Scala facade depends on the fyi.oxide:pdf-oxide Java binding, which owns the single JNI native bridge. Scala 3.3+ is required.

Quick Start

Build a PDF from Markdown, then open it and extract text back out. Using.resource closes each handle for you.

import fyi.oxide.pdf.{Pdf, PdfDocument, producerOption}
import scala.util.Using

Using.resource(Pdf.fromMarkdown("# Hello pdf_oxide\n\nThis is a **Scala** binding.\n")): pdf =>
  Using.resource(PdfDocument.open(pdf.save())): doc =>
    println(s"pages:    ${doc.pageCount()}")
    println(s"producer: ${doc.producerOption.getOrElse("(none)")}")
    println(doc.extractText(0))

Pdf.fromMarkdown returns a Pdf handle; pdf.save() serializes it to an Array[Byte]. PdfDocument.open accepts those bytes and exposes the document API.

Text Extraction

Plain Text

Extract plain text from any page by its zero-based index.

import fyi.oxide.pdf.PdfDocument
import scala.util.Using

Using.resource(PdfDocument.open(pdfBytes)): doc =>
  assert(doc.isOpen)
  val text = doc.extractText(0)
  println(text)

Markdown and HTML

Convert the whole document to Markdown or HTML in a single call.

import fyi.oxide.pdf.PdfDocument
import scala.util.Using

Using.resource(PdfDocument.open(pdfBytes)): doc =>
  println(doc.toMarkdown())
  println(doc.toHtml())

Page Elements

doc.page(i) returns a PdfPage. The facade exposes each element extractor as a Scala Seq via the *Seq extension methods: wordsSeq, linesSeq, charsSeq, tablesSeq, imagesSeq, and annotationsSeq. Each TextWord carries its text and a bbox.

import fyi.oxide.pdf.{PdfDocument, wordsSeq, linesSeq, charsSeq, tablesSeq, imagesSeq, annotationsSeq}
import scala.util.Using

Using.resource(PdfDocument.open(pdfBytes)): doc =>
  val page = doc.page(0)
  println(s"size: ${page.width()} x ${page.height()}")

  page.wordsSeq.take(8).foreach { w =>
    println(s"  ${w.text} @ ${w.bbox}  (w=${w.bbox.width})")
  }

  println(s"lines:       ${page.linesSeq.size}")
  println(s"chars:       ${page.charsSeq.size}")
  println(s"tables:      ${page.tablesSeq.size}")
  println(s"images:      ${page.imagesSeq.size}")
  println(s"annotations: ${page.annotationsSeq.size}")

You can also iterate every page as a Seq with doc.pagesSeq (its size matches doc.pageCount()).

import fyi.oxide.pdf.{PdfDocument, pagesSeq, wordsSeq}
import scala.util.Using

Using.resource(PdfDocument.open(pdfBytes)): doc =>
  doc.pagesSeq.zipWithIndex.foreach { (page, i) =>
    println(s"page $i: ${page.wordsSeq.size} words")
  }

doc.searchSeq(query) returns a Seq[SearchMatch]. Each match exposes its text.

import fyi.oxide.pdf.{PdfDocument, searchSeq}
import scala.util.Using

Using.resource(PdfDocument.open(pdfBytes)): doc =>
  val matches = doc.searchSeq("Hello")
  println(s"${matches.size} match(es)")
  matches.foreach(m => println(s"  ${m.text}"))

Metadata as Option

Nullable document metadata surfaces as Option[String] through producerOption and creatorOption, so you handle absent values the Scala way.

import fyi.oxide.pdf.{PdfDocument, producerOption, creatorOption}
import scala.util.Using

Using.resource(PdfDocument.open(pdfBytes)): doc =>
  println(doc.producerOption.getOrElse("(unknown producer)"))
  println(doc.creatorOption.getOrElse("(unknown creator)"))

  // Form fields come back as a Seq too:
  println(s"form fields: ${doc.formFieldsSeq.size}")

Rendering

doc.render(i) rasterizes a page and returns the encoded image bytes.

import fyi.oxide.pdf.PdfDocument
import scala.util.Using

Using.resource(PdfDocument.open(pdfBytes)): doc =>
  val png = doc.render(0)
  java.nio.file.Files.write(java.nio.file.Path.of("page-0.png"), png)

Auto Extraction

AutoExtractor.of(doc).extractDocument() returns an AutoResult with the extracted text, optional markdown/html renderings, and the list of pages that still need OCR — all exposed idiomatically via the facade (markdownOption, htmlOption, pagesNeedingOcrSeq).

import fyi.oxide.pdf.{PdfDocument, AutoExtractor, markdownOption, htmlOption, pagesNeedingOcrSeq}
import scala.util.Using

Using.resource(PdfDocument.open(pdfBytes)): doc =>
  val result = AutoExtractor.of(doc).extractDocument()
  println(result.text)
  result.markdownOption.foreach(println)
  result.htmlOption.foreach(println)
  println(s"pages needing OCR: ${result.pagesNeedingOcrSeq}")

Editing

DocumentEditor.open opens an existing PDF for structural edits. Here we scrub metadata and serialize the result back to bytes.

import fyi.oxide.pdf.DocumentEditor
import scala.util.Using

Using.resource(DocumentEditor.open(pdfBytes)): editor =>
  assert(editor.isOpen)
  editor.scrubMetadata()
  val cleaned: Array[Byte] = editor.save()
  java.nio.file.Files.write(java.nio.file.Path.of("scrubbed.pdf"), cleaned)

Next Steps