What is the fastest Python PDF library?

PDF Oxide is the fastest Python PDF library, with 0.8ms mean text extraction time — 5.8× faster than PyMuPDF (4.6ms) and 15× faster than pypdf (12.1ms). Benchmarked on 3,830 real-world PDFs with 100% pass rate.

Is PDF Oxide free for commercial use?

Yes. PDF Oxide is MIT licensed — free for all uses including commercial products, SaaS, and proprietary software. No license fees, no sales calls, no AGPL restrictions.

Can PDF Oxide handle scanned PDFs with OCR?

Yes. PDF Oxide includes built-in OCR via PaddleOCR and ONNX Runtime. No Tesseract installation needed — just pip install pdf_oxide and use extract_text_ocr(). Supports PP-OCRv3, v4, and v5 models.

Does PDF Oxide support XFA forms?

Yes. PDF Oxide is the only Python PDF library that can detect, analyze, and extract data from XFA forms (XML Forms Architecture). PyMuPDF, pypdf, pdfplumber, and pdfminer cannot read XFA form data.

How does PDF Oxide compare to PyMuPDF?

PDF Oxide is 5.8× faster than PyMuPDF (0.8ms vs 4.6ms mean), has a 100% pass rate vs 99.3%, and is MIT licensed vs PyMuPDF's AGPL-3.0. PDF Oxide also has built-in Markdown/HTML output and XFA form support that PyMuPDF lacks.

Can PDF Oxide convert PDF to Markdown?

Yes. PDF Oxide has built-in PDF to Markdown conversion with heading detection, table preservation, and list formatting — ideal for LLM and RAG pipelines. No separate package needed, unlike PyMuPDF which requires pymupdf4llm (69× slower).

DocumentBuilder 流式 API

DocumentBuilder 是一套流式 API，用于逐页构建 PDF 文档，可全面控制文本布局、字体、注释、表单字段和内容元素。

绑定覆盖情况（v0.3.38）。 DocumentBuilder + FluentPageBuilder + EmbeddedFont 已在 Rust、Python、Node/TypeScript、C#、Go 和 WASM 中提供。所有绑定都暴露相同的能力：多页构建、通过嵌入字体支持 CJK／西里尔／希腊文本、15 种注释方法、5 种 AcroForm 控件类型、图形基本图元（rect、filled_rect、line）、AES-256 加密以及 HTML+CSS 流水线。

快速示例

Rust

use pdf_oxide::writer::{DocumentBuilder, PageSize, DocumentMetadata};

let mut builder = DocumentBuilder::new()
    .metadata(DocumentMetadata::new().title("My Document"));

builder
    .page(PageSize::Letter)
        .at(72.0, 720.0)
        .heading(1, "Hello, World!")
        .paragraph("This is a PDF document created with DocumentBuilder.")
        .done();

let bytes = builder.build()?;
std::fs::write("output.pdf", bytes)?;

Python

from pdf_oxide import DocumentBuilder, EmbeddedFont

font = EmbeddedFont.from_file("DejaVuSans.ttf")
pdf = (DocumentBuilder()
    .register_embedded_font("DejaVu", font)
    .a4_page()
        .font("DejaVu", 12).at(72, 720).text("Hello, World!")
        .highlight((1.0, 1.0, 0.0))
    .done()
    .build())
open("output.pdf", "wb").write(pdf)

Node / TypeScript

import { DocumentBuilder, EmbeddedFont } from "pdf-oxide";

const font = await EmbeddedFont.fromFile("DejaVuSans.ttf");
const bytes = new DocumentBuilder()
  .registerEmbeddedFont("DejaVu", font)
  .a4Page()
    .font("DejaVu", 12).at(72, 720).text("Hello, World!")
    .highlight([1.0, 1.0, 0.0])
  .done()
  .build();

using PdfOxide;

using var font = EmbeddedFont.FromFile("DejaVuSans.ttf");
var bytes = DocumentBuilder.Create()
    .RegisterEmbeddedFont("DejaVu", font)
    .A4Page()
        .Font("DejaVu", 12).At(72, 720).Text("Hello, World!")
        .Highlight(1.0, 1.0, 0.0)
    .Done()
    .Build();
File.WriteAllBytes("output.pdf", bytes);

Go（CGo 方式；需要 CGO_ENABLED=1）

import pdfoxide "github.com/yfedoseev/pdf_oxide/go"

font, _ := pdfoxide.EmbeddedFontFromFile("DejaVuSans.ttf")
defer font.Close()

builder := pdfoxide.NewDocumentBuilder()
builder.RegisterEmbeddedFont("DejaVu", font)
builder.A4Page().
    Font("DejaVu", 12).At(72, 720).Text("Hello, World!").
    Highlight(1.0, 1.0, 0.0).
    Done()
bytes, _ := builder.Build()

WASM（浏览器／Node／打包器 — API 相同）

import init, { DocumentBuilder, EmbeddedFont } from "pdf-oxide-wasm/web";
await init();

const font = await fetch("/DejaVuSans.ttf").then(r => r.arrayBuffer());
const bytes = new DocumentBuilder()
  .registerEmbeddedFont("DejaVu", EmbeddedFont.fromBytes(new Uint8Array(font), "DejaVu"))
  .a4Page()
    .font("DejaVu", 12).at(72, 720).text("Hello, World!")
  .done()
  .build();

C++

#include <pdf_oxide/pdf_oxide.hpp>

auto font = pdf_oxide::EmbeddedFont::from_file("DejaVuSans.ttf");
auto builder = pdf_oxide::DocumentBuilder::create();
builder.register_embedded_font("DejaVu", font);
builder.a4_page()
    .font("DejaVu", 12).at(72, 720).text("Hello, World!")
    .highlight(1.0, 1.0, 0.0)
    .done();
std::vector<std::uint8_t> bytes = builder.build();

Swift

import PdfOxide

let font = try EmbeddedFont.fromFile("DejaVuSans.ttf")
let builder = try DocumentBuilder.create()
try builder.registerEmbeddedFont("DejaVu", font)
let page = try builder.a4Page()
try page.font("DejaVu", 12).at(72, 720).text("Hello, World!")
    .highlight(1.0, 1.0, 0.0)
    .done()
let bytes = try builder.build()

Dart

import 'package:pdf_oxide/pdf_oxide.dart';

final font = EmbeddedFont.fromFile('DejaVuSans.ttf');
final builder = DocumentBuilder.create();
builder.registerEmbeddedFont('DejaVu', font);
builder.a4Page()
    .font('DejaVu', 12.0).at(72.0, 720.0).text('Hello, World!')
    .highlight(1.0, 1.0, 0.0)
    .done();
final bytes = builder.build();

library(pdfoxide)

font <- pdf_embedded_font_from_file("DejaVuSans.ttf")
b    <- pdf_builder_create()
pdf_builder_register_embedded_font(b, "DejaVu", font)
pg <- pdf_builder_a4_page(b)
pdf_page_font(pg, "DejaVu", 12)
pdf_page_at(pg, 72, 720)
pdf_page_builder_text(pg, "Hello, World!")
pdf_page_highlight(pg, 1.0, 1.0, 0.0)
pdf_page_done(pg)
bytes <- pdf_builder_build(b)

Julia

using PdfOxide

emb = embedded_font_from_file("DejaVuSans.ttf")
b   = DocumentBuilder()
register_embedded_font(b, "DejaVu", emb)
pg = a4_page(b)
font(pg, "DejaVu", 12)
at(pg, 72, 720)
text(pg, "Hello, World!")
highlight(pg, 1.0, 1.0, 0.0)
done(pg)
bytes = build(b)

Zig

const pdf_oxide = @import("pdf_oxide");
const a = std.heap.page_allocator;

var font = try pdf_oxide.EmbeddedFont.fromFile("DejaVuSans.ttf");
var builder = try pdf_oxide.DocumentBuilder.create();
try builder.registerEmbeddedFont("DejaVu", &font);
var page = try builder.a4Page();
try page.font("DejaVu", 12);
try page.at(72, 720);
try page.text("Hello, World!");
try page.highlight(1.0, 1.0, 0.0);
try page.done();
const bytes = try builder.build(a);

Objective-C

#import "POXPdfOxide.h"
NSError *err = nil;

POXEmbeddedFont *font = [POXEmbeddedFont fromPath:@"DejaVuSans.ttf" error:&err];
POXDocumentBuilder *builder = [POXDocumentBuilder createWithError:&err];
[builder registerEmbeddedFont:@"DejaVu" font:font error:&err];
POXPageBuilder *page = [builder a4PageWithError:&err];
[page font:@"DejaVu" size:12 error:&err];
[page at:72 y:720 error:&err];
[page text:@"Hello, World!" error:&err];
[page highlightR:1.0 g:1.0 b:0.0 error:&err];
[page done:&err];
NSData *bytes = [builder buildWithError:&err];

Elixir

{:ok, font} = PdfOxide.font_from_file("DejaVuSans.ttf")
{:ok, db}   = PdfOxide.builder()
:ok = PdfOxide.builder_register_embedded_font(db, "DejaVu", font)
{:ok, page} = PdfOxide.builder_a4_page(db)
:ok = PdfOxide.page_font(page, "DejaVu", 12)
:ok = PdfOxide.page_at(page, 72, 720)
:ok = PdfOxide.page_text(page, "Hello, World!")
:ok = PdfOxide.page_highlight(page, 1.0, 1.0, 0.0)
:ok = PdfOxide.page_done(page)
{:ok, bytes} = PdfOxide.builder_build(db)

完整 API 参考