CapyPDF is a low level library for generating PDF files. It does not have its own document model, it merely exposes PDF primitives directly.
- Aims to eventually support all functionality in PDF, including accessibility features
- Deprecated PDF features are not supported by default, but support may be added for important, widely used features
- Reads PNG, JPEG and TIFF files
- Fully color managed using LittleCMS 2
- Not implemented in C
- Provides a plain C API for easy integration into scripting languages
- Ships with a
ctypesPython binding and a C++ wrapper header - Minimal dependencies
- Creates PDF 2.0 unless chosen PDF type (X, A, etc) requires a specific older version
- Reading PDF files
- Modifying PDF files
- Cryptographic operations (i.e. document signing)
- Supporting any other output format than PDF
- Parsing any vector data, eg SVG
- Data conversions in general (apart from colorspaces)
The simplest way to see how the library is meant to be used is to take a look at the unit tests. Each test is written to also be a readable example on how to generate a specific type of document.
A low res image of all tests that generate output can be found in the test outputs directory.
Until 1.0 there is no guarantee of any kind. Anything can be changed. However we try not to change things without a good reason. Once 1.0 happens, we aim to provide the following:
- The plain C interface is both API and ABI stable
- The Python API shall be stable as well
- Nothing else is stable
The library shall always generate PDFs that are syntactically valid. Any deviation is a bug that should be reported.
The output is not guaranteed to be semantically valid. PDF has structural validity requirements that a plain PDF generation library can not guarantee to hold. That work needs to be done by the generating application. CapyPDF does have some semantic checks, such as not permitting RGB images in PDF/X3 documents, but they are implemented on a best effort basis.
The basic functionality is there and the library can be used to generate fairly complex documents. The APIs are not stable yet, however they are not expected to change much any more.
The use of any and all AI tools for this project is prohibited. Issues and pull requests created using AI will not be looked at, but instead closed immediately upon detection.