Tensorium_tex is a symbolic parser and code generation tool designed to bridge LaTeX mathematical notation and tensor-based computation backends (such as MLIR or C++). The goal of this project is to provide a robust frontend that parses LaTeX expressions, builds a structured Abstract Syntax Tree (AST), and enables automatic code generation for high-performance tensor algebra and general relativity.
- LaTeX Parser: Converts LaTeX expressions (including tensor indices and subscripts/superscripts) into an internal AST representation.
- AST Utilities: Tools for manipulating and analyzing the parsed abstract syntax tree.
- Tensor Indexing Support: Support for complex tensor index manipulation (covariant, contravariant, partial derivatives, etc.).
- Code Generation Ready: Backend-agnostic, with planned support for MLIR and C++ code generation.
- Test Suite: Simple tests to validate parser and AST functionality.
Requirements:
- C++17 or newer compiler (tested with Clang and GCC)
- CMake (recommended)
- (Optional) MLIR/LLVM if targeting MLIR backend (future)
mkdir build
cd build
cmake ..
makeGiven a LaTeX file like:
$R_{\mu\nu} = R^{\lambda}_{\ \mu\lambda\nu}$Tensorium_tex will parse the expression and build the corresponding AST, ready for further processing or code generation.
- Robust LaTeX → AST pipeline for tensor and index-rich expressions
- Symbolic manipulation and simplification
- Code generation for MLIR and C++ (planned)
- Extensible design for additional backends and transformations
PRs, issues, and suggestions are welcome! Please use English for all code and documentation.
This project is licensed under the MIT License. See the LICENSE file for details.
For questions or discussions, please open an issue on the GitHub repository or contact the maintainers. @at0m741