atomlib is a package for creating, modifying, and controlling atomic structures. It draws heavy inspiration from previous tools like Atomsk and ASE, but attempts to provide a cleaner, more consistent interface that can be used from Python or a command line.

atomlib has minimal dependencies: numpy, scipy, and polars are required for core atom structure manipulation, and click is required for command line functionality.

Atomic structures are stored as polars DataFrames, providing a clean, immutable interface that maximizes expressiveness and minimizes errors. For formats that allow arbitrary properties, these properties can be passed through transparently. atomlib has first-class support for fractional occupancy, Debye-Waller factors, atomic forces, and labels.

Translational symmetry is stored in Cell objects, which represent a fully generic cell. Atoms can be modified in any coordinate system that makes sense (global, local real-space, cell fraction, box fraction, etc.). Support for non-translational symmetry operations is limited at this point.

For more information, check out the example notebooks and the API documentation.

File format support is still a work in progress. Where possible, parsers are implemented from scratch in-repo. Most formats are implemented in two steps: Parsing to an intermediate representation which preserves all format-specific information, and then conversion to the generic Atoms & AtomCell types for manipulation & display. This means you can write your own code to utilize advanced format features even if they're not supported out of the box.