v3.4.1

Fixed:

• Prevent downgrading file versions when opening in write mode.

v3.4.0

Added:

• GSD_TYPE_CHARACTER chunk type for string data - valid in file layer versions 2.1 and later (#391).

Changed:

• The log dictionary in a hoomd frame now accepts string values. read_log correspondingly produces a NumPy array of variable length strings (#391).
• Require NumPy >= 2.0 (#391).

v3.3.2

Added:

• Support Python 3.13 (#388).

Removed:

• Support for Python 3.9 (#388).

v3.3.1

Changed:

• Improved build instructions (#376).

v3.3.0

Added:

• Support numpy 2.0 (#367).

Changed:

• Navigate the documentation with arrow keys (#365).
• Add button to copy code snippets from the documentation (#365).
• Remove singularity reference from the documentation (#365).

Removed:

• Support for Python 3.8 (#367).

v3.2.1

Fixed:

• Write all pending index entries to the file when gsd_flush() is called after gsd_write_chunk() and before gsd_end_frame() (#319).
• Readthedocs builds with pandas 2.2.0 (#322).
• Import without seg fault when built with CMake on macOS (#323).
• Internal cached data remains valid when users modify frames obtained by indexing trajectories (#324).

Changed:

• Provide support via GitHub discussions (#308).
• Use ruff (#317).
• Perform fewer implicit flushes when using the gsd.hoomd python API (#325).

v3.2.0

Added:

• Support Python 3.12 (#283).

v3.1.1

Fixed:

• Raise a FileExistsError when opening a file that already exists with mode = 'x'.

v3.1.0

Fixed:

• hoomd.read_log no longer triggers a numpy deprecation warning.

Added:

• HOOMDTrajectory.flush - flush buffered writes on an open HOOMDTrajectory.

v3.0.1

Fixed:

• Prevent ValueError: signal only works in main thread of the main interpreter when importing gsd in a non-main thread.