BamboostOther Data
Beyond parameters,
metadata, series,
and meshes, you can store any arbitrary data
directly in the HDF5 hierarchy using the generic Group and
Dataset API exposed through sim.root.
Storing custom groups and datasets
sim.root is a Group pointing at
/ inside data.h5. Navigate and create sub-groups exactly as you
would with h5py, but through the bamboost wrappers that handle lazy
file access and MPI safely.
import numpy as np
sim = coll["my-simulation"].edit()
# Create a group (idempotent, like h5py.require_group)
stats = sim.root.require_group("statistics")
# Store an attribute on the group
stats.attrs["final_residual"] = 3.2e-6
# Store a numpy array as a dataset
stats["residual_history"] = np.array([1e-1, 5e-2, 1e-2, 5e-3, 1e-3])
# Nest groups as deeply as you need
checkpoints = stats.require_group("checkpoints")
checkpoints["iterations"] = np.array([10, 20, 30], dtype=np.int64)Reading custom data back
sim = coll["my-simulation"]
stats = sim.root["statistics"]
# Read an attribute
residual = stats.attrs["final_residual"]
# Read a dataset (lazy reference—slice to load into memory)
history = stats["residual_history"][:]
# Read a nested dataset
iters = sim.root["statistics/checkpoints/iterations"][:]Git snapshot
bamboost includes a built-in helper to record the current state of a
git repository alongside the simulation data. This is useful for full
reproducibility.
sim = coll["my-simulation"].edit()
sim.git.add("my_solver", repo_path="/path/to/solver")This stores the remote origin URL, the current commit hash, the active
branch, and a git diff HEAD patch in the .git group inside
data.h5.
# Read back git information
git_item = sim.git["my_solver"]
print(git_item.commit) # e.g. "a3b1f29..."
print(git_item.branch) # e.g. "main"
print(git_item.origin) # e.g. "https://github.com/..."
print(git_item.patch) # the diff as a string