HDF5 dimension scale
An HDF5 dimension scale is a dataset associated with an array dimension to supply application-defined coordinates or labels.
Identity: a scale's name is separate from its dataset path and need not be unique. h5py name lookup returns the first matching attached scale, while the HDF5 storage profile does not define scale-list order. Do not treat insertion order as a portable selection rule. [2] [1]
Consistency: an axis label does not attach a scale. Coordinate interpretation and compatible extents are application responsibilities; several scales or unequal extents may be intentional. [1]
Observed consequence: in two synthetic HDF5 builds, selecting between duplicate time-scale names changed an endpoint rate from 10 to 1 count/s with identical value-array bytes. Thirteen of fourteen cases were numerically readable, but only two met the experiment's single-time-axis policy. These are configuration outcomes, not reliability frequencies. [3]
Relations
Citations
- The HDF Group: HDF5 High Level Dimension Scales, storage profile and limitations
- h5py: Dimension Scales
- Hesela: HDF5 coordinate selection, two-build controlled observations
Sources & caveats
- Hesela's synthetic files were observed with h5py 3.10.0/HDF5 1.14.2 and h5py 3.15.0/HDF5 1.14.6. Writers closed before fresh readers opened; no live acquisition, H5MD consumer or power-loss behavior was tested.