How-to guides

Start with Sample for routine recording, transformation, selection, correction, aggregation, claims and archive hand-offs. These guides take one task further. Where a task needs the underlying container, keep the same sample and use sample.core rather than starting another history.

Working with spectra

• Resample a spectrum onto a new grid — move data between axes with linear or SciPy-backed interpolation.
• Control the operation registry — add custom operations or lock an application down to approved ones.
• Handle missing extras — fail helpfully when an optional extra is not installed.

Scientific Python

• Use scientific Python libraries — DataFrames in and out, your own NumPy or SciPy operations, and direct plotting.
• Author a connector — write the loader that turns your file’s bytes into runs and checkpoints.

Combining runs

• Replace a checkpoint — when better evidence arrives, rather than a mistake to fix.
• Align mixed-resolution members — aggregate replicates measured on different grids.

Reproducibility

• Replay a committed chain — re-derive a result from its recorded recipe.
• Anchor verification with receipts — prove the manifest is the one you recorded.
• Inspect a damaged archive — see exactly what is wrong without loading the file.

Disclosure and claims

• Attach author disclosure — ship environment records and source bundles, stored but never executed.
• Record scientific assertions — typed claims with evidence, supersession, and retraction.
• Check external evidence — verify a citation into another archive, offline.

Finding and fixing problems

• Handle selection and data errors — recover from a missing run, a repeated run name, or data that the commit boundary refuses.

Datasets

• Prepare a machine-learning dataset — align archives onto one grid with leak-free splits and a provenance digest.