Expand description
Checked, vendor-neutral raw and processed NMR data models.
read detects and materializes a raw acquisition or processed spectrum.
raw::open is the lazy, raw-only reader for trace and region access. Readers do
not apply FFT, filtering, phasing, scan normalization, interpolation, or
NUS reconstruction.
Public complex samples use R + iI. Readers apply a vendor-to-public sign
transform only when supported by format evidence. Known nucleus labels use
mass-number-first ASCII notation such as 1H and 13C; original vendor text
remains available through vendor/source metadata.
Detection identifies and selects a candidate format. It does not promise that
the selected scientific layout is supported by read or raw::open.
§Experimental algorithms
The supported scientific foundation is checked data models, evidence-aware
reading and explicit deterministic processing. ACME, AsLS, the fixed 512/128
IST kernel and DensePipeline automatic policy have a separate experimental
scope: successful execution is not a general inference-quality guarantee.
Their accepted data domains, diagnostics and future replacement profiles are
documented independently of the stable data and explicit-operation contracts.
This project has not been released. APIs, algorithms and schemas are updated
in place during development, retaining their initial version identifiers.
Development snapshots have no backward-compatibility or migration guarantee.
Closed mathematical sets (raw/processed kind and positive/negative Fourier
exponent) remain exhaustive. Extensible format/error/result classifications
are non-exhaustive. Constructible request variants retain their listed fields;
new contracts use new variants or checked parameter objects. Read-only result
variants may acquire facts and require .. in external matches.
Algorithm quality and bitwise numerical reproducibility are separate from interface compatibility. Histories record the current algorithm identifiers and execution environment. Replay success does not compare output samples; numerical comparison is caller-owned.
let loaded = nmr::read("data/experiment")?;
println!("{:?}", loaded.source_format());Re-exports§
pub use acquisition::AxisIndex;pub use dataset::Dataset;pub use dataset::ProcessedDatasetWithContext;pub use dataset::RawDatasetWithContext;pub use execution::CancellationToken;pub use execution::ExecutionContext;
Modules§
- acquisition
- Canonical, vendor-neutral acquisition semantics.
- axis
- Vendor-neutral mathematical axis vocabulary.
- data
- Owned region blocks and bounded access to an immutable Dataset.
- dataset
- Materialized datasets with private storage and retained reading context.
- derivation
- Evidence-bearing library derivations that establish a new processing input.
- execution
- Synchronous cooperative execution control, independent of a host task runtime.
- execution_
report - Portable, versioned execution reports and independent numerical comparison.
- export
- Deterministic, bounded PlotData NPZ V1 export.
- external
- Checked boundaries for algorithms executed by the host application.
- formats
- Vendor-specific parsers and advanced in-memory decoding APIs.
- plot
- Checked scalar data prepared for plotting or export.
- processed
- Dense scientific data products with independently modeled axis domains.
- processing
- Bounded, history-bearing minimal processing for dense rank-1 through rank-3 data.
- provenance
- Source provenance and canonical identity shared by raw and processed datasets.
- raw
- Vendor-neutral raw-dataset model and bounded path access.
- resource
- Payload-capacity resource controls, not process RSS or allocator guarantees.
- snapshot
- Versioned embedded snapshots with exact binary samples and explicit history acceptance. Paths are recorded locators and are never accessed by restoration. The host owns compression, file transactions, and storage of earlier replay inputs.
Structs§
- Dataset
Identity - Portable identity fields established without sequence-name guessing.
- Parameter
Error - Structured vendor parameter failure.
- Read
Candidate - A resolver candidate retained in ambiguity and mismatch errors.
- Read
Error - Error produced while detecting, reading, or accessing a dataset.
- Read
Limits - Resource limits applied while opening and reading a dataset.
- Read
Options - Builder for unified detection and materialized reads.
- Sampling
Declaration - A complete sampling assertion, retaining original indices and their source. This cannot override vendor evidence or change physical trace/lane order.
Enums§
- Dataset
Kind - Whether a loaded dataset contains acquisition or processed samples.
- Format
- A precise supported input format.
- Format
Family - Vendor family shared by raw and processed representations.
- Input
Source - The concrete input associated with a reader failure.
- Metadata
Field - Portable metadata field implicated by a warning.
- Parameter
Error Kind - Stable classification of a vendor parameter failure.
- Read
Error Kind - Stable machine-readable category for a
ReadError. - Read
Error Reason - Structured reason for a detection, decoding, or dataset-access failure.
- Read
Preference - Policy for choosing between raw and processed candidates.
- Read
Resource - Resource governed by a reader byte limit.
- Read
Warning - Non-fatal issue retained by a successful read.
- Sampling
Index Base - Index convention of a caller-supplied sampling table.
- Warning
Impact - Scientific effect of omitted optional information.
Functions§
- detect
- Detects and selects a path using default selection policy.
- read
- Materializes a path using default selection policy.
- read_
with_ context - Reads with default selection policy and shared execution control.
Type Aliases§
- Complex64
- Alias for a
Complex<f64>