# Measurement fixtures
The shared reader uses one canonical beamline collection in `xas/` across Rust,
Python, TypeScript/WebAssembly and GUI tests. Tests read local files; they never
download or regenerate measurements. The fixture bundles are excluded from
crates.io. The two original manifests below are historical catalogs of that
shared data, not two physical copies.
| [Beamline measurements](xas/README.md) | 154 valid files, grouped by facility and beamline | [Per-file source manifest](xas/manifest.json), [citations](xas/CITATIONS.md), [license index](xas/LICENSE.md) |
| [Expanded measurement corpus](xas/collections/rexafs-corpus/INDEX.md) | 222 files: 182 parsed, including five partial HDF5 recoveries; 40 known rejections | [Original manifest](xas/collections/rexafs-corpus/manifest.json), [licenses](xas/collections/rexafs-corpus/LICENSE.md), [snapshot checksums](xas/collections/rexafs-corpus/SNAPSHOT.json) |
| [Saved sessions](sessions/README.md) | 6 XTUNES files, 15 valid Larix sessions and 6 intentionally invalid Larix sessions | Separate original bundle manifests, source records and MIT notices |
| [Copper mixture recovery](analysis/cu-mixtures/README.md) | 100 generated mixtures and three prepared standards | Exact fractions, seed, source checksum and explicit permission record; upstream open-data license not established |
| [Experimental Larch comparisons](analysis/experimental-larch/README.md) | Two measured Cu foils and Aichi RuO₂, referenced from `xas/`; full MBACK and wavelet expected arrays | Source hashes, CC0/CC-BY-NC-SA attribution, pinned Larch functions and package versions |
The `xas/` and session suites cover **175 readable inputs**. Readability includes preserved tables,
images and saved results that require explicit interpretation before conversion
to an absorption spectrum. The six invalid sessions are rejection cases, not
supported measurements. See the [reader guide](../../../../doc/measurement-reader.md)
for the distinction and current format limitations.
The expanded corpus was copied from the private `rexafs-format` gathering
checkout after integrity and provenance review. All 571 original data and
attribution files remain available byte for byte through its
[path map](xas/collections/rexafs-corpus/paths.json). Its README/research documents
describe the historical collection; no external checkout is needed for tests.
The [audit](../../../../doc/validation/2026-09-14-measurement-corpus/review.md)
records format gaps and numerical regressions. Across `xas/` and `rexafs-corpus/`
there are **243 unique measurement/project payloads** (133 occur in both
manifests and now share their stored file). Of those, 203 parse, including partial recoveries, and 40 have
explicit rejection expectations. The 21 valid application sessions are separate.
## Layout and tests
- `xas/samples/` retains original beamline files in facility/beamline/source order.
- `xas/candidates/kek-pf/` retains five QD measurements for academic, nonmilitary testing under the [original custom usage notice](xas/candidates/kek-pf/README.md).
- `xas/collections/rexafs-corpus/` retains the expanded collection's original
manifests, snapshot and historical research records. `paths.json` maps every
former path to its canonical file under `xas/`; `SNAPSHOT.json` continues to
pin the original bytes. The integrity check verifies all entries, including
attribution sidecars. No symbolic links or generated fixture copies are needed.
- `sessions/` retains complete application-generated bundles, including original
licenses, manifests, expected values and generation evidence.
- [`../measurement_reader.rs`](../measurement_reader.rs) runs self-contained
parser and conversion contracts from `../measurement_reader/`. Synthetic inputs
are declared in those tests and remain in the published crate archive.
- [`../measurement_fixtures.rs`](../measurement_fixtures.rs) runs the new corpus,
reference-format and session checks from `../measurement_fixtures/`. Its
[coverage table](../measurement_fixtures/coverage.csv) stores expected reader
output separately from the original data manifests.
- [`../beamline_formats.rs`](../beamline_formats.rs) retains the original
specialized-reader regression target.
From the repository root:
```sh
cargo test --locked -p rexafs --test measurement_reader --test measurement_fixtures --test beamline_formats
uv run --no-project --python 3.12 scripts/check-beamline-fixtures.py
uv run --no-project --python 3.12 scripts/check-beamline-fixtures.py --package
```
## Adding data
For a raw beamline example, follow the [corpus contribution workflow](xas/INTEGRATION.md).
Retain its source URL, revision, upstream license text, original filename, byte
length and checksum. For an application session, retain those records beside
the saved file and identify the source measurements, application version and
save/reopen or generation procedure. Record which expectations come from an
independent reference and which describe rexafs behavior. Keep synthetic damage
cases clearly labeled and linked to their valid parent.
Data licenses apply per source. The library's MIT/Apache-2.0 license does not
replace upstream fixture terms or license third-party applications. Do not infer
redistribution permission merely because a repository is public. Keep notices
with any copied or derived data and preserve historical records unchanged.
The 2026-09-14 consolidation removed 356 identical copies, including 133
measurements, saving 105,329,736 bytes of checkout space. It relocated 216 unique
expanded-collection files without changing their bytes. All coverage expectations,
source URLs, license records and historical manifests remain available.