Changelog

Unreleased

The package license in pyproject.toml changes from MIT to CC0-1.0, the
license in the repo's LICENSE.
The documentation is now in Spanish (docs/aug/es/) and English
(docs/aug/en/). Both are published to
docs.jjsm.science
and linked as translations of the same page.

aug.analysis.aurora_lbo simulates LBO injections with full Aurora
transport (aurora_sim, every charge state, no coronal equilibrium) and
fits D(ρ), v(ρ) and the source duration to the measured curve (make_sim,
run_lbo, fit_injection, measured_curve, transport_profiles). It
works without a geqdsk (approximate geometry) and without OMFIT (a minimal
stand-in for atomic_element).
load_adf11_subset reads files with several elements (Fe, Ni) and a
varying number of charge states; element parameter.
Docs: aurora_lbo.md.

read_imp/read_rest: in the -rest.txt files the Zeff signal comes in
two identical columns, but the header names it only once. Before this fix,
every signal after Zeff was shifted by one column: tau_tot held Zeff
(which is why it read ~1.2 to 2), H/L-facs;8 held tau_tot, P_OH held the
H factor, IBu1 held P_OH, and the real RMP current ended up in _col0.
A Zeff_dup column is now inserted after Zeff when the duplicate is
detected. Shots without Zeff (#38858/9) were already correct.
Effect on 0.3.0: tau_E_estimate added the H factor (~0.6 to 1.1) as if it
were P_OH in MW, underestimating τ_E by ~10 to 20 %. τ_imp and the columns
before Zeff do not change.

aug.io.read_lbo_profiles: every IDA profile (ne, Te) in a -lboN.txt,
with values.
aug.analysis.plasma: per-injection parameters for correlations. It covers
τ_E and the H factor from the TOT diagnostic, τ_E = W/P_loss, H98 computed
with IPB98(y,2), I_RMP (IBu1), and edge gradients and peaking of ne/Te
(profile_gradients, rest_parameters, parameter_table).
aug.analysis.atomic: W ionization equilibrium with Aurora + ADF11
(load_w_atom_data, w_fractional_abundance), where W44+ sits in a profile
(w44_profile_metrics), a toy LBO with fixed transport (toy_lbo_signal),
and shape metrics (signal_metrics).
Docs: plasma.md, atomic.md.

aug.analysis.compare: metrics for comparing species. Decay curvature index
(decay_shape, segment_tau), mean residual and curves collapsed in t/τ
(residual_profiles, collapsed_curves), delay between charge states of
the same species (multiline_timing), tau_E_estimate, plasma_response
and comparison_table.
SPECIES_Z, SPECIES_LINES.
Docs: compare.md.

aug.analysis.injections: per-injection pipeline (detection in the signal,
onset/peak, rise times, decay window, context, flags) and InjectionSet.
It generalizes the 09W3 code to every species.
aug.models.exponential: fit_single_exp, fit_double_exp (multi-start,
physical bounds, identifiability criterion), select_decay_model,
block_bootstrap_tau, fit_rise_decay, exp1, exp2, rise_decay.
aug.io.qc.check_files: QC of the Trace= header and MD5 duplicates.
aug.plot.injections: overview, aligned curves, fit mosaic, single vs.
double, window sensitivity.
Shot.emission, Shot.default_signal, Shot.find_injections,
Shot.analyze_injections, Shot.folder; GroupShots.analyze_injections,
GroupShots.qc_table.
DEFAULT_SIGNALS (core line per species).

Shot.fit_tau_imp fits from the detected peak, on the normalized curve,
with a trimmed window. It used to fit from t_lbo, rise included. The
t_pre, t_post and time_interval arguments are ignored.
Shot.decay uses the physical emission (ysc sign corrected) and
default_signal. It used to take signals[0], which for Fe was Fe14_284.
GroupShots.from_directory automatically flags shots whose files belong to
another shot (flag_bad_files=True): #38884.
REGIME_MAP: EDA_Fe-LBO-H becomes ECRH-high (it was ICRF, but the
data has no ICRF); RMP_W-LBO and *-NBI become NBI+ECRH.
GroupShots.summary() includes the folder column.

Shot, GroupShots, Spectrum; readers read_imp, read_rest,
read_lbo, build_catalog; fit_exponential_decay,
fit_double_exponential, compare_decay_models, bootstrap_decay_fit;
figures aug.plot.shot and aug.plot.group.