Versione verificata
| Provider | math.numeric 1.14.0 |
|---|---|
| API | NUMERIC_API_v1 · api_version=1 |
| Framework | GX-TXT 0.9.8.10-dev-031 |
1. In parole semplici
È la “cassetta degli attrezzi matematica” comune di GX-TXT. Un modulo che deve interpolare, fare una regressione, calcolare una derivata o risolvere una matrice non dovrebbe incorporare una propria implementazione privata se la stessa operazione è già disponibile qui.
2. Perché è un'API e non un modulo da Run
math.numeric è un servizio protetto runner=service. I consumer scoprono l'API, verificano capability e invocano operazioni. Questo separa il “motore matematico” dal workflow scientifico che decide quali Dataset usare e come interpretare il risultato.
3. Catalogo delle possibilità
| Famiglia | Operazioni |
|---|---|
| Elementwise / array | array_unary.*, array_binary.*, array_compare.*, array_mask.*, array_select, array_where |
| Reductions | array_reduce.sum/mean/median/std/variance/rms/min/max/quantile/product/argmin/argmax/count/any/all, reduce_summary, reduce_accumulate |
| Complex | complex_unary.*, complex_binary.*, unwrap |
| Interpolation / grid | interpolate, resample, resample_uniform, spline_create/evaluate/integral, grid_info, validate_monotonic, uniform_grid, common_range |
| Calculus | integrate, cumulative_integrate, differentiate, cumulative_sum |
| Statistics | statistics, rank, covariance, correlation, mad, histogram, ecdf, normal_cdf, student_t_cdf/quantile, f_cdf, hypothesis_test, one_way_anova, matrix_statistics, residual_metrics, quantiles, weighted_quantiles, weighted_statistics |
| Linear algebra | matrix_svd, linear_matrix |
| Vector / random | vector_normalize, vector_similarity, resample_indices |
| Fitting / optimization | least_squares, linear_regression, polynomial_fit/evaluate, nonlinear_fit, root_find, minimize_1d, minimize |
| Signal primitives | smooth, digital_filter, find_peaks, baseline, fft, ifft, analytic_signal, frequency_grid, one_sided, detrend, window, configure_fft_threads |
| Uncertainty | propagate_uncertainty |
| Dataset-aware | dataset_array_*, dataset_integrate/cumulative_integrate/differentiate, dataset_interpolate/resample/resample_uniform, dataset_statistics/reduce_summary/weighted_statistics/grid_info, dataset_polynomial_fit/nonlinear_fit, dataset_smooth/find_peaks/baseline |
| Advanced foundation | foundation(family, action, a, b, c, options) |
4. Statistica
La stessa API contiene sia primitive classiche (statistics, covariance/correlation, MAD, histogram, ECDF, quantili) sia inferenza (hypothesis_test, one_way_anova, probability CDF/quantile) e un gruppo foundation.statistics con metodi robusti, bootstrap/permutation, multiple testing, logistic regression, GLM, survival e clustering.
5. Fitting e ottimizzazione
least_squares, linear_regression, polynomial_fit sono data-only. nonlinear_fit, root_find, minimize_1d e minimize usano callback nel binding Octave/Script Bridge.
6. Complex linear algebra
linear_matrix copre inspect, identity/transpose, multiply, solve, inverse/pseudoinverse, LU/QR/Cholesky/SVD, eigen/generalized eigen e matrix exponential/integer power. Supporta matrici reali e complesse con diagnostica di rank/condition number.
7. Foundation dispatcher
| Family | Actions |
|---|---|
family | align, surface, derivative, extrema, envelope |
tensor | transpose, conjugate_transpose, trace, symmetric_part, antisymmetric_part, contract, apply, outer |
scattered | interpolate |
vector | dot, cross, norm, normalize, outer, projection, angle, triple, contract |
grid | interpolate, sample, compatible, resample |
calculus | integrate2, integrate3, weighted_integral, gradient_grid, line_integral, line_scalar_integral, surface_flux, surface_integral, gradient, jacobian, hessian, directional |
special | erf, erfc, gamma, log_gamma, beta, incomplete_gamma, incomplete_beta, Bessel J/Y/I/K, spherical Bessel, airy, exponential_integral, sinc |
geometry | plane/frame operations, centroid, trajectory/plane intersections, mesh section, nearest/KD-tree, distances, areas/volumes, barycentric, fit/rigid transform, Delaunay/Voronoi, orientations, smoothing/remesh/boolean/volume slice |
sparse | construct, matvec, solve, least_squares, eigen_subset |
equations | solve, multiple_start, continuation |
variational | action, functional, legendre, first_variation, euler_lagrange_residual, constrained |
statistics | confidence_interval, robust_location, robust_regression, covariance_diagnostics, partial_correlation, multiple_testing, bootstrap_ci, permutation_test, logistic_regression, glm, survival_kaplan_meier, kmeans |
ode | solve |
constrained | solve |
8. Regole comuni
- Molte curve operations richiedono X strettamente crescente.
- Nonfinite values richiedono policy esplicita dove supportata.
- Le unità non vengono magicamente convertite dai raw mathematical calls.
- Dataset-aware calls usano Dataset ID + Variable ID e possono applicare Selection pubbliche.
- Weights zero vengono esclusi; semantics relative/inverse-variance sono distinte nei fitting che le supportano.
9. Discovery
Tcl:
::GX::NumericAPIv1::listOperations
::GX::NumericAPIv1::operationMetadata <id>
Octave:
gxtxt_numeric_v1('list_operations')
gxtxt_numeric_v1('operation_metadata', id)10. Script Bridge
Nella dev-031 la binding Script Bridge espone i dispatcher principali array_*, foundation, linear_matrix, statistiche Dataset-aware selezionate e le operazioni callback di root/minimize/nonlinear fit. Il catalogo pubblico completo è più ampio del solo elenco top-level del Bridge.