jiahao/GSL.jl

Fetching contributors…
Cannot retrieve contributors at this time
135 lines (109 sloc) 4.8 KB

GSL.jl

Julia interface to the GNU Scientific Library (GSL)

Jiahao Chen

Tested with libgsl 1.16

How to install

In Julia:

Pkg.add("GSL")

How to use

This package aims to provide wrappers to all documented functions and structs in the GSL manual. For GSL functions, the gsl_ prefix is not necessary in the name. More information is available below.

Example:

    using GSL
x = randn()
sf_hyperg_U(-1.0, -1.0, x) - (1 + x)
#Answer: 0.0

Translated examples from the GSL documentation are available in examples/.

Convenience methods provided

Roots of polynomials

    roots{T<:Real}(c::Vector{T}, realOnly::Bool) -> Vector{Complex128}

Finds the roots of the polynomial with real coefficients c [ 0 = \sum_{i=1}^{length(c)} c_i z^{i-1} ] The coefficients are returned in ascending order of the power If the degree of the polynomial is <= 3, then realOnly=true finds only the real roots.

Special functions

Hypergeometric functions

    hypergeom(a, b, x::Float64) -> Float64

Computes the appropriate hypergeometric pFq function, where p and q are the lengths of the input vectors a and b respectively. Singleton a and/or b may be specified as scalars, and length-0 a and/or b may be input as simply []. Supported values of (p, q) are (0, 0), (0, 1), (1, 1), (2, 0) and (2, 1). This only calls the floating-point versions of the GSL functions gsl_sf_hyperg_?F? (GSL manual, Section 7.24)

    hypergeom_e(a, b, x::Float64) -> (Float64, Float64)

An error-estimating version of hypergeom.

Test functions

    @sf_test(sf, args...)

Macro to help test equality of a function and its error-propagating variant. Requires Base.Test. Example:

    x = randn()
@eval @sf_test sf_dawson \$x

Current status

What is available

• Functions: all except the ones described below. Functions have the gsl_ prefix stripped.
• Most special functions: All except for the following categories:
• Some array-valued functions sf_*_array
• Available: sf_bessel_*_array, sf_gegenpoly_array.
• Not available: all others. The wrappers do not currently work.
• Not available: sf_*_e10_e that return the sf_result_e10 struct. (Currently returns bus error.)
• All documented gsl_* structs - These do NOT have the gsl_ prefix stripped.
• GSL_* constants: strip GSL_ and GSL_CONST_ prefixes
• GSL_* macros: Most available but untested.

What is not available

• GSL's BLAS and CBLAS wrappers blas_*, cblas_*. Use Julia's interface instead.
• Data I/O functions, such as *_fprintf and *_fscanf. Work in progress. Wrappers to these functions exist but most likely won't work
• Row and column views of matrices, matrix_row* and matrix_column* (Sec. 8.4.6)
• GSL_* macros:
• COMPLEX_AT, COMPLEX_FLOAT_AT, COMPLEX_LONG_DOUBLE_AT
• SF_RESULT_SET and others in gsl_sf_result.h
• SET_COMPLEX, SET_REAL, SET_IMAG, SET_COMPLEX_PACKED

Current tests

• Special functions
• Basic tests comparing equality of basic and error-propagating special functions
• Some identity tests for hypergeometric functions
• Available but untested:
• sf_bessel_sequence_Jnu_e, sf_bessel_Jnu
• sf_coulomb_CL_e, sf_coulomb_CL_e
• sf_coupling_6j, sf_elljac_e
• Mathieu Functions (Section 7.26). (Needs convenience function)
• Trigonometric Functions for Complex Arguments (Section 7.31)
• Conversion Functions (Section 7.31.4)
• Restriction Functions (Section 7.31.5)
• Trigonometric Functions With Error Estimates (Section 7.31.6)
• All other functions are untested
• All macros are untested

How you can help

The wrappers are automatically generated using util/makewrapper.py.

1. Test function wrappers for correctness.
2. Write convenience methods to further wrap the function calls with a Julia- friendly syntax.