Interface Changes for the naginterfaces Python Package¶

This page lists backwards-incompatible API changes in naginterfaces up to and including version 30.3.0.0.

Interfaces Changes in version 28.2.0.0¶

opt - Optimization

`opt` - Optimization ¶

Solver naginterfaces.library.opt.handle_solve_ipopt() now requires the return vector from its Hessian callback to be updated in place.

Old definition:
```
def hess(x, idf, sigma, lamda, nnzh, inform):
    hx = foo
    return hx, inform
```
New definition:
```
def hess(x, idf, sigma, lamda, hx, inform):
    hx[:] = foo
    return inform
```
Also in this solver, length argument nnzu has been dropped and the (optional) multipliers u have been added on input for use in a future release. The size of the returned array u will match that supplied for u on input, or be zero if it was not present.

Old call:
```
handle_solve_ipopt(handle, x, nnzu, objfun=...)
```
New call:
```
u = np.empty(nnzu)
handle_solve_ipopt(handle, x, objfun=..., u=u)
```

Solver naginterfaces.library.opt.handle_solve_bxnl() now requires the vector rdx from its lsqgrd callback to be updated in place.

Old definition:

def lsqgrd(x, nres, rdx, inform):
    rdx = foo
    return rdx, inform

New definition:

def lsqgrd(x, nres, rdx, inform):
    rdx[:] = foo
    return inform

Interfaces Changes in version 27.3.0.0¶

`opt` - Optimization¶

Argument hx in callback lsqhes in naginterfaces.library.opt.handle_solve_bxnl() is now to be modified in place.

Old definition:

def lsqhes(x, lamda, inform):
    hx = something
    return hx, inform

New definition:

def lsqhes(x, lamda, hx, inform):
    hx[:, :] = something
    return inform

Interfaces Changes in version 27.1.0.0¶

`anova` - Analysis of Variance¶

Values for the maximum number of treatment means to be computed and the maximum number of terms in the analysis of variance table are now computed internally in naginterfaces.library.anova.factorial().

Old call:
```
factorial(y, lfac, nblock, inter, irdf, mterm, maxt)
```
New call:
```
factorial(y, lfac, nblock, inter, irdf)
```

`correg` - Correlation and Regression Analysis¶

Argument x in naginterfaces.library.correg.coeffs_kspearman_overwrite() and naginterfaces.library.correg.coeffs_kspearman_miss_case_overwrite() is now modified in place on the right-hand side of the function call.

Sample old call:
```
x, rr = coeffs_kspearman_overwrite(x, itype)
```
New call:
```
rr = coeffs_kspearman_overwrite(x, itype)
```

`dot` - Inner Products¶

Redundant size and step arguments in naginterfaces.library.dot.real_prec() and naginterfaces.library.dot.complex_prec() have been removed.

Sample old call:
```
real_prec(a[::2], b, len(a[::2]), 1, 1, c1, c2, sw)
```
New call:
```
real_prec(a[::2], b, c1, c2, sw)
```

`glopt` - Global Optimization¶

The return value of naginterfaces.library.glopt.bnd_mcs_option_check() is now a bool.

Old call:

if bnd_mcs_option_check('Some Option', comm) == 1:
   ...

New call:

if bnd_mcs_option_check('Some Option', comm):
   ...

`matop` - Matrix Operations, Including Inversion¶

Output array w in naginterfaces.library.matop.real_gen_sparse_lu() has been replaced by a single scalar increase which records the growth value of the computed factorization.

Old call:
```
fac = real_gen_sparse_lu(...)
increase = fac.w[0]
```
New call:
```
fac = real_gen_sparse_lu(...)
increase = fac.increase
```

`mip` - Operations Research¶

The interface for naginterfaces.library.mip.iqp_dense() has been improved.

Old call:
```
iqp_dense(a, bl, bu, cvec, h, intvar, istate, xs, strtgy, comm, qphess=...)
```
New call:
```
iqp_dense(bl, bu, h, intvar, xs, strtgy, comm, a=a, cvec=cvec, istate=istate, qphess=...)
```
in which optionality has been activated for a, cvec and istate and where the returned ax is now None when there were no linear constraints.

`mv` - Multivariate Methods¶

Optionality has been improved in naginterfaces.library.mv.multidimscal_ordinal().

Old call:

multidimscal_ordinal(typ, d, x, itera, iopt, ...)

New call:

multidimscal_ordinal(d, x, typ=typ, itera=itera, iopt=iopt, ...)

`omp` - OpenMP Utilities¶

The return value of naginterfaces.library.omp.get_nested(), naginterfaces.library.omp.in_parallel() and naginterfaces.library.omp.using_threaded_impl() is now a bool.

Sample old call:
```
if in_parallel() == 1:
   ...
```
New call:
```
if in_parallel():
   ...
```

`opt` - Optimization¶

Some callbacks in naginterfaces.library.opt.handle_solve_bounds_foas(), naginterfaces.library.opt.handle_solve_dfls() and have modified signatures with respect to the inform argument.

Sample old call:

objfun = lambda x, _nres: fx
handle_solve_dfls(...)

New call:

objfun = lambda x, _nres, inform: (fx, inform)
handle_solve_dfls(...)

Solver naginterfaces.library.opt.handle_solve_ipopt() also now requires return vectors from its gradient callbacks to be updated in place.

Old call:

objfun = lambda x: fx
objgrd = lambda x, _nnzfd: fdx
confun = lambda x, _ncnln: gx
congrd = lambda x, _nnzgd: gdx
handle_solve_ipopt(...)

New call:

objfun = lambda x, inform: (fx, inform)
def objgrd(x, fdx, inform):
    fdx[:] = [...]
    return inform
confun = lambda x, _ncnln, inform: (gx, inform)
def congrd(x, gdx, inform):
    gdx[:] = [...]
    return inform
handle_solve_ipopt(...)

Please refer to one of the specific documents to confirm a routine’s new interface.

Argument objfun is now optional in naginterfaces.library.opt.handle_solve_dfno().

Old call:

handle_solve_dfno(handle, objfun, x, ...)

New call:

handle_solve_dfno(handle, x, objfun=objfun, ...)

Argument maxcal is now optional in naginterfaces.library.opt.lsq_uncon_quasi_deriv_comp().

Old call:

lsq_uncon_quasi_deriv_comp(m, selct, lsqfun, maxcal, ..., lsqmon=None, ...)

New call:

lsq_uncon_quasi_deriv_comp(m, selct, lsqfun, ..., maxcal=maxcal, lsqmon=None, ...)

The interfaces for naginterfaces.library.opt.lp_solve(), naginterfaces.library.opt.lsq_lincon_solve() and naginterfaces.library.opt.qp_dense_solve() have been improved.

For naginterfaces.library.opt.lp_solve(); old call:
```
lp_solve(nclin, a, bl, bu, cvec, istate, x, comm, io_manager=None)
```
New call:
```
lp_solve(bl, bu, x, comm, a=a, cvec=cvec, istate=istate, io_manager=None)
```
in which optionality has been activated for a, cvec and istate, nclin is now inferred from a, and where the returned ax is now None when there were no linear constraints.

For naginterfaces.library.opt.lsq_lincon_solve(); old call:
```
lsq_lincon_solve(c, bl, bu, cvec, istate, kx, x, a, b, comm, io_manager=None)
```
New call:
```
lsq_lincon_solve(bl, bu, x, comm, c=c, cvec=cvec, istate=istate, kx=kx, a=a, b=b, io_manager=None)
```
in which optionality has been activated for c, cvec, istate, kx, a and b, and where the returned b is now None is some situations.

For naginterfaces.library.opt.qp_dense_solve(); old call:
```
qp_dense_solve(nclin, a, bl, bu, cvec, h, istate, x, comm, qphess=None, data=None, io_manager=None)
```
New call:
```
qp_dense_solve(bl, bu, h, x, comm, a=a, cvec=cvec, qphess=None, istate=istate, data=None, io_manager=None)
```
in which optionality has been activated for a, cvec and istate, nclin is now inferred from a, and where the returned ax is now None when there were no linear constraints.

`pde` - Partial Differential Equations¶

In the DAE solvers naginterfaces.library.pde.dim1_parab_dae_fd(), naginterfaces.library.pde.dim1_parab_dae_coll(), naginterfaces.library.pde.dim1_parab_dae_keller(), naginterfaces.library.pde.dim1_parab_convdiff_dae(), naginterfaces.library.pde.dim1_parab_convdiff_remesh(), naginterfaces.library.pde.dim1_parab_remesh_fd() and naginterfaces.library.pde.dim1_parab_remesh_keller() the keyword arguments for configuring the sizes of workspaces used when sparse matrix methods are requested have been renamed and now have more descriptive defaults (0).

Sample old call:
```
dim1_parab_dae_coll(..., lrsave=lrsave, lisave=lisave, ...)
```
New call:
```
dim1_parab_dae_coll(..., lrsave_estim=lrsave, lisave_estim=lisave, ...)
```

`quad` - Quadrature¶

Argument nagcal has been removed from naginterfaces.library.quad.dim1_gen_vec_multi_rcomm(). Its inclusion was erroneous and the argument had no effect.

Old call:
```
dim1_gen_vec_multi_rcomm(..., comm, nagcal)
```
New call:
```
dim1_gen_vec_multi_rcomm(..., comm)
```

`rand` - Random Number Generators¶

Reference vector r in naginterfaces.library.rand.multivar_normal(), naginterfaces.library.rand.multivar_students_t(), naginterfaces.library.rand.times_garch_asym1(), naginterfaces.library.rand.times_garch_asym2(), naginterfaces.library.rand.times_garch_exp(), naginterfaces.library.rand.times_garch_gjr() and naginterfaces.library.rand.times_smooth_exp() has been promoted to a communication structure. Consequently, the structure will now be initialized for you if you pass in an empty dict, and is modified in place.

Sample old calls:
```
mode = 0
r_n = np.empty((m*(m+1) + 1))
r_n, _ = multivar_normal(mode, n, df, xmu, c, r_n, statecomm)
...
```
or
```
mode = 0
r_t = np.empty((m*(m+1) + 2))
r_t, _ = multivar_students_t(mode, n, df, xmu, c, r_t, statecomm)
...
```
New calls:
```
mode = 0
comm_n = {}
multivar_normal(mode, n, df, xmu, c, comm_n, statecomm)
...

mode = 0
comm_t = {}
multivar_students_t(mode, n, df, xmu, c, comm_t, statecomm)
...
```

`sparse` - Large Scale Linear Systems¶

Many of the input arguments in naginterfaces.library.sparse.real_gen_precon_ilu() and naginterfaces.library.sparse.complex_gen_precon_ilu() have been made optional.

Sample old call:
```
real_gen_precon_ilu(
    nnz, a, irow, icol, lfill, dtol, milu, ipivp, ipivq,
    pstrat=pstrat,
)
```
New call:
```
real_gen_precon_ilu(
    nnz, a, irow, icol, n=?,
    lfill=lfill, dtol=dtol, milu=milu,
    ipivp=?, ipivq=?, pstrat=pstrat,
)
```
The new optionality of the pivot arrays ipivp and ipivq means that if they are omitted then n, the order of the matrix $A$ , must be supplied instead.
The weight array is now optional in naginterfaces.library.sparse.real_gen_basic_solver() and naginterfaces.library.sparse.complex_gen_basic_solver().

Sample old call:
```
real_gen_basic_solver(irevcm, u, v, wgt, comm)
```
New call:
```
real_gen_basic_solver(irevcm, u, v, comm, wgt=wgt)
```

Many of the input arguments in naginterfaces.library.sparse.real_gen_basic_setup() have been made optional.

Old call:

real_gen_basic_setup(
    method, precon, n, m, tol, maxitn, anorm, sigmax, monit,
    norm=norm, weight=weight, iterm=iterm,
)

New call:

real_gen_basic_setup(
    n, method=method, precon=precon, m=m, tol=tol, maxitm=maxitn,
    anorm=anorm, sigmax=sigmax, monit=monit,
    norm=norm, weight=weight, iterm=iterm,
)

Many of the input arguments in naginterfaces.library.sparse.complex_gen_basic_setup() have been made optional.

Additionally, lwork and lwreq have been dropped from this routine and also from naginterfaces.library.sparse.real_symm_basic_setup() and naginterfaces.library.sparse.complex_herm_basic_setup().

An old call sequence of the form

lwreq, comm = complex_gen_basic_setup(method, ..., lwork, ...)
if lwork < lwreq:
  _, comm = complex_gen_basic_setup(method, ..., lwreq, ...)

should be updated to

comm = complex_gen_basic_setup(
    n, method=method, precon=precon, m=m, tol=tol, maxitm=maxitn,
    anorm=anorm, sigmax=sigmax, monit=monit,
    norm=norm, weight=weight, iterm=iterm,
)

`sparseig` - Large Scale Eigenproblems¶

Argument mb is now required to be zero-sized when the standard eigenvalue problem is being solved in naginterfaces.library.sparseig.real_symm_band_solve(). (This is the default mode of operation.)

Old call:

comm = {}
# The default problem form:
real_symm_option('Standard', comm)
...
mb = np.empty((2*kl+`ku`+1, n))
real_symm_band_solve(..., mb, ...)

New call:

comm = {}
real_symm_option('Standard', comm)
...
mb = np.empty((0, n))
real_symm_band_solve(..., mb, ...)

`tsa` - Time Series Analysis¶

Reference vector r in naginterfaces.library.tsa.uni_smooth_exp() has been promoted to a communication structure. Consequently, the structure will now be initialized for you if you pass in an empty dict, and is modified in place.

Old call:

mode = 0
r = np.empty(13 + (0 if itype in (1, 2, 3) else p))
e = uni_smooth_exp(mode, itype, p, param, y, k, init, nf, r)
r = e.r
...

New call:

mode = 0
comm = {}
e = uni_smooth_exp(mode, itype, p, param, y, k, init, nf, comm)
...

Argument l has been dropped from naginterfaces.library.tsa.multi_kalman_sqrt_invar().

Old call:
```
multi_kalman_sqrt_invar(transf, l, a, b, ...)
```
New call:
```
multi_kalman_sqrt_invar(transf, a, b, ...)
```

Interfaces Changes in version 27.0.0.0¶

Cross-Chapter Changes¶

‘Diagnostic channel’ arguments have been removed.

The following table lists the affected routines and the name of the corresponding removed argument.

Routine name	Argument
`correg.glm_normal`	`nout`
`correg.glm_binomial`	`nout`
`correg.glm_poisson`	`nout`
`correg.glm_gamma`	`nout`
`correg.robustm`	`nout`
`correg.robustm_wts`	`nadv`
`correg.robustm_user`	`nadv`
`correg.robustm_corr_huber`	`nadv`
`correg.robustm_corr_user_deriv`	`nadv`
`correg.robustm_corr_user`	`nadv`
`contab.condl_logistic`	`nout`
`contab.binary`	`nadv`
`linsys.real_gen_sparse_lsqsol`	`nadv`
`matop.real_gen_sparse_lu`	`nadv`
`matop.real_gen_sparse_lu_reuse`	`nadv`
`mesh.dim2_gen_inc`	`nadv`
`mesh.dim2_gen_delaunay`	`nadv`
`mesh.dim2_gen_front`	`nadv`
`mesh.dim2_gen_boundary`	`nadv`
`mesh.dim2_smooth_bary`	`nadv`
`mesh.dim2_renumber`	`nadv`
`mesh.dim2_transform_affine`	`nadv`
`mesh.dim2_join`	`nadv`
`ode.ivp_stiff_exp_fulljac`	`nadv`
`ode.ivp_stiff_exp_bandjac`	`nadv`
`ode.ivp_stiff_exp_sparjac`	`nadv`
`ode.ivp_stiff_imp_fulljac`	`nadv`
`ode.ivp_stiff_imp_bandjac`	`nadv`
`ode.ivp_stiff_imp_sparjac`	`nadv`
`ode.ivp_stiff_exp_revcom`	`nadv`
`ode.ivp_stiff_imp_revcom`	`nadv`
`pde.dim2_ellip_mgrid`	`nadv`
`pde.dim1_parab_fd`	`nadv`
`pde.dim1_parab_coll`	`nadv`
`pde.dim1_parab_keller`	`nadv`
`pde.dim1_parab_convdiff`	`nadv`
`pde.dim1_parab_dae_fd`	`nadv`
`pde.dim1_parab_dae_coll`	`nadv`
`pde.dim1_parab_dae_keller`	`nadv`
`pde.dim1_parab_convdiff_dae`	`nadv`
`pde.dim1_parab_remesh_fd`	`nadv`
`pde.dim1_parab_remesh_keller`	`nadv`
`pde.dim1_parab_convdiff_remesh`	`nadv`
`pde.dim2_gen_order2_rectangle`	`nadv`
`pde.dim2_gen_order2_rectilinear`	`nadv`
`surviv.coxmodel`	`nout`
`tsa.uni_arima_estim`	`lp`
`tsa.uni_arima_estim_easy`	`nadv`
`tsa.uni_arima_resid`	`nadv`
`tsa.multi_inputmod_estim`	`nadv`
`tsa.multi_varma_estimate`	`lp`
`tsa.multi_varma_diag`	`nadv`

A previous call of the form

m = FileObjManager()
f = some_file_object
m.register((f,))
nag_solver(..., nadv=m.unit_from_fileobj(f), io_manager=m)

is equivalent in the new scheme to

m = FileObjManager()
f = some_file_object
m.register_to_advunit(f)
nag_solver(..., io_manager=m)

The previous form for output relying on the default association of the advisory unit

m = FileObjManager()
nag_solver(..., nadv=m.advunit, io_manager=m)

is now performed via

m = FileObjManager()
nag_solver(..., io_manager=m)

For calls in the old regime in which output from different solvers was directed to different file objects:

m = FileObjManager()
f1 = some_file_object
f2 = another_file_object
m.register((f1,f2))
nag_solver_a(..., nadv=m.unit_from_fileobj(f1), io_manager=m)
nag_solver_b(..., nadv=m.unit_from_fileobj(f2), io_manager=m)

use two I/O managers instead:

m1 = FileObjManager()
f1 = some_file_object
m1.register_to_advunit(f1)
m2 = FileObjManager()
f2 = another_file_object
m2.register_to_advunit(f2)
nag_solver_a(..., io_manager=m1)
nag_solver_b(..., io_manager=m2)

Some ‘result’ return values have been given more-meaningful names.

In the cases listed below this affects the name of the attribute in the function’s return tuple and/or in the return_data attribute on any warning raised by the function: for example, a call of the form

soln = quad.dim1_fin_well([...])
definite_integral = soln.result

in the new scheme must be converted to

soln = quad.dim1_fin_well([...])
definite_integral = soln.din

The following table lists the new names. In all cases the old name was result:

Routine name	New attr. name
`blas.ddiv`	`div`
`blas.zdiv`	`sdiv`
`nonpar.gofstat_anddar`	`gofs`
`ode.ivp_stiff_errest`	`errnrm`
`quad.dim1_fin_well`	`din`
`roots.lambertw_real`	`lambw`
`specfun.exp_complex`	`ez`
`specfun.erfc_complex`	`erfcz`
`specfun.tan`	`tx`
`specfun.sinh`	`shx`
`specfun.cosh`	`chx`
`specfun.integral_exp`	`e1x`
`specfun.gamma`	`gx`
`specfun.gamma_log_real`	`lgx`
`specfun.erfcx_real`	`ecx`
`specfun.bessel_k1_real`	`k1`
`specfun.ellipint_legendre_1`	`f`
`specfun.ellipint_legendre_3`	`p`
`specfun.ellipint_complete_1`	`k`
`specfun.hyperg_gauss_real`	`f21`
`stat.prob_gamma`	`p`
`stat.prob_studentized_range`	`p`
`stat.prob_dickey_fuller_unit`	`pint`
`stat.inv_cdf_students_t`	`x`
`stat.inv_cdf_chisq`	`x`
`stat.inv_cdf_f`	`x`
`stat.inv_cdf_beta`	`x`
`stat.inv_cdf_gamma`	`x`
`stat.inv_cdf_studentized_range`	`x`
`stat.inv_cdf_landau`	`invlan`
`stat.prob_f_noncentral`	`p`
`stat.prob_beta_noncentral`	`p`
`stat.prob_multi_normal`	`p`
`stat.prob_multi_students_t`	`p`
`tsa.uni_dickey_fuller_unit`	`ts`
`univar.outlier_peirce_2var`	`outl`

Some user-supplied function arguments have been renamed to f.

This affects calls to the routines listed below in which you pass the given function as a named keyword argument: for example, a call of the form

sum.invlaplace_crump(fun=my_f, [...])

must be converted to

sum.invlaplace_crump(f=my_f, [...])

The following table lists the old affected names. In all cases the new name is f.

Routine name	Old arg. name
`numdiff.fwd`	`fun`
`quad.dim1_indef`	`fun`
`quad.md_gauss`	`fun`
`quad.md_adapt`	`functn`
`quad.md_mcarlo`	`functn`
`quad.md_simplex`	`functn`
`quad.dim1_inf_exp_wt`	`fun`
`sum.invlaplace_crump`	`fun`

Improved identification of argument optionality has resulted in the following arguments becoming optional (and hence perhaps moving to a different position in the argument list). Please refer to the specific document for a given routine to see the new interface.

Routine name	Argument(s)
`correg.quantile_linreg`	`comm`
`correg.ridge`	`pec`
`correg.robustm_user`	`chi`
`eigen.real_gen_sparse_arnoldi`	`option`
`eigen.real_symm_sparse_arnoldi`	`option`
`fit.dim1_spline_deriv_vector`	`iwrk`
`glopt.nlp_pso`	`confun`
`glopt.nlp_multistart_sqp`	`a`, `start`, `confun`
`glopt.nlp_multistart_sqp_lsq`	`a`, `confun`
`lapackeig.dsbtrd`	`q`
`lapackeig.zhbtrd`	`q`
`lapackeig.dgees`	`select`
`lapackeig.dgeesx`	`select`
`lapackeig.zgees`	`select`
`lapackeig.zgeesx`	`select`
`lapackeig.zhseqr`	`z`
`lapackeig.zhsein`	`vl`, `vr`
`lapackeig.dtrevc`	`select`, `vl`, `vr`
`lapackeig.dtrsna`	`select`, `vl`, `vr`
`lapackeig.ztrevc`	`select`, `vl`, `vr`
`lapackeig.ztrsna`	`select`, `vl`, `vr`
`lapackeig.dgges`	`selctg`
`lapackeig.dggesx`	`selctg`
`lapackeig.dgges3`	`selctg`
`lapackeig.zgges`	`selctg`
`lapackeig.zggesx`	`selctg`
`lapackeig.zgges3`	`selctg`
`lapackeig.dtgevc`	`select`, `vl`, `vr`
`lapackeig.dtgsna`	`select`, `vl`, `vr`
`lapackeig.ztgevc`	`select`, `vl`, `vr`
`lapackeig.ztgsna`	`select`, `vl`, `vr`
`lapacklin.dgesvx`	`af`, `ipiv`, `r`, `c`
`mip.sqp`	`a`, `d`, `confun`
`mip.iqp_dense`	`qphess`
`mip.iqp_sparse`	`qphx`
`mesh.dim2_gen_boundary`	`fbnd`
`ode.ivp_rk_zero_simple`	`output`, `g`
`ode.ivp_adams_zero_simple`	`output`, `g`
`ode.ivp_bdf_zero_simple`	`pederv`, `output`, `g`
`ode.ivp_stiff_exp_fulljac`	`jac`
`ode.ivp_stiff_exp_bandjac`	`jac`
`ode.ivp_stiff_exp_sparjac`	`jac`
`ode.dae_dassl_gen`	`jac`
`ode.ivp_stiff_imp_fulljac`	`jac`
`ode.ivp_stiff_imp_bandjac`	`jac`
`ode.ivp_stiff_imp_sparjac`	`jac`
`ode.ivp_adams_roots`	`g`
`ode.bvp_fd_nonlin_gen`	`jacobf`, `jacobg`, `jaceps`, `jacgep`
`ode.bvp_shoot_genpar_algeq`	`eqn`, `constr`, `prsol`
`opt.miqp_mps_write`	`idxc`, `c`, `a`, `irowa`, `iccola`, `h`, `irowh`, `iccolh`, `intvar`
`opt.qp_dense_solve`	`qphess`
`opt.qpconvex1_sparse_solve`	`qphx`
`opt.qpconvex2_sparse_solve`	`qphx`, `c`
`opt.handle_set_quadobj`	`idxc`, `c`, `irowh`, `icolh`, `h`
`opt.handle_set_nlnobj`	`idxfd`
`opt.handle_set_linconstr`	`bl`, `bu`, `irowb`, `icolb`, `b`
`opt.handle_set_nlnconstr`	`bl`, `bu`, `irowgd`, `icolgd`
`opt.handle_set_nlnhess`	`irowh`, `icolh`
`opt.handle_set_nlnls`	`irowrd`, `icolrd`
`opt.handle_set_linmatineq`	`blksizea`
`opt.handle_solve_ipopt`	`objfun`, `objgrd`, `confun`, `congrd`, `hess`
`opt.handle_solve_pennon`	`u`, `uc`, `ua`
`opt.nlp1_solve`	`confun`, `istate`, `clamda`, `r`
`opt.nlp1_sparse_solve`	`confun`, `objfun`
`opt.lsq_gencon_deriv`	`a`, `confun`
`opt.nlp2_solve`	`confun`
`pde.dim1_parab_convdiff`	`pdedef`
`pde.dim1_parab_dae_fd`	`odedef`
`pde.dim1_parab_dae_coll`	`odedef`
`pde.dim1_parab_dae_keller`	`odedef`
`pde.dim1_parab_convdiff_dae`	`pdedef`, `odedef`
`pde.dim1_parab_remesh_fd`	`odedef`
`pde.dim1_parab_remesh_keller`	`odedef`
`pde.dim1_parab_convdiff_remesh`	`pdedef`, `odedef`
`quad.md_sphere`	`region`
`rand.sample_wgt`	`ipop`
`rand.int_binomial`	`comm`
`rand.int_geom`	`comm`
`rand.int_general`	`comm`
`rand.int_hypergeom`	`comm`
`rand.int_log`	`comm`
`rand.int_multinomial`	`comm`
`rand.int_negbin`	`comm`
`rand.int_poisson`	`comm`
`rand.bb_make_bridge_order`	`move`
`sparse.sym_rcm`	`mask`
`tsa.inhom_iema`	`sinit`
`tsa.inhom_iema_all`	`sinit`
`tsa.inhom_ma`	`sinit`
`tsa.uni_arima_estim`	`piv`
`tsa.kalman_unscented_state_revcom`	`lx`

`blas` - Linear Algebra Support Routines¶

Arguments m and k have been dropped from naginterfaces.library.blas.dgemm().

Old call:
```
dgemm(transa, transb, m, k, alpha, ...)
```
New call:
```
dgemm(transa, transb, alpha, ...)
```
Argument m has been dropped from naginterfaces.library.blas.dlange().

Old call:
```
dlange(norm, m, a)
```
New call:
```
dlange(norm, a)
```

`blgm` - Linear Model Specification¶

A new argument what has been added to naginterfaces.library.blgm.lm_submodel() for controlling the labels that are to be produced.

An old call, in which hform was produced by naginterfaces.library.blgm.lm_formula(), of the form

lm_submodel(hform, [...])

must be converted to

lm_submodel('S', hform, [...])

Additionally, some error exits have been renumbered as follows

Meaning	Old code	New code
Illegal `what`	N/A	11
`what` not valid for the supplied `hxdesc`	N/A	12
`hform` not initialized or corrupt	11	21
`hform` not a G22 handle generated by `lm_formula`	12	22
A variable name used when creating `hform` is not present in `hxdesc`	13	23
Model and design matrix not consistent (w.r.t. mean effect)	14	24
Model and design matrix not consistent (w.r.t. terms)	15	25
Model and design matrix not consistent (w.r.t. mean or main effect)	16	26
Model and design matrix not consistent (w.r.t. contrasts)	17	27
`hxdesc` not initialized or corrupt	21	31
`hxdesc` not a G22 handle generated by `lm_formula`	22	32
Constraint violated for `lisx` and $m_{x}$	41	61
Constraint violated for `lplab` and $p$	71	81
`lenlab` too small	81	91
Constraint violated for `lvinfo`	91	101
`lisx`, `lplab` or `lvinfo` too small	92	102

`interp` - Interpolation¶

Argument n has been dropped from naginterfaces.library.interp.dim1_aitken().

Old call:
```
dim1_aitken(a, b, n, x)
```
New call:
```
dim1_aitken(a, b, x)
```
Argument n has been dropped from naginterfaces.library.interp.dim1_everett().

Old call:
```
dim1_everett(n, p, a)
```
New call:
```
dim1_everett(p, a)
```
Argument lck has been dropped from naginterfaces.library.interp.dim1_spline().

Old call:
```
dim1_spline(x, y, lck)
```
New call:
```
dim1_spline(x, y)
```

`lapackeig` - Least Squares and Eigenvalue Problems (LAPACK)¶

Output argument work in naginterfaces.library.lapackeig.dgesvd() has been renamed to workb.

This affects calls using namedtuple referencing on the output.

Old call:
```
soln = dgesvd([...])
foo = soln.work
```
New call:
```
soln = dgesvd([...])
foo = soln.workb
```

`matop` - Matrix Operations, Including Inversion¶

Callback-termination flags iflag have been removed from naginterfaces.library.matop.complex_gen_matrix_cond_num(), naginterfaces.library.matop.complex_gen_matrix_cond_usd(), naginterfaces.library.matop.complex_gen_matrix_fun_num(), naginterfaces.library.matop.complex_gen_matrix_fun_usd(), naginterfaces.library.matop.complex_herm_matrix_fun(), naginterfaces.library.matop.real_gen_matrix_cond_num(), naginterfaces.library.matop.real_gen_matrix_cond_usd(), naginterfaces.library.matop.real_gen_matrix_fun_num(), naginterfaces.library.matop.real_gen_matrix_fun_usd() and naginterfaces.library.matop.real_symm_matrix_fun().

The previous functionality is superseded by the UserCallbackTerminate mechanism. See also the User-supplied Function Arguments subsection.

Old code of the form
```
f = lambda iflag, z: (iflag, F(z))
fa, iflag, imnorm = matop.real_gen_matrix_fun_num(a, f)
```
should be updated to become
```
f = lambda z: F(z)
fa, imnorm = matop.real_gen_matrix_fun_num(a, f)
```

`mip` - Operations Research¶

Argument comm has been dropped from the callback qphess in naginterfaces.library.mip.iqp_dense().

Old signature:
```
qphess(jthcol, h, x, comm, data)
```
New signature:
```
qphess(jthcol, h, x, data)
```

`ode` - Ordinary Differential Equations¶

Argument mmax has been added to the callback guess in naginterfaces.library.ode.bvp_coll_nlin_solve().

Old signature:
```
guess(x, m, data)
```
New signature:
```
guess(x, m, mmax, data)
```
Argument neq has been dropped from naginterfaces.library.ode.ivp_stiff_interp().

Old call:
```
ivp_stiff_interp(tsol, m, neq, comm)
```
New call:
```
ivp_stiff_interp(tsol, m, comm)
```
ysav has been subsumed into a communication structure in the monitr callback of naginterfaces.library.ode.ivp_stiff_exp_bandjac(), naginterfaces.library.ode.ivp_stiff_exp_fulljac(), naginterfaces.library.ode.ivp_stiff_exp_sparjac(), naginterfaces.library.ode.ivp_stiff_imp_bandjac(), naginterfaces.library.ode.ivp_stiff_imp_fulljac() and naginterfaces.library.ode.ivp_stiff_imp_sparjac().

Old signature:
```
monitr(t, hlast, hnext, y, ydot, ysav, r, acor, imon, hmin, hmax, nqu, comm, data=None),
```
New signature:
```
monitr(t, hlast, hnext, y, ydot, comm, r, acor, imon, hmin, hmax, nqu, data=None)
```
Likewise, the same array has been transformed in the interpolators naginterfaces.library.ode.ivp_stiff_c1_interp() and naginterfaces.library.ode.ivp_stiff_nat_interp().

Old calls:
```
ivp_stiff_c1_interp(xsol, m, ysav, acor, x, nqu, hu, h, comm)
ivp_stiff_nat_interp(xsol, m, ysav, neq, x, nqu, hu, h)
```
New calls:
```
ivp_stiff_c1_interp(xsol, m, comm, acor, x, nqu, hu, h)
ivp_stiff_nat_interp(xsol, m, comm, neq, x, nqu, hu, h):
```

Argument comm has been dropped from the callback monit in naginterfaces.library.ode.bvp_shoot_genpar_algeq().

Old signature:

monit(istate, iflag, ifail1, p, f, pnorm, pnorm1, eps, d, comm, data)

New signature:

monit(istate, iflag, ifail1, p, f, pnorm, pnorm1, eps, d, data)

`opt` - Optimization¶

Argument mon has been renamed to monit in naginterfaces.library.opt.handle_solve_dfls() and naginterfaces.library.opt.handle_solve_ipopt().

This affects calls using named keyword arguments.

For example, the old call
```
handle_solve_ipopt([...], mon=my_monit, [...])
```
must be converted to
```
handle_solve_ipopt([...], monit=my_monit, [...])
```

Argument cjac in naginterfaces.library.opt.nlp1_solve() has been made optional and hence it moves position in the argument list.

Old call:

nlp1_solve(a, bl, bu, confun, objfun, istate, cjac, clamda, r, x, comm, ...)

New call:

nlp1_solve(a, bl, bu, confun, objfun, istate, clamda, r, x, comm, cjac=cjac, ...)

Argument ncnln has been dropped from naginterfaces.library.opt.lsq_gencon_deriv().

Old call:

lsq_gencon_deriv(ncnln, a, bl, bu, y, confun, objfun, x, comm, ...)

New call:

lsq_gencon_deriv(a, bl, bu, y, confun, objfun, x, comm, ...)

Argument comm has been dropped from the callback qphess in naginterfaces.library.opt.qp_dense_solve().

Old signature:
```
qphess(jthcol, h, x, comm, data)
```
New signature:
```
qphess(jthcol, h, x, data)
```

`pde` - Partial Differential Equations¶

The DAE solvers naginterfaces.library.pde.dim1_parab_dae_fd(), naginterfaces.library.pde.dim1_parab_dae_coll(), naginterfaces.library.pde.dim1_parab_dae_keller(), naginterfaces.library.pde.dim1_parab_convdiff_dae(), naginterfaces.library.pde.dim1_parab_convdiff_remesh(), naginterfaces.library.pde.dim1_parab_remesh_fd() and naginterfaces.library.pde.dim1_parab_remesh_keller() have gained keyword arguments lrsave and lisave that can be used to configure the sizes of workspaces used.

Generally the routines are able to use initial estimates for the sizes which they compute internally. However, when sparse matrix methods are requested the initial estimates may be insufficient. The two new output arguments lrsave_min and lisave_min allow programmatic access to better lower bounds that can now be passed back to the routines in subsequent calls.
Argument comm has been dropped from the callback numflx in naginterfaces.library.pde.dim1_parab_convdiff(), naginterfaces.library.pde.dim1_parab_convdiff_dae() and naginterfaces.library.pde.dim1_parab_convdiff_remesh().

Old signature in dim1_parab_convdiff:
```
numflx(t, x, uleft, uright, ires, comm, data)
```
New signature in dim1_parab_convdiff:
```
numflx(t, x, uleft, uright, ires, data)
```
Old signature in dim1_parab_convdiff_dae and dim1_parab_convdiff_remesh:
```
numflx(t, x, v, uleft, uright, ires, comm, data)
```
New signature in dim1_parab_convdiff_dae and dim1_parab_convdiff_remesh:
```
numflx(t, x, v, uleft, uright, ires, comm, data)
```
Likewise, comm has been dropped from inidom in naginterfaces.library.pde.dim2_gen_order2_rectilinear() as well.

Old signature:
```
inidom(maxpts, ierr, comm, data)
```
New signature:
```
inidom(maxpts, ierr, data)
```

`rand` - Random Number Generators¶

The population vector ipop is now optional in naginterfaces.library.rand.sample().

To preserve prior operation
```
sample(ipop, m, statecomm)
```
it is now necessary to use
```
sample(m, statecomm, ipop=ipop)
```
Additional convenience now comes if the input population is $1, \dots, n$ , in which case ipop can be omitted and n supplied instead; that is, an existing call of the form
```
sample(list(range(1, n+1)), m, statecomm)
```
should become
```
sample(m, statecomm, n=n)
```
Note that it is not permitted to omit both ipop and n in the new call.

`roots` - Roots of One or More Transcendental Equations¶

A number of arguments have been made optional in naginterfaces.library.roots.contfn_brent(), naginterfaces.library.roots.contfn_brent_interval() and naginterfaces.library.roots.contfn_cntin().

Old calls:

contfn_brent(a, b, eps, f, ...)
contfn_brent_interval(x, h, eps, f, ...)
contfn_cntin(x, eps, f, nfmax, ...)

New calls:

contfn_brent(a, b, f, eps=eps, ...)
contfn_brent_interval(x, f, h=h, eps=eps, ...)
contfn_cntin(x, f, nfmax=nfmax, eps=eps, ...)

`sparse` - Large Scale Linear Systems¶

Argument check has been made optional in naginterfaces.library.sparse.complex_gen_matvec(), naginterfaces.library.sparse.complex_gen_precon_ilu_solve(), naginterfaces.library.sparse.complex_gen_precon_jacobi(), naginterfaces.library.sparse.complex_gen_precon_ssor_solve(), naginterfaces.library.sparse.complex_herm_matvec(), naginterfaces.library.sparse.complex_herm_precon_ilu_solve(), naginterfaces.library.sparse.complex_herm_precon_ssor_solve(), naginterfaces.library.sparse.real_gen_matvec(), naginterfaces.library.sparse.real_gen_precon_ilu_solve(), naginterfaces.library.sparse.real_gen_precon_jacobi(), naginterfaces.library.sparse.real_gen_precon_ssor_solve(), naginterfaces.library.sparse.real_symm_matvec(), naginterfaces.library.sparse.real_symm_precon_ichol_solve() and naginterfaces.library.sparse.real_symm_precon_ssor_solve().

Example old call:
```
real_gen_precon_ilu_solve(..., idiag, check, y)
```
Example new call:
```
real_gen_precon_ilu_solve(..., idiag, y, check=check)
```

`stat` - Simple Calculations on Statistical Data¶

Argument rcomm in naginterfaces.library.stat.summary_onevar() is now modified in place on the right-hand side of the function call.

A previous call sequence of the form

pn = 0
rcomm = [0.]*20
while ...:
    summary = summary_onevar(x, ..., pn=pn, rcomm=rcomm)
    pn = summary.pn
    rcomm = summary.comm

should be updated to be

pn = 0
rcomm = numpy.empty(20)
while ...:
    summary = summary_onevar(x, ..., pn=pn, rcomm=rcomm)
    pn = summary.pn

`sum` - Summation of Series¶

The signature for callback f in naginterfaces.library.sum.invlaplace_crump() has been simplified.

Old signature:
```
(fr, fi) = f(pr, pi, data=None)
```
New signature:
```
fp = f(p, data=None)
```
in which p = complex(pr, pi) and fp = complex(fr, fi).

`tsa` - Time Series Analysis¶

Argument npar has been removed from naginterfaces.library.tsa.uni_arima_prelim().

Old call:
```
uni_arima_prelim(mr, r, xv, npar)
```
New call:
```
uni_arima_prelim(mr, r, xv)
```
Argument hp in naginterfaces.library.tsa.uni_garch_asym1_estim() has been renamed to pht.

This affects calls using named keyword arguments.

Old call:
```
uni_garch_asym1_estim([...], hp=foo, [...])
```
New call:
```
uni_garch_asym1_estim([...], pht=foo, [...])
```
Argument hp in naginterfaces.library.tsa.uni_garch_asym2_estim() has been renamed to pht.

This affects calls using named keyword arguments.

Old call:
```
uni_garch_asym2_estim([...], hp=foo, [...])
```
New call:
```
uni_garch_asym2_estim([...], pht=foo, [...])
```
Argument hp in naginterfaces.library.tsa.uni_garch_gjr_estim() has been renamed to pht.

This affects calls using named keyword arguments.

Old call:
```
uni_garch_gjr_estim([...], hp=foo, [...])
```
New call:
```
uni_garch_gjr_estim([...], pht=foo, [...])
```
Argument hp in naginterfaces.library.tsa.uni_garch_exp_estim() has been renamed to pht.

This affects calls using named keyword arguments.

Old call:
```
uni_garch_exp_estim([...], hp=foo, [...])
```
New call:
```
uni_garch_exp_estim([...], pht=foo, [...])
```
Argument sparam is now returned as a rank-2 array from naginterfaces.library.tsa.cp_pelt() and naginterfaces.library.tsa.cp_binary() when ctype is not $5$ or $6$ .

`wav` - Wavelet Transforms¶

Argument lenc has been removed from naginterfaces.library.wav.dim1_sngl_fwd().

Old call:
```
dim1_sngl_fwd(x, lenc, comm)
```
New call:
```
dim1_sngl_fwd(x, comm)
```
Argument lenc has been removed from naginterfaces.library.wav.dim1_multi_fwd().

Old call:
```
dim1_multi_fwd(x, lenc, nwl, comm)
```
New call:
```
dim1_multi_fwd(x, nwl, comm)
```
Argument lenc has been removed from naginterfaces.library.wav.dim1_mxolap_fwd().

Old call:
```
dim1_mxolap_fwd(x, lenc, comm)
```
New call:
```
dim1_mxolap_fwd(x, comm)
```

Argument lenc has been removed from naginterfaces.library.wav.dim1_mxolap_multi_fwd().

Old call:

dim1_mxolap_multi_fwd(x, keepa, lenc, nwl, comm)

New call:

dim1_mxolap_multi_fwd(x, keepa, nwl, comm)

Argument lenc has been removed from naginterfaces.library.wav.dim2_multi_fwd().

Old call:
```
dim2_multi_fwd(a, lenc, nwl, comm)
```
New call:
```
dim2_multi_fwd(a, nwl, comm)
```
Argument lenc has been removed from naginterfaces.library.wav.dim3_sngl_fwd().

Old call:
```
dim3_sngl_fwd(a, lenc, comm)
```
New call:
```
dim3_sngl_fwd(a, comm)
```
Argument lenc has been removed from naginterfaces.library.wav.dim3_multi_fwd().

Old call:
```
dim3_multi_fwd(a, lenc, nwl, comm)
```
New call:
```
dim3_multi_fwd(a, nwl, comm)
```

`zeros` - Zeros of Polynomials¶

Previously-split complex data which appeared as floats in various forms has been coalesced back to complex in the following routines.

Old call

zsm, zlg = quadratic_complex(ar, ai, br, bi, cr, ci)

New call

z_small, z_large = quadratic_complex(a, b, c)

in which a = complex(ar, ai), and so on, and z_small = complex(zsm[0], zsm[1]), etc.

Old call

zsm, zlg = quadratic_real(a, b, c)

New call

z_small, z_large = quadratic_real(a, b, c)

in which z_small = complex(zsm[0], zsm[1]), and so on.

Old call

zeror, zeroi, errest = quartic_real(e, a, b, c, d)

New call

zero, errest = quartic_real(e, a, b, c, d)

in which zero = zeror + zeroi*1j.

Old call

zeror, zeroi, errest = quartic_complex(e, a, b, c, d)

New call

zero, errest = quartic_complex(e, a, b, c, d)

in which zero = zeror + zeroi*1j.

Naming in naginterfaces.library.zeros.poly_complex() has been unified with other routines in the submodule.

This affects calls using named keyword arguments.

Old call
```
poly_complex(ca=ca, scal=scal)
```
New call
```
poly_complex(a=ca, scal=scal)
```

Interface Changes for the naginterfaces Python Package¶

Interfaces Changes in version 28.2.0.0¶

opt - Optimization¶

Interfaces Changes in version 27.3.0.0¶

opt - Optimization¶

Interfaces Changes in version 27.1.0.0¶

anova - Analysis of Variance¶

correg - Correlation and Regression Analysis¶

dot - Inner Products¶

glopt - Global Optimization¶

matop - Matrix Operations, Including Inversion¶

mip - Operations Research¶

mv - Multivariate Methods¶

omp - OpenMP Utilities¶

opt - Optimization¶

pde - Partial Differential Equations¶

quad - Quadrature¶

rand - Random Number Generators¶

sparse - Large Scale Linear Systems¶

sparseig - Large Scale Eigenproblems¶

tsa - Time Series Analysis¶

Interfaces Changes in version 27.0.0.0¶

Cross-Chapter Changes¶

blas - Linear Algebra Support Routines¶

blgm - Linear Model Specification¶

interp - Interpolation¶

lapackeig - Least Squares and Eigenvalue Problems (LAPACK)¶

matop - Matrix Operations, Including Inversion¶

mip - Operations Research¶

ode - Ordinary Differential Equations¶

opt - Optimization¶

pde - Partial Differential Equations¶

rand - Random Number Generators¶

roots - Roots of One or More Transcendental Equations¶

sparse - Large Scale Linear Systems¶

stat - Simple Calculations on Statistical Data¶

sum - Summation of Series¶

tsa - Time Series Analysis¶

wav - Wavelet Transforms¶

zeros - Zeros of Polynomials¶

`opt` - Optimization ¶

`opt` - Optimization¶

`anova` - Analysis of Variance¶

`correg` - Correlation and Regression Analysis¶

`dot` - Inner Products¶

`glopt` - Global Optimization¶

`matop` - Matrix Operations, Including Inversion¶

`mip` - Operations Research¶

`mv` - Multivariate Methods¶

`omp` - OpenMP Utilities¶

`opt` - Optimization¶

`pde` - Partial Differential Equations¶

`quad` - Quadrature¶

`rand` - Random Number Generators¶

`sparse` - Large Scale Linear Systems¶

`sparseig` - Large Scale Eigenproblems¶

`tsa` - Time Series Analysis¶

`blas` - Linear Algebra Support Routines¶

`blgm` - Linear Model Specification¶

`interp` - Interpolation¶

`lapackeig` - Least Squares and Eigenvalue Problems (LAPACK)¶

`matop` - Matrix Operations, Including Inversion¶

`mip` - Operations Research¶

`ode` - Ordinary Differential Equations¶

`opt` - Optimization¶

`pde` - Partial Differential Equations¶

`rand` - Random Number Generators¶

`roots` - Roots of One or More Transcendental Equations¶

`sparse` - Large Scale Linear Systems¶

`stat` - Simple Calculations on Statistical Data¶

`sum` - Summation of Series¶

`tsa` - Time Series Analysis¶

`wav` - Wavelet Transforms¶

`zeros` - Zeros of Polynomials¶