A zero order Taylor coefficient is nan.on

[metab0t]

The following code errors:

#include <iostream>
#include <unordered_map>
#include <string>

#include "cppad/cppad.hpp"

using CppAD::AD;
using CppAD::ADFun;
using CppAD::Independent;
using CppAD::sparse_rc;
using CppAD::sparse_rcv;

typedef CPPAD_TESTVECTOR(double) d_vector;
typedef CPPAD_TESTVECTOR(AD<double>) ad_vector;
typedef CPPAD_TESTVECTOR(size_t) s_vector;

void debug_mvp()
{
	size_t n = 3;
	size_t m = 1;
	CPPAD_TESTVECTOR(AD<double>) ax(n), ay(m);
	Independent(ax);
	ay[0] = ax[0] / ax[1] - 1.0 / (ax[2] * ax[2]);
	ADFun<double> f;
	f.Dependent(ax, ay);

	// n by n identity matrix sparsity
	sparse_rc<s_vector> pattern_in;
	pattern_in.resize(n, n, n);
	for (size_t k = 0; k < n; k++)
		pattern_in.set(k, k, k);

	// sparsity for J(x)
	bool transpose = false;
	bool dependency = false;
	bool internal_bool = true;
	sparse_rc<s_vector> pattern_jac;
	f.for_jac_sparsity(pattern_in, transpose, dependency, internal_bool, pattern_jac);

	ADFun<AD<double>, double> af = f.base2ad();

	CPPAD_TESTVECTOR(AD<double>) axp(n);
	{
		Independent(axp);
		for (size_t j = 0; j < n; ++j)
			ax[j] = axp[j];
		// compute entire forward mode Jacobian
		sparse_rcv<s_vector, ad_vector> subset(pattern_jac);
		CppAD::sparse_jac_work work;
		std::string coloring = "cppad";
		size_t n_color = af.sparse_jac_rev(ax, subset, pattern_jac, coloring, work);
		ADFun<double> g;
		g.Dependent(axp, subset.val());
	}
}

int main(void)
{
	debug_mvp();
}

The error happens at the size_t n_color = af.sparse_jac_rev(ax, subset, pattern_jac, coloring, work); line as follows:

cppad-20240602 error from a known source:
yq = f.Forward(q, xq): a zero order Taylor coefficient is nan.
Corresponding independent variables vector was written to binary a file.
vector_size = 3
file_name = C:\Users\yangyue\AppData\Local\Temp\uv5o.0
index = 0

Error detected by false result for
    if( CppAD::isnan( yq[ (q+1) * index + 0 ] )
at line 265 in the file
    E:\codeworkspace\cpp_try\cppad_try\thirdparty\include\cppad/core/forward/forward.hpp
Assertion failed: false, file E:\codeworkspace\cpp_try\cppad_try\thirdparty\include\cppad/utility/error_handler.hpp, line 205

[metab0t]

The problem seems that AD<double> defaults to 0.0, and causes the evaluation function problematic.
Just assign correct values to axp like

CPPAD_TESTVECTOR(AD<double>) axp(n);
axp[0] = 1.0;
axp[1] = 2.0;
axp[2] = 3.0;

[bradbell]

During the recording of a function, the function is evaluated at the independent variable values. These values are unspecified if you do not set them before the call to Independent.

You can suppress the check for nan using the check_for_nan setting; the argument b on
https://cppad.readthedocs.io/latest/check_for_nan.html#b

[metab0t]

Yes, I understand the reasons now. Thanks!

[bradbell]

I added the script
https://github.com/coin-or/CppAD/blob/master/bin/run_source.sh
for running test cases like the one above. I had to change

#include "cppad/cppad.hpp"

to

#include <cppad/cppad.hpp>

to get it to compile using run_source.sh.