From 52039ee51bf62ac91fd29e2cc24520f2129cde81 Mon Sep 17 00:00:00 2001
From: Thomas Lynn <32374143+lynntf@users.noreply.github.com>
Date: Mon, 1 Apr 2024 16:12:24 +0200
Subject: [PATCH] Add clarity to jacobian/hessianstructure methods

Original tutorial is not entirely clear on what the jacobian/hessian jacobian/hessianstructure methods should return when providing them as sparse matrices.
---
 docs/source/tutorial.rst | 13 ++++++++-----
 1 file changed, 8 insertions(+), 5 deletions(-)

diff --git a/docs/source/tutorial.rst b/docs/source/tutorial.rst
index c0a06729..fe0a8300 100644
--- a/docs/source/tutorial.rst
+++ b/docs/source/tutorial.rst
@@ -253,15 +253,18 @@ methods should return the non-zero values of the respective matrices as
 flattened arrays. The hessian should return a flattened lower triangular
 matrix.
 
-The Jacobian and Hessian can be dense or sparse. If sparse, you must also
-define:
+The Jacobian and Hessian can be dense or sparse. If sparse,
+:func:`cyipopt.Problem.jacobian` and :func:`cyipopt.Problem.hessian` methods
+should return only the non-zero values of the respective matrices and you must
+also define:
 
 - :func:`cyipopt.Problem.jacobianstructure`
 - :func:`cyipopt.Problem.hessianstructure`
 
-which should return a tuple of indices that indicate the location of the
-non-zero values of the Jacobian and Hessian matrices, respectively. If not
-defined then these matrices are assumed to be dense.
+which should return a tuple of indices (row indices, column indices) that
+indicate the location of the non-zero values of the Jacobian and Hessian
+matrices, respectively. If not defined then these matrices are assumed to be
+dense.
 
 The :func:`cyipopt.Problem.intermediate` method is called every Ipopt iteration
 algorithm and can be used to perform any needed computation at each iteration.