diff --git a/internal/ceres/accelerate_sparse.cc b/internal/ceres/accelerate_sparse.cc
index b01414d..4331c29 100644
--- a/internal/ceres/accelerate_sparse.cc
+++ b/internal/ceres/accelerate_sparse.cc
@@ -114,12 +114,12 @@
   // Accelerate's columnStarts is a long*, not an int*.  These types might be
   // different (e.g. ARM on iOS) so always make a copy.
   column_starts_.resize(A->num_rows() + 1);  // +1 for final column length.
-  std::copy_n(A->rows(), column_starts_.size(), &column_starts_[0]);
+  std::copy_n(A->rows(), column_starts_.size(), column_starts_.data());
 
   ASSparseMatrix At;
   At.structure.rowCount = A->num_cols();
   At.structure.columnCount = A->num_rows();
-  At.structure.columnStarts = &column_starts_[0];
+  At.structure.columnStarts = column_starts_.data();
   At.structure.rowIndices = A->mutable_cols();
   At.structure.attributes.transpose = false;
   At.structure.attributes.triangle = SparseUpperTriangle;
diff --git a/internal/ceres/block_jacobian_writer.cc b/internal/ceres/block_jacobian_writer.cc
index 727d649..d7e1e95 100644
--- a/internal/ceres/block_jacobian_writer.cc
+++ b/internal/ceres/block_jacobian_writer.cc
@@ -94,7 +94,7 @@
   jacobian_layout_storage->resize(num_jacobian_blocks);
 
   int e_block_pos = 0;
-  int* jacobian_pos = &(*jacobian_layout_storage)[0];
+  int* jacobian_pos = jacobian_layout_storage->data();
   for (int i = 0; i < residual_blocks.size(); ++i) {
     const ResidualBlock* residual_block = residual_blocks[i];
     const int num_residuals = residual_block->NumResiduals();
@@ -144,7 +144,8 @@
 
   auto preparers = std::make_unique<BlockEvaluatePreparer[]>(num_threads);
   for (unsigned i = 0; i < num_threads; i++) {
-    preparers[i].Init(&jacobian_layout_[0], max_derivatives_per_residual_block);
+    preparers[i].Init(jacobian_layout_.data(),
+                      max_derivatives_per_residual_block);
   }
   return preparers;
 }
diff --git a/internal/ceres/compressed_row_sparse_matrix.h b/internal/ceres/compressed_row_sparse_matrix.h
index d164bef..d98f61f 100644
--- a/internal/ceres/compressed_row_sparse_matrix.h
+++ b/internal/ceres/compressed_row_sparse_matrix.h
@@ -110,8 +110,8 @@
   int num_rows() const final { return num_rows_; }
   int num_cols() const final { return num_cols_; }
   int num_nonzeros() const final { return rows_[num_rows_]; }
-  const double* values() const final { return &values_[0]; }
-  double* mutable_values() final { return &values_[0]; }
+  const double* values() const final { return values_.data(); }
+  double* mutable_values() final { return values_.data(); }
 
   // Delete the bottom delta_rows.
   // num_rows -= delta_rows
@@ -133,11 +133,11 @@
   void set_num_cols(const int num_cols) { num_cols_ = num_cols; }
 
   // Low level access methods that expose the structure of the matrix.
-  const int* cols() const { return &cols_[0]; }
-  int* mutable_cols() { return &cols_[0]; }
+  const int* cols() const { return cols_.data(); }
+  int* mutable_cols() { return cols_.data(); }
 
-  const int* rows() const { return &rows_[0]; }
-  int* mutable_rows() { return &rows_[0]; }
+  const int* rows() const { return rows_.data(); }
+  int* mutable_rows() { return rows_.data(); }
 
   StorageType storage_type() const { return storage_type_; }
   void set_storage_type(const StorageType storage_type) {
diff --git a/internal/ceres/covariance_impl.cc b/internal/ceres/covariance_impl.cc
index d6ce69b..bc4e75f 100644
--- a/internal/ceres/covariance_impl.cc
+++ b/internal/ceres/covariance_impl.cc
@@ -600,9 +600,9 @@
   cholmod_jacobian.ncol = num_cols;
   cholmod_jacobian.nzmax = num_nonzeros;
   cholmod_jacobian.nz = nullptr;
-  cholmod_jacobian.p = reinterpret_cast<void*>(&transpose_rows[0]);
-  cholmod_jacobian.i = reinterpret_cast<void*>(&transpose_cols[0]);
-  cholmod_jacobian.x = reinterpret_cast<void*>(&transpose_values[0]);
+  cholmod_jacobian.p = reinterpret_cast<void*>(transpose_rows.data());
+  cholmod_jacobian.i = reinterpret_cast<void*>(transpose_cols.data());
+  cholmod_jacobian.x = reinterpret_cast<void*>(transpose_values.data());
   cholmod_jacobian.z = nullptr;
   cholmod_jacobian.stype = 0;  // Matrix is not symmetric.
   cholmod_jacobian.itype = CHOLMOD_LONG;
diff --git a/internal/ceres/inner_product_computer.cc b/internal/ceres/inner_product_computer.cc
index a75cbe5..f2d5457 100644
--- a/internal/ceres/inner_product_computer.cc
+++ b/internal/ceres/inner_product_computer.cc
@@ -255,7 +255,7 @@
   int nnz = 0;
 
   // Process the first term.
-  const InnerProductComputer::ProductTerm* current = &product_terms[0];
+  const InnerProductComputer::ProductTerm* current = product_terms.data();
   FILL_CRSM_COL_BLOCK;
 
   // Process the rest of the terms.
diff --git a/internal/ceres/reorder_program.cc b/internal/ceres/reorder_program.cc
index eb37dc3..912fe3d 100644
--- a/internal/ceres/reorder_program.cc
+++ b/internal/ceres/reorder_program.cc
@@ -129,7 +129,7 @@
     if (parameter_block_ordering.NumGroups() <= 1) {
       // The user did not supply a useful ordering so just go ahead
       // and use AMD.
-      ss.Ordering(block_jacobian_transpose, OrderingType::AMD, &ordering[0]);
+      ss.Ordering(block_jacobian_transpose, OrderingType::AMD, ordering);
     } else {
       // The user supplied an ordering, so use CAMD.
       vector<int> constraints;
@@ -142,9 +142,9 @@
       // Renumber the entries of constraints to be contiguous integers
       // as CAMD requires that the group ids be in the range [0,
       // parameter_blocks.size() - 1].
-      MapValuesToContiguousRange(constraints.size(), &constraints[0]);
+      MapValuesToContiguousRange(constraints.size(), constraints.data());
       ss.ConstrainedApproximateMinimumDegreeOrdering(
-          block_jacobian_transpose, &constraints[0], ordering);
+          block_jacobian_transpose, constraints.data(), ordering);
     }
   } else if (linear_solver_ordering_type == ceres::NESDIS) {
     // If nested dissection is chosen as an ordering algorithm, then
@@ -152,7 +152,7 @@
     CHECK(SuiteSparse::IsNestedDissectionAvailable())
         << "Congratulations, you found a Ceres bug! "
         << "Please report this error to the developers.";
-    ss.Ordering(block_jacobian_transpose, OrderingType::NESDIS, &ordering[0]);
+    ss.Ordering(block_jacobian_transpose, OrderingType::NESDIS, ordering);
   } else {
     LOG(FATAL) << "Congratulations, you found a Ceres bug! "
                << "Please report this error to the developers.";
@@ -344,7 +344,7 @@
   // Renumber the entries of constraints to be contiguous integers as
   // CAMD requires that the group ids be in the range [0,
   // parameter_blocks.size() - 1].
-  MapValuesToContiguousRange(constraints.size(), &constraints[0]);
+  MapValuesToContiguousRange(constraints.size(), constraints.data());
 
   // Compute a block sparse presentation of J'.
   std::unique_ptr<TripletSparseMatrix> tsm_block_jacobian_transpose(
@@ -355,7 +355,7 @@
 
   vector<int> ordering(parameter_blocks.size(), 0);
   ss.ConstrainedApproximateMinimumDegreeOrdering(
-      block_jacobian_transpose, &constraints[0], &ordering[0]);
+      block_jacobian_transpose, constraints.data(), ordering.data());
   ss.Free(block_jacobian_transpose);
 
   const vector<ParameterBlock*> parameter_blocks_copy(parameter_blocks);
@@ -551,7 +551,7 @@
         *tsm_block_jacobian_transpose,
         parameter_blocks,
         parameter_block_ordering,
-        &ordering[0]);
+        ordering.data());
   } else if (sparse_linear_algebra_library_type == ACCELERATE_SPARSE) {
     // Accelerate does not provide a function to perform reordering without
     // performing a full symbolic factorisation.  As such, we have nothing
@@ -565,7 +565,7 @@
     OrderingForSparseNormalCholeskyUsingEigenSparse(
         linear_solver_ordering_type,
         *tsm_block_jacobian_transpose,
-        &ordering[0]);
+        ordering.data());
   }
 
   // Apply ordering.
diff --git a/internal/ceres/suitesparse.cc b/internal/ceres/suitesparse.cc
index 85fc5b2..3f091b9 100644
--- a/internal/ceres/suitesparse.cc
+++ b/internal/ceres/suitesparse.cc
@@ -189,7 +189,7 @@
   cc_.nmethods = 1;
   cc_.method[0].ordering = CHOLMOD_GIVEN;
   cholmod_factor* factor =
-      cholmod_analyze_p(A, const_cast<int*>(&ordering[0]), nullptr, 0, &cc_);
+      cholmod_analyze_p(A, const_cast<int*>(ordering.data()), nullptr, 0, &cc_);
 
   if (cc_.status != CHOLMOD_OK) {
     *message =
@@ -236,8 +236,8 @@
   block_matrix.nrow = num_row_blocks;
   block_matrix.ncol = num_col_blocks;
   block_matrix.nzmax = block_rows.size();
-  block_matrix.p = reinterpret_cast<void*>(&block_cols[0]);
-  block_matrix.i = reinterpret_cast<void*>(&block_rows[0]);
+  block_matrix.p = reinterpret_cast<void*>(block_cols.data());
+  block_matrix.i = reinterpret_cast<void*>(block_rows.data());
   block_matrix.x = nullptr;
   block_matrix.stype = A->stype;
   block_matrix.itype = CHOLMOD_INT;