VectorTemplate, IntVector and PermutationVector

[eivindjahren]

This removes VectorTemplate, IntVector and PermutationVector which is deprecated in version 6.3.

[Copilot]

Pull request overview

This PR removes the Python-side typed vector wrappers (notably IntVector and related templates) and continues migrating C++/pybind interfaces away from int_vector_type / bool_vector_type toward standard containers (std::vector, std::set) and NumPy arrays, with corresponding updates to tests and bindings.

Changes:

• Deleted the Python IntVector/PermutationVector implementation and its tests, and removed exports from resdata.util.util.
• Updated multiple C++ APIs and pybind bindings to use std::vector<int> / std::vector<bool> and NumPy arrays instead of ERT typed vectors.
• Updated Python and C++ tests to construct ACTNUM/index lists using built-in lists, std::vector, and returned containers.

Reviewed changes

Copilot reviewed 59 out of 59 changed files in this pull request and generated 4 comments.

Show a summary per file

File	Description
tests/util_tests/test_vectors.py	Removes IntVector-focused unit tests (file deleted).
tests/rd_tests/test_region.py	Stops using IntVector for actnum in region tests.
tests/rd_tests/test_rd_3dkw.py	Replaces IntVector ACTNUM setup with Python lists.
tests/rd_tests/test_layer.py	Replaces IntVector ACTNUM setup with Python lists.
tests/rd_tests/test_kw_function.py	Replaces IntVector ACTNUM setup with Python lists.
tests/rd_tests/test_grid.py	Replaces IntVector ACTNUM setup; adjusts type-error parametrization.
tests/rd_tests/test_grid_equinor.py	Replaces IntVector ACTNUM setup with Python lists.
python/resdata/util/util/vector_template.py	Removes Python vector template implementation (file deleted).
python/resdata/util/util/permutation_vector.py	Removes Python permutation vector wrapper (file deleted).
python/resdata/util/util/int_vector.py	Removes Python IntVector wrapper (file deleted).
python/resdata/util/util/init.py	Stops exporting IntVector/PermutationVector/VectorTemplate.
python/resdata/summary/rd_sum.py	Removes IntVector import from summary module.
python/resdata/grid/rd_region.py	Changes index-list return types to list[int].
python/resdata/grid/rd_grid.py	Changes exportACTNUM() to return a NumPy array type.
python/resdata/grid/faults/fault_block.pyi	Updates stubs to return list[int] instead of IntVector.
python/resdata/geometry/geo_region.py	Changes getActiveList() to return list[int].
lib/util/vector_template.cpp.in	Removes bool_vector usage from documentation snippet.
lib/util/type_vector_functions.cpp	Removes bool/int vector conversion helpers (file deleted).
lib/util/tests/ert_util_vector_test.cpp	Drops include of removed int_vector header.
lib/util/tests/ert_util_type_vector_test.cpp	Removes typed-vector C tests (file deleted).
lib/util/tests/ert_util_type_vector_functions.cpp	Removes type-vector-functions tests (file deleted).
lib/util/tests/ert_util_string_util.cpp	Removes string_util tests relying on int/bool vectors (file deleted).
lib/util/tests/ert_util_copy_file.cpp	Drops unused string_util include.
lib/util/string_util.cpp	Removes string_util implementation (file deleted).
lib/tests/test_rd_region.cpp	Migrates region list assertions to std::vector APIs.
lib/tests/test_rd_kw.cpp	Migrates indexed set usage to std::vector<int>.
lib/tests/test_layer.cpp	Drops include of removed int_vector header.
lib/tests/test_geometry.cpp	Migrates geo_region index list assertions to std::vector.
lib/resdata/tests/rd_region.cpp	Migrates region list checks to std::vector.
lib/resdata/tests/rd_nnc_vector.cpp	Drops include of removed int_vector header.
lib/resdata/tests/rd_nnc_test.cpp	Drops include of removed int_vector header.
lib/resdata/tests/rd_nnc_info_test.cpp	Drops include of removed int_vector header.
lib/resdata/tests/rd_kw_init.cpp	Drops include of removed bool_vector header.
lib/resdata/rd_sum.cpp	Replaces bool/int typed vectors with std::vector<bool> and std::vector<int>.
lib/resdata/rd_region.cpp	Replaces internal typed vectors with std::vector<int>; updates indexed kw operations.
lib/resdata/rd_region_pybind.cpp	Returns STL containers directly via pybind11 instead of IntVector wrappers.
lib/resdata/rd_kw.cpp	Updates indexed operations to accept const std::vector<int>&.
lib/resdata/rd_kw_pybind.cpp	Updates _fix_uninitialized binding to take a NumPy array.
lib/resdata/rd_grid.cpp	Drops include of removed int_vector header.
lib/resdata/rd_grid_pybind.cpp	Changes _init_actnum to return a NumPy array.
lib/resdata/rd_grav_common.cpp	Iterates region active indices via STL container.
lib/resdata/rd_file_view.cpp	Drops include of removed int_vector header.
lib/resdata/fault_block.cpp	Migrates internal lists to std::vector; regions to std::set.
lib/resdata/fault_block_pybind.cpp	Returns STL containers directly; drops IntVector wrapping.
lib/resdata/fault_block_layer.cpp	Drops include of removed int_vector header.
lib/resdata/cwrap_pybind.cpp	Removes IntVector helper and int_vector_type caster.
lib/private-include/detail/resdata/cwrap_pybind.hpp	Removes IntVector forward declaration.
lib/include/resdata/rd_region.hpp	Changes region list getters to return const std::vector<int>&.
lib/include/resdata/rd_kw.hpp	Updates indexed kw APIs to use std::vector<int> (but still exposes int_vector_type elsewhere).
lib/include/resdata/rd_file_view.hpp	Drops include of removed int_vector header.
lib/include/resdata/nnc_vector.hpp	Drops include of removed int_vector header.
lib/include/resdata/nnc_info.hpp	Drops include of removed int_vector header.
lib/include/resdata/fault_block.hpp	Migrates members to STL containers and updates signatures.
lib/include/ert/util/type_vector_functions.hpp	Removes type-vector-functions header (file deleted).
lib/include/ert/util/string_util.hpp	Removes string_util header (file deleted).
lib/include/ert/geometry/geo_region.hpp	Changes geo_region index list getter to return std::vector<int>.
lib/geometry/geo_region.cpp	Replaces cached int_vector index list with on-demand std::vector<int> construction.
lib/geometry/geo_region_pybind.cpp	Returns STL container directly from binding.
lib/CMakeLists.txt	Stops generating/building bool vector and string/type-vector utility code and tests.

Suppressed comments (1)

lib/include/resdata/rd_kw.hpp:12

• This header still depends on ert/util/int_vector.hpp and exposes int_vector_type in the public API (rd_kw_fread_indexed_data(..., const int_vector_type *index_map, ...)), while the newer indexed operations in the same header have been migrated to std::vector<int>. If the goal is to remove the typed vector utilities entirely, rd_kw_fread_indexed_data (and any callers/impl) should be migrated to std::vector<int>/std::span<const int> as well so the int_vector dependency can be dropped.

#include <ert/util/type_macros.hpp>
#include <ert/util/int_vector.hpp>
#include <ert/util/util.hpp>

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[Copilot]

Pull request overview

Copilot reviewed 67 out of 67 changed files in this pull request and generated 1 comment.

Suppressed comments (5)

lib/resdata/rd_grav_common.cpp:139

• Same issue as above: const auto copies the region’s active-index vector; keep it as a reference to avoid copying before iteration.

        const auto index_vector = rd_region_get_active_list(region);

lib/resdata/rd_region.cpp:1169

• rd_region_get_kw_index_list() returns a const std::vector<int>&, but this code copies it into target_index. That adds unnecessary allocation/copying on every call (potentially for large regions).

    const std::vector<int> target_index =
        rd_region_get_kw_index_list(rd_region, rd_kw, force_active);
    rd_kw_copy_indexed(rd_kw, target_index, src_kw);

lib/resdata/rd_grav_common.cpp:69

• rd_region_get_active_list() returns a const std::vector<int>&, but using const auto here drops the reference and copies the whole index list before iterating. That is unnecessary work and could be expensive for large regions.

This issue also appears on line 139 of the same file.

        const auto index_vector = rd_region_get_active_list(region);

lib/include/ert/geometry/geo_region.hpp:19

• Returning const std::vector<int> by value is counterproductive: it can inhibit moves (the returned rvalue is const) and adds no safety. Prefer returning std::vector<int> (non-const) by value.

void geo_region_reset(geo_region_type *region);
const std::vector<int> geo_region_get_index_list(geo_region_type *region);

lib/geometry/geo_region.cpp:139

• Same issue as the header: this function returns const std::vector<int> by value. Dropping the top-level const avoids inhibiting move semantics and matches typical C++ practice.

const std::vector<int> geo_region_get_index_list(geo_region_type *region) {

[Copilot]

Pull request overview

Copilot reviewed 48 out of 48 changed files in this pull request and generated 1 comment.

Suppressed comments (5)

lib/resdata/rd_grav_common.cpp:140

• Same as above: const auto index_vector = rd_region_get_active_list(region); copies the vector. Prefer iterating the returned reference directly to avoid per-call copies.

        const auto index_vector = rd_region_get_active_list(region);
        for (const auto index : index_vector) {

lib/resdata/rd_region.cpp:1169

• rd_region_get_kw_index_list() returns a const std::vector<int>&, but target_index is currently copied. This creates unnecessary work for potentially large regions; bind to a const reference instead.

    const std::vector<int> target_index =
        rd_region_get_kw_index_list(rd_region, rd_kw, force_active);
    rd_kw_copy_indexed(rd_kw, target_index, src_kw);

lib/include/resdata/rd_kw.hpp:12

• This header still includes and publicly exposes int_vector_type via rd_kw_fread_indexed_data(..., const int_vector_type *index_map, ...), which keeps the ert/util/int_vector.hpp dependency even though most indexed APIs were migrated to std::vector<int>. If the goal is to remove int_vector entirely, consider migrating rd_kw_fread_indexed_data (and its call sites) to take std::vector<int>/std::span<const int> and then drop the int_vector include here.

#include <vector>

#include <ert/util/type_macros.hpp>
#include <ert/util/int_vector.hpp>
#include <ert/util/util.hpp>

lib/resdata/rd_grav_common.cpp:70

• rd_region_get_active_list() now returns a const std::vector<int>&, but storing it in const auto index_vector = ... makes an unnecessary copy of (potentially large) index lists. Iterate the returned container by reference to avoid extra allocations/copies on every call.

This issue also appears on line 139 of the same file.

        const auto index_vector = rd_region_get_active_list(region);
        for (const auto index : index_vector) {

lib/include/ert/geometry/geo_region.hpp:18

• geo_region_get_index_list() now returns a std::vector<int> by value, which allocates and copies on every call. If this function is called repeatedly (e.g., from Python), this can be a noticeable regression compared to the previous cached list. Consider restoring caching in geo_region_struct and returning a const std::vector<int>& (or another non-owning view) to avoid repeated allocations.

geo_region_type *geo_region_alloc(const geo_pointset_type *pointset,
                                  bool preselect);
void geo_region_free(geo_region_type *region);
void geo_region_reset(geo_region_type *region);
const std::vector<int> geo_region_get_index_list(geo_region_type *region);

[Copilot]

Pull request overview

Copilot reviewed 48 out of 48 changed files in this pull request and generated no new comments.

Suppressed comments (1)

lib/resdata/rd_region.cpp:172

• The module-level API example above still documents these getters as returning const int * and assigns rd_region_get_global_list(...) to a pointer (lines 41–74). That example no longer compiles with the new const std::vector<int>& return type; update it to use a vector reference and .size()/indexing.

const std::vector<int> &rd_region_get_global_list(rd_region_type *region) {
    rd_region_assert_global_index_list(region);
    return region->global_index_list;

[Copilot]

Pull request overview

Copilot reviewed 48 out of 48 changed files in this pull request and generated no new comments.

Suppressed comments (2)

python/resdata/grid/rd_region.py:929

• active_size() and global_size() immediately above still compute len(self.get_*_list()). These bindings now convert the full C++ vectors into Python lists, so a size query has changed from O(1) to an O(n) allocation/copy; this is costly for million-cell grids and repeated size checks. Please expose dedicated size bindings (or otherwise query the C++ vector sizes directly) and use those from the size methods.
  lib/geometry/geo_region.cpp:143
• This removes the cached index list and now scans the entire point set on every call. In particular, Python GeoRegion.__len__ calls this getter, and __repr__ calls it twice, turning formerly cached/O(1) queries into repeated O(point-count) scans and list allocations. Preserve a cached std::vector<int> with invalidation, or add a direct selected-count path for size queries.

const std::vector<int> geo_region_get_index_list(geo_region_type *region) {
    std::vector<int> index_list;
    index_list.reserve(region->size());
    for (size_t i = 0; i < region->size(); i++)
        if (region->active_mask[i])

[Copilot]

Pull request overview

Copilot reviewed 48 out of 48 changed files in this pull request and generated no new comments.

[Copilot]

Pull request overview

Copilot reviewed 49 out of 49 changed files in this pull request and generated no new comments.

Suppressed comments (2)

python/resdata/grid/rd_region.py:929

• Returning a Python list here makes the existing __nonzero__ implementation at lines 137-139 copy every selected global index merely to determine whether the region is empty. This regresses bool(region) from a constant-size wrapper/length check to an O(n) allocation; update __nonzero__ to use the new global_size() helper instead.
  lib/geometry/geo_region.cpp:161
• Reserve only the number of selected entries. Reserving region->size() allocates space for every point even when the active list is empty or sparse, so getActiveList() can transiently consume O(total point count) memory for an O(selected count) result.

    index_list.reserve(region->size());

[Copilot]

Pull request overview

Copilot reviewed 49 out of 49 changed files in this pull request and generated 1 comment.

[ajaust]

What is the purpose of this PR? The subject mentions the removal of IntVector and bool_vector_type, but the body says

This removes VectorTemplate, IntVector and PermutationVector which is deprecated in version 6.3.

which does not mention bool_vector_type at all. I also still find the bool_vector_type:

$ git branch
  fix_build
  knowledge
  main
  python-trusted-publishing
* remove_bool_vector_type
  remove_double_vector
  remove_time_t_vector
  use-uv-with-lockfile

$ rg bool_vector_type
lib/util/type_vector_functions.cpp
7:int_vector_type *bool_vector_alloc_active_list(const bool_vector_type *mask) {
18:bool_vector_type *int_vector_alloc_mask(const int_vector_type *active_list) {
19:    bool_vector_type *mask = bool_vector_alloc(0, false);

lib/util/string_util.cpp
84:        bool_vector_type *mask = int_vector_alloc_mask(active_list);
116:                                    bool_vector_type *active_mask) {
131:bool_vector_type *string_util_alloc_active_mask(const char *range_string) {
132:    bool_vector_type *mask = bool_vector_alloc(0, false);

lib/util/vector_template.cpp.in
1185:   bool_vector_type   * v2;

lib/util/tests/ert_util_string_util.cpp
29:static void test2(const bool_vector_type *active_mask) {
46:    bool_vector_type *active_mask =

lib/util/tests/ert_util_type_vector_functions.cpp
14:        bool_vector_type *mask = int_vector_alloc_mask(index_list);
29:    bool_vector_type *mask = bool_vector_alloc(0, false);

lib/include/ert/util/type_vector_functions.hpp
11:int_vector_type *bool_vector_alloc_active_list(const bool_vector_type *mask);
12:bool_vector_type *int_vector_alloc_mask(const int_vector_type *active_list);

lib/include/ert/util/string_util.hpp
13:                                    bool_vector_type *active_mask);
14:bool_vector_type *string_util_alloc_active_mask(const char *range_string);

[ajaust]

I am unsure if we are at risk of an out-of-bounds access due to how add_cell works.

In add_cell we have the following:

    i_list.push_back(i);
    j_list.push_back(j);
    global_index_list.push_back(rd_grid_get_global_index3(grid, i, j, k));

If i_list.push_back(i) succeeds, but j_list.push_back(j) fails, j_list will hold one fewer element than i_list and we have an out-of-bounds access.

Do you think this is possible or will the application always (safely) abort if a memory allocation in push_back fails?

[eivindjahren]

You are right, I think the right course of action is to switch to a list of tuple. I added a commit that I will squash with the fault_block commit which addresses this.

[eivindjahren]

What is the purpose of this PR? The subject mentions the removal of IntVector and bool_vector_type, but the body says

This removes VectorTemplate, IntVector and PermutationVector which is deprecated in version 6.3.

I appologize for the confusion the bool_vector_type removal was moved to the "Remove unused utilities" PR in the stack, and I didn't notice the title was out of date.