Add external group (generic I/O of arrays beyond the trexio format)

Makefile.am

@@ -37,7 +37,7 @@ CLEANFILES = trexio.mod
 if HAVE_FORTRAN
 BUILT_SOURCES = trexio.mod
 else
-BUILT_SOURCES = 
+BUILT_SOURCES =
 endif
 EXTRA_DIST = .git_hash
 
@@ -116,6 +116,7 @@ TESTS_C += \
   tests/io_dset_float_hdf5 \
   tests/io_dset_int_hdf5 \
   tests/io_dset_sparse_hdf5 \
+  tests/io_dset_external_hdf5 \
   tests/io_determinant_hdf5 \
   tests/io_jastrow_hdf5 \
   tests/io_safe_dset_float_hdf5 \
@@ -303,7 +304,7 @@ DEB_FILES = \
   helpers-debian/libtrexio0.install \
   helpers-debian/libtrexio-dev.install \
   helpers-debian/source        \
-  helpers-debian/README.source 
+  helpers-debian/README.source
 
 debian_from_dist: $(DEB_FILES) $(SOURCES) $(trexio_h)
 	cp ../trexio-$(PACKAGE_VERSION).tar.gz ../libtrexio_$(PACKAGE_VERSION).orig.tar.gz

python/test/benzene_data.py

@@ -47,3 +47,6 @@
 
 orb_up_test = [0, 65, 128, 129]
 orb_dn_test = [1, 64, 128, 129]
+
+external_2Dfloat_name = "test external float matrix"
+external_1Dint32_name = "test external int32 vector"

python/test/test_api.py

@@ -149,6 +149,19 @@ def test_array_2D(self):
         assert trexio.has_nucleus_coord(self.test_file)
 
 
+    def test_external_array(self):
+        """Write external arrays."""
+        self.open()
+
+        assert not trexio.has_external_array(self.test_file, external_2Dfloat_name)
+        trexio.write_external_array(self.test_file, nucleus_coord, external_2Dfloat_name)
+        assert trexio.has_external_array(self.test_file, external_2Dfloat_name)
+
+        assert not trexio.has_external_array(self.test_file, external_1Dint32_name)
+        trexio.write_external_array(self.test_file, np.array(nucleus_charge,dtype=np.int32), external_1Dint32_name)
+        assert trexio.has_external_array(self.test_file, external_1Dint32_name)
+
+
     def test_indices(self):
         """Write array of indices."""
         self.open()
@@ -252,6 +265,21 @@ def test_read_array_2D(self):
         np.testing.assert_array_almost_equal(coords_np, np.array(nucleus_coord).reshape(nucleus_num,3), decimal=8)
 
 
+    def test_read_external_array(self):
+        """Read external arrays."""
+        self.open(mode='r')
+        # read nuclear coordinates without providing optional argument dim
+        coords_external_np = trexio.read_external_array(self.test_file, name=external_2Dfloat_name, dtype="float64", size=nucleus_num*3)
+        assert coords_external_np.dtype is np.dtype(np.float64)
+        assert coords_external_np.size == nucleus_num * 3
+        np.testing.assert_array_almost_equal(coords_external_np.reshape(nucleus_num,3), np.array(nucleus_coord).reshape(nucleus_num,3), decimal=8)
+
+        charge_external_np = trexio.read_external_array(self.test_file, name=external_1Dint32_name, dtype="int32", size=nucleus_num)
+        assert charge_external_np.dtype is np.dtype(np.int32)
+        assert charge_external_np.size == nucleus_num
+        np.testing.assert_array_almost_equal(charge_external_np, np.array(nucleus_charge, dtype=np.int32))
+
+
     def test_read_errors(self):
         """Test some reading errors."""
         self.open(mode='r')

src/pytrexio.i

@@ -108,6 +108,9 @@ import_array();
 /* For some reasons SWIG does not apply the proper bitfield_t typemap, so one has to manually specify int64_t* ARGOUT_ARRAY1 below */
 %apply (int64_t* ARGOUT_ARRAY1, int32_t DIM1) {(bitfield_t* const bit_list, const int32_t N_int)};
 
+/* For passing dimensions of external arrays fron Python front to C back */
+%apply (uint64_t* IN_ARRAY1, int32_t DIM1) {(const uint64_t* dims_in, const int32_t dims_dim_in)};
+
 /* This tells SWIG to treat char ** dset_in pattern as a special case
    Enables access to trexio_[...]_write_dset_str set of functions directly, i.e.
    by converting input list of strings from Python into char ** of C