From b78625b6c37026b6d398618a18de44b169605d54 Mon Sep 17 00:00:00 2001
From: Steph Prince <40640337+stephprince@users.noreply.github.com>
Date: Wed, 2 Oct 2024 14:45:30 -0700
Subject: [PATCH] Fix scalar dataset with compound dtype for export (#1185)

* convert compound dtype to list on read

* revert dtype list checks in validator

* update CHANGELOG.md

---------

Co-authored-by: Ryan Ly <rly@lbl.gov>
---
 CHANGELOG.md                      | 9 +++------
 src/hdmf/backends/hdf5/h5tools.py | 4 ++++
 src/hdmf/validate/validator.py    | 4 ++--
 3 files changed, 9 insertions(+), 8 deletions(-)

diff --git a/CHANGELOG.md b/CHANGELOG.md
index b72ead1c5..888a3cb70 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -1,11 +1,6 @@
 # HDMF Changelog
 
-## HDMF 3.14.6 (Upcoming)
-
-### Bug fixes
-- Fixed mamba-related error in conda-based GitHub Actions. @rly [#1194](https://github.com/hdmf-dev/hdmf/pull/1194)
-
-## HDMF 3.14.5 (September 17, 2024)
+## HDMF 3.14.5 (Upcoming)
 
 ### Enhancements
 - Added support for overriding backend configurations of `h5py.Dataset` objects in `Container.set_data_io`. @pauladkisson [#1172](https://github.com/hdmf-dev/hdmf/pull/1172)
@@ -13,6 +8,8 @@
 ### Bug fixes
 - Fixed bug in writing of string arrays to an HDF5 file that were read from an HDF5 file that was introduced in 3.14.4. @rly @stephprince
   [#1189](https://github.com/hdmf-dev/hdmf/pull/1189)
+- Fixed export of scalar datasets with a compound data type. @stephprince [#1185](https://github.com/hdmf-dev/hdmf/pull/1185)
+- Fixed mamba-related error in conda-based GitHub Actions. @rly [#1194](https://github.com/hdmf-dev/hdmf/pull/1194)
 
 ## HDMF 3.14.4 (September 4, 2024)
 
diff --git a/src/hdmf/backends/hdf5/h5tools.py b/src/hdmf/backends/hdf5/h5tools.py
index da7f78a91..36aeb7c8f 100644
--- a/src/hdmf/backends/hdf5/h5tools.py
+++ b/src/hdmf/backends/hdf5/h5tools.py
@@ -700,6 +700,10 @@ def __read_dataset(self, h5obj, name=None):
                 kwargs['dtype'] = d.dtype
             elif h5obj.dtype.kind == 'V':  # scalar compound data type
                 kwargs['data'] = np.array(scalar, dtype=h5obj.dtype)
+                cpd_dt = h5obj.dtype
+                ref_cols = [check_dtype(ref=cpd_dt[i]) or check_dtype(vlen=cpd_dt[i]) for i in range(len(cpd_dt))]
+                d = BuilderH5TableDataset(h5obj, self, ref_cols)
+                kwargs['dtype'] = HDF5IO.__compound_dtype_to_list(h5obj.dtype, d.dtype)
             else:
                 kwargs["data"] = scalar
         else:
diff --git a/src/hdmf/validate/validator.py b/src/hdmf/validate/validator.py
index 2668da1ec..6ce211f96 100644
--- a/src/hdmf/validate/validator.py
+++ b/src/hdmf/validate/validator.py
@@ -147,7 +147,7 @@ def get_type(data, builder_dtype=None):
     # Case for h5py.Dataset and other I/O specific array types
     else:
         # Compound dtype
-        if builder_dtype and len(builder_dtype) > 1:
+        if builder_dtype and isinstance(builder_dtype, list):
             dtypes = []
             string_formats = []
             for i in range(len(builder_dtype)):
@@ -441,7 +441,7 @@ def validate(self, **kwargs):
             except EmptyArrayError:
                 # do not validate dtype of empty array. HDMF does not yet set dtype when writing a list/tuple
                 pass
-        if builder.dtype is not None and len(builder.dtype) > 1 and len(np.shape(builder.data)) == 0:
+        if isinstance(builder.dtype, list) and len(np.shape(builder.data)) == 0:
             shape = ()  # scalar compound dataset
         elif isinstance(builder.dtype, list):
             shape = (len(builder.data), )  # only 1D datasets with compound types are supported