Skip to content

perf(io): open the HDF5 file once per DataStub read #910

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
ehennestad merged 2 commits into from
Oct 6, 2026
Merged

perf(io): open the HDF5 file once per DataStub read #910

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
0b7dca3
perf(io): open the HDF5 file once per load_mat_style call
ehennestad Sep 30, 2026
d3c3a53
perf(io): read a contiguous range of a 1-D dataset as one hyperslab
ehennestad Oct 1, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
27 changes: 20 additions & 7 deletions +io/+backend/+hdf5/@HDF5LazyArray/load_mat_style.m
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -9,7 +9,6 @@

%% Select from Space
dataDimensions = obj.dims;
spaceId = obj.getSpace();
userSelection = varargin;

selectionErrorId = 'NWB:DataStub:Load:InvalidSelection';
Expand Down Expand Up @@ -41,8 +40,15 @@
data = obj.load_mat_style(1);
data = getEmptyRepresentation(data);
return
end

% The dataspace and the read use the same open dataset, so each call
% opens the file once.
fileId = H5F.open(obj.Filename);
datasetId = H5D.open(fileId, obj.DatasetPath);
spaceId = H5D.get_space(datasetId);

elseif isscalar(userSelection) && ~ischar(userSelection{1})
if isscalar(userSelection) && ~ischar(userSelection{1})
% linear index into the fast dimension.
orderedSelection = unique(userSelection{1});

Expand All @@ -66,11 +72,20 @@
dataDimensions = [dataDimensions, 1];
end

[points{:}] = ind2sub(dataDimensions, orderedSelection);
readSpaceId = H5S.copy(spaceId);
H5S.select_none(readSpaceId);
H5S.select_elements(readSpaceId, 'H5S_SELECT_SET', ...
cell2mat(flipud(points)) - 1);
isOneRun = orderedSelection(end) - orderedSelection(1) + 1 == numel(orderedSelection);
if isscalar(obj.dims) && isOneRun
% One contiguous run of a one-dimensional dataset is a single
% hyperslab, which HDF5 selects faster than one point per element.
% The selection can be of any integer class, and H5S takes doubles.
H5S.select_hyperslab(readSpaceId, 'H5S_SELECT_SET', ...
double(orderedSelection(1)) - 1, [], numel(orderedSelection), []);
else
[points{:}] = ind2sub(dataDimensions, orderedSelection);
H5S.select_elements(readSpaceId, 'H5S_SELECT_SET', ...
cell2mat(flipud(points)) - 1);
end
memorySpaceId = H5S.create_simple(length(selectionDimensions), ...
selectionDimensions, selectionDimensions);
else
Expand All @@ -81,8 +96,6 @@
H5S.close(spaceId);

%% Read Data
fileId = H5F.open(obj.Filename);
datasetId = H5D.open(fileId, obj.DatasetPath);
data = H5D.read(datasetId, 'H5ML_DEFAULT', memorySpaceId, readSpaceId, 'H5P_DEFAULT');

%% Retype Data
Expand Down
18 changes: 18 additions & 0 deletions +tests/+unit/+io/+backend/HDF5LazyArrayTest.m
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -70,6 +70,24 @@ function reorderedFirstSubscriptBeforeFoldedColon(testCase, Subscript)

testCase.verifyEqual(lazyArray.load_mat_style(Subscript, ':'), data(Subscript, :));
end

function loadMatStyleReadsOneDimensionalSelections(testCase)
% A selection that is one contiguous run is read as a hyperslab
% and any other selection as points. Both return the elements
% in the order they were asked for.
filename = "lazy-array-1d.h5";
data = (101:120)';
h5create(filename, "/data", numel(data));
h5write(filename, "/data", data);
lazyArray = io.backend.hdf5.HDF5LazyArray(filename, "/data");

selections = {5:9, (9:-1:5)', [3 3 4], 1:20, 20, uint64(5:9), [2 5 9]};
for iSelection = 1:numel(selections)
selection = selections{iSelection};
testCase.verifyEqual(lazyArray.load_mat_style(selection), data(selection), ...
sprintf('Unexpected data for selection %s.', mat2str(selection)));
end
end
end

methods (Access = private)
Expand Down
Loading