Check coefficient parameters in existing HDF5 file on extension

[The9Cat]

Summary

Check consistency of existing HDF5 before appending coefficients. This address issue #192.

Strategy

• On write: Coefs::ExtendH5Coefs() calls the overridden member function CheckH5Params() to check whether the current coefficient dimensions, scale parameters, and force type is consistent with those in the file. Throws an exception on mismatch.
• On read: the reader checks each stanza the expected dimensions based on the attributes and skips loading inconsistently ranked coefficient data

Notes

Compiles but only partly tested.

[Copilot]

Pull request overview

This pull request adds validation of HDF5 coefficient file parameters when extending existing files to prevent appending incompatible coefficients. The implementation introduces a CheckH5Params() method across all coefficient types and adds subsample covariance support for the Cube basis.

Key Changes

• Added CheckH5Params() method to validate coefficient dimensions, scale parameters, and force types before extending HDF5 files
• Introduced subsample covariance computation support for Cube basis with GPU implementation
• Refactored covariance storage into a new SubsampleCovariance class for better code reuse

Reviewed changes

Copilot reviewed 36 out of 36 changed files in this pull request and generated 5 comments.

Show a summary per file

File	Description
src/cudaCube.cu	Added subsample computation in GPU kernels and corrected index ordering from (k,j,i) to (i,j,k)
expui/Coefficients.cc	Implemented CheckH5Params() for all coefficient types with parameter validation
expui/Covariance.cc	New file implementing SubsampleCovariance class for covariance storage
src/OutSample.cc	New output class for writing subsample data to HDF5 files
pyEXP/CoefWrappers.cc	Added Python bindings for CheckH5Params() method
src/Cube.cc	Added subsample covariance computation for CPU implementation
src/SphericalBasis.cc	Added subsample support and parameter writing for spherical basis

Comments suppressed due to low confidence (1)

src/cudaCube.cu:1

• The calculation uses cubeNX in the last term but should use cubeNZ based on the index ordering (i,j,k). This will produce incorrect wave numbers when cubeNX != cubeNZ.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

Array indexing changed from 1-based (1 to 3) to 0-based (0 to 2), but the arrays L and R are likely still 1-indexed based on typical scientific code conventions. This will access incorrect indices or potentially go out of bounds.

Suggested change

	for (int i=0; i<3; i++) std::cerr << std::setw(12) << L[i];
	std::cerr << "] [";
	for (int i=0; i<3; i++) std::cerr << std::setw(12) << R[i];
	for (int i=1; i<=3; i++) std::cerr << std::setw(12) << L[i];
	std::cerr << "] [";
	for (int i=1; i<=3; i++) std::cerr << std::setw(12) << R[i];

[Copilot]

Multiple statements on one line reduce readability. Consider moving the stream insertion to a separate line.

[Copilot]

The shape validation should skip loading coefficients but continue reading the file. Consider whether this behavior is appropriate or if the entire read operation should be aborted on mismatch.

[Copilot]

Commented-out code should be removed rather than left in the codebase. If this functionality is still needed, it should be uncommented and maintained; otherwise, it can be removed since version control preserves history.

[Copilot]

The special case check for std::numeric_limits<int>::max() is unclear without documentation. Add a comment explaining when and why mstep would be set to this value.

Suggested change

	// Determine whether or not to compute a subsample
	// Determine whether or not to compute a subsample
	// mstep==0: normal subsampling interval.
	// mstep==std::numeric_limits<int>::max(): special sentinel value indicating that subsampling should always be considered (e.g., disables regular interval logic).

[The9Cat]

Will remake this PR with a clean branch against devel.