[newchem-cpp] hdf5 io refactor

[mabruzzo]

To be reviewed after #454 is merged

---

Refactor a bunch of hdf5 file-reading functionality.

• This addresses one of the items from Tabulated UVB/cooling needs refactoring #128.
• Overall, we are now a lot more "careful" (i.e. we verify that a lot of related properties are the same rather than just assume that they are the same)

The biggest change here is that I factored out the parsing of the attributes of the tabulated interpolation datasets. This is now handled by parse_GridTableProps and returns an object that summarizes the information

• the cost to this is that we are allocating some extra memory (this shouldn't be a huge deal since this is all done at initialization)
• a major benefit is that we could easily choose to do the following (we don't actually do any of this in this PR):
  • we could validate that the names of quantities along axes are are what we expect (e.g. "hdens", "redshift", "Temperature", "metal * log")
  • we could validate that different datasets have consistent attributes
  • I think that this validation is something we should be doing, but it will introduce slightly more overhead (and I didn't want to do it before talking it over)
• Realistically, the main reason I did this is so I can reuse the machinery when it comes time to parse the datafiles I made holding the multi-species dust information¹.

Footnotes

1. Honestly, if I had to do it all over again, I would do it slightly differently. This was a lot more effort than I expected. But I think it is a very useful improvement for us to have. ↩

[mabruzzo]

I would really like to make this change. The only drawback is that it will introduce slight overhead at startup. But, I think it's probably worth it

Suggested change

	// Ideally we would uncomment the following block of logic that verifies
	// that the Grid Table properties are identical to the cooling table
	// -> uncommenting the logic would be useful for enforcing consistency
	// in any new data files (all files currently shipped with Grackle
	// already satisfy this requirement)
	// -> the only downside to uncommenting the block is a little overhead
	//if (h5io::assert_has_consistent_GridTableProps(file_id, dset_name,
	// grid_props) != GR_SUCCESS){
	// h5io::drop_GridTableProps(&grid_props);
	// H5Fclose(file_id);
	// return GR_FAIL;
	//}
	// validate that Heating table has have identical GridTableProps
	if (h5io::assert_has_consistent_GridTableProps(file_id, dset_name,
	grid_props) != GR_SUCCESS){
	h5io::drop_GridTableProps(&grid_props);
	H5Fclose(file_id);
	return GR_FAIL;
	}

[mabruzzo]

Just like the preceding comment, I would really like to make this change

Suggested change

	// Ideally we would uncomment the following block of logic that verifies
	// that the Grid Table properties are identical to the cooling table
	// -> uncommenting the logic would be useful for enforcing consistency
	// in any new data files (all files currently shipped with Grackle
	// already satisfy this requirement)
	// -> the only downside to uncommenting the block is a little overhead
	//if (h5io::assert_has_consistent_GridTableProps(file_id, mmw_dset_name,
	// grid_props) != GR_SUCCESS){
	// h5io::drop_GridTableProps(&grid_props);
	// H5Fclose(file_id);
	// return GR_FAIL;
	//}
	// validate that MMW table has have identical GridTableProps
	if (h5io::assert_has_consistent_GridTableProps(file_id, mmw_dset_name,
	grid_props) != GR_SUCCESS){
	h5io::drop_GridTableProps(&grid_props);
	H5Fclose(file_id);
	return GR_FAIL;
	}