Error `H5type` is set in `h5createAttribute()`

[lazappi]

If I try to create an attribute using h5createAttribute() with the H5type argument set I get an error:

path <- tempfile(fileext = ".h5")
file <- rhdf5::H5Fcreate(path)

rhdf5::h5write(
  1:10,
  name = "test",
  file = file
)
  
rhdf5::h5createAttribute(
  "test",
  attr = "test_attr",
  dims = 1,
  file = file,
  H5type = "H5T_STD_I32LE"
)
#> Error in rhdf5::h5createAttribute("test", attr = "test_attr", dims = 1, : object 'tid' not found

^{Created on 2025-06-26 with reprex v2.1.1}

I think it's because nothing is matched here

rhdf5/R/h5create.R

Lines 566 to 568 in fced3af

	if(grepl(pattern = "^[[:digit:]]+$", H5type)) {
	tid <- H5type
	}

Perhaps it could use .setDataType() like in h5createDataset()? Or alternatively, I've misunderstood something and this isn't supposed to work.

[grimbough]

Thanks for pointing this out.

Immediate workarounds:

• I think in this specific case you can actually use storage.mode = "integer" and it'll use that datatype.
• More generally you can used the following to use any of the inbuilt HDF5 datatypes

t1 <- H5Tcopy(dtype_id = "H5T_STD_I32LE")
rhdf5::h5createAttribute(
  "test", 
  attr = "test_attr",
  dims = 1,
  file = file,
  H5type = t1
)

However, you're right that what you tried seems much more natural, and it's odd for this to be inconsistent between h5createDataset and h5createAttrbute. Looking at the code there are some inconsistencies in behaviour e.g. how an R logical will be stored, so I can't just immediately switch to using .setDataType() without potentially breaking something for another package. I'll have to spend a bit of time looking at the implications, but it would be good for these two functions to work similarly long term.