`h5read()` crashes after `h5set_extent()` on a dataset of variable-length strings

[christiansteck]

Hi,

I want to use an extendable dataset of variable length strings.
So far, I was not able to make it work, as h5read() crashes after using h5set_extent().

Below is a minimal reproducible example.

rhdf5::h5createFile("foo.h5")

rhdf5::h5createDataset(
  file = "foo.h5",
  dataset = "charset_test",
  dims = 0,
  maxdim = rhdf5::H5Sunlimited(),
  chunk = 10,
  size = NULL,
  storage.mode = "character"
)

rhdf5::h5set_extent(
  file = "foo.h5",
  dataset = "charset_test",
  dims = 2
)

rhdf5::h5read(
  file = "foo.h5",
  name = "charset_test"
)

This results in a segfault with the following traceback:

 *** caught segfault ***
address (nil), cause 'memory not mapped'

Traceback:
 1: H5Dread(h5dataset = h5dataset, h5spaceFile = h5spaceFile, h5spaceMem = h5spaceMem,     compoundAsDataFrame = compoundAsDataFrame, drop = drop, ...)
 2: doTryCatch(return(expr), name, parentenv, handler)
 3: tryCatchOne(expr, names, parentenv, handlers[[1L]])
 4: tryCatchList(expr, classes, parentenv, handlers)
 5: tryCatch({    obj <- H5Dread(h5dataset = h5dataset, h5spaceFile = h5spaceFile,         h5spaceMem = h5spaceMem, compoundAsDataFrame = compoundAsDataFrame,         drop = drop, ...)}, error = function(e) {    err <- h5checkFilters(h5dataset)    on.exit(H5Dclose(h5dataset))    if (nchar(err) > 0)         stop(err, call. = FALSE)    else stop(e)})
 6: h5readDataset(h5dataset, index = index, start = start, stride = stride,     block = block, count = count, compoundAsDataFrame = compoundAsDataFrame,     drop = drop, ...)
 7: rhdf5::h5read(file = "foo.h5", name = "charset_test")

Am I using something incorrectly or do I need to supply other/more parameters to h5createDataset()?
Any help would be greatly appreciated.

---

Some notes and additional information:

• Writing data to the dataset after calling h5set_extent() but before calling h5read(), e.g. h5write(c("hello", "world"), "foo.h5", "charset_test") apparently fixes the issue and h5read() works fine, but I'd rather have the file itself use a suitable fill value for the new entries.
• Setting fillValue = "" in h5createDataset() gives Error in H5Tset_size(tid, size) : HDF5. Invalid arguments to routine. Bad value.
• Directly after h5set_extent(), the dataset contains NULL values and it seems to me that those are related to the crash:

$ h5dump -d charset_test foo.h5
HDF5 "foo.h5" {
DATASET "charset_test" {
   DATATYPE  H5T_STRING {
      STRSIZE H5T_VARIABLE;
      STRPAD H5T_STR_NULLPAD;
      CSET H5T_CSET_ASCII;
      CTYPE H5T_C_S1;
   }
   DATASPACE  SIMPLE { ( 2 ) / ( H5S_UNLIMITED ) }
   DATA {
   (0): NULL, NULL
   }
}
}

• Setting e.g. size = 10 and fillValue = "1234567890" in h5createDataset() works fine, but I need (or at least prefer) variable-length strings as the experiment results come in one after another.

> R.version.string
[1] "R version 4.2.2 (2022-10-31)"
> packageVersion("rhdf5")
[1] ‘2.42.1’
> rhdf5::H5get_libversion()
majnum minnum relnum
     1     10      7