These are all constants used in HDF5. They are stored in an environment with locked bindings so that
they cannot be changed. An overview of all constants can be seen with h5const$overview, listing all of them.
Each constant can be accessed using $ and the name of the constant. See the examples below.
Holger Hoefling
There are also some flags that govern edge cases of conversion from HDF5 to R. This is related to how integers are being treated and
the issue of R not being able to natively represent 64bit integers and not at all being able to represent unsigned 64bit integers
(even using add-on packages).
The constants all start with the term H5TOR. There are currently possible values
Doesn't do any conversion. Every integer datatype with more than 32 bit is returned as 64bit integers. For unsigned 64bit integers, the conversion to signed 64bit integers is done by truncation
Under this setting, whenever a 64 bit integer would be returned, it is checked if it would also fit into a 32 bit integer without data loss and returned as such if possible
Under this setting, whenever a 64 bit integer would be returned, it is checked if it would also fit into a 64 bit floating point value without data loss and returned as such if possible
Combines H5TOR_CONV_INT64_INT_NOLOSS and H5TOR_CONV_INT64_FLOAT_NOLOSS and is set
as the default in the hdf5r.h5tor_default option.
Under this setting, whenever a 64 bit integer would be returned, it is coerced to a double
even if this results in a loss of precision. If a loss of precision occurs, a warning is issued. Please note that this also overrides
the use of H5TOR_CONV_UNIT64_NA. As loss of precision is already accepted, UINT64-values that are larger than LLONG_MAX will be represented
as their next possible floating point value.
When converting an unsigned 64bit integer, any values that don't fit into a signed 64bit integer are
set to NA. If this flag is not set, then the values will be truncated to LLONG_MAX, the largest 64bit signed integer.
Is both H5TOR_CONV_INT64_INT and H5TOR_CONV_UNIT64_FLOAT
h5const$overview
h5const$H5F_ACC_RDWR
h5const$H5F_ACC_DEFAULT
# Combining flags
bitwOr(h5const$H5TOR_CONV_UINT64_NA, h5const$H5TOR_CONV_INT64_INT_NOLOSS)
Run the code above in your browser using DataLab