Hi Werner, On Jun 3, 2010, at 8:46 AM, Werner Benger wrote:
> Hi Quincey, > > On Thu, 03 Jun 2010 06:39:27 -0500, Quincey Koziol <[email protected]> > wrote: > >> Hi Mark, >> >> On Jun 2, 2010, at 9:36 PM, Mark Miller wrote: >> >>> Yes. But the procedure you'd use depends a bit on what kind of object >>> you are talking about (dataset or attribute) and how you chose to define >>> the object. Did you use H5T_C_S1 as the datatype for the string? Did you >>> maybe just define a dataset/attribute of type H5T_NATIVE_CHAR and then >>> an 1D dataspace of some size? I think those choices effect how you'd go >>> about obtaining length. >>> >>> A general algorithm might look like... >>> >>> 1. whether its a dataset or attribute, you need to open it first >>> (H5Dopen/H5Aopen) >>> 2. then get its datatype (H5Dget_type/H5Aget_type) >>> 3. then get the size of that datatype (H5Tget_size) >>> 4. Then get the size of the dataspace >>> 5. Use results from steps 3 and 4 to determine total size. >> >> Actually, I think that Ken is trying to determine the actual string >> length for an individual element of a dataset or attribute, and he's correct >> that there's no way to retrieve the length before reading it in. The method >> you describe above is correct for determining the size of the array an >> application will need to allocate for reading the strings, but the size >> returned from H5Tget_size() should be the same as sizeof(char *) (for >> variable-length strings - for variable-length sequences of another datatype, >> it should be sizeof(hvl_t)). The memory for the actual strings themselves >> will be allocated at read time (and the malloc/free routines used can be >> controlled with H5Pset_vlen_mem_manager). >> > Is the H5Dread() routing robust against H5Pset_vlen_mem_manager's allocate > function returning NULL? Yes, I'm confident that it is. Getting NULL back from the allocate function will cause the H5Dread() call to fail and return an error. It won't release the memory from a partially completed I/O, however. > If so, it might be possible to provide an allocation function > > typedef void *(*H5MM_allocate_t)(size_t size, void *alloc_info) ; > > that divides the size parameter by H5Tget_size( H5Dget_type() ) to get > the number of elements that would be allocated for a specific element > (if that one is sufficiently simple). > This function would also need to count which element within the dataspace > it is just reading, so needs to be one allocate-function per dataspace > to be in sync. Yes, I thought about suggesting this, but decided it was pretty hackish. :-) Quincey > But, would only work if the H5Dread() deals with null pointers by just > skipping the reading, but not exiting with an error. Didn't try this yet. > > Werner > > > >> Quincey >> >>> Mark >>> >>> On Wed, 2010-06-02 at 18:55, Ken Sullivan wrote: >>>> Quick question, is there a way to find the length of a variable length >>>> string without first reading it? >>>> >>>> Thanks, >>>> Ken >>> -- >>> Mark C. Miller, Lawrence Livermore National Laboratory >>> ================!!LLNL BUSINESS ONLY!!================ >>> [email protected] urgent: [email protected] >>> T:8-6 (925)-423-5901 M/W/Th:7-12,2-7 (530)-753-851 >>> >>> >>> _______________________________________________ >>> Hdf-forum is for HDF software users discussion. >>> [email protected] >>> http://mail.hdfgroup.org/mailman/listinfo/hdf-forum_hdfgroup.org >> >> >> _______________________________________________ >> Hdf-forum is for HDF software users discussion. >> [email protected] >> http://mail.hdfgroup.org/mailman/listinfo/hdf-forum_hdfgroup.org >> >> > > > -- > ___________________________________________________________________________ > Dr. Werner Benger Visualization Research > Laboratory for Creative Arts and Technology (LCAT) > Center for Computation & Technology at Louisiana State University (CCT/LSU) > 211 Johnston Hall, Baton Rouge, Louisiana 70803 > Tel.: +1 225 578 4809 Fax.: +1 225 578-5362 _______________________________________________ Hdf-forum is for HDF software users discussion. [email protected] http://mail.hdfgroup.org/mailman/listinfo/hdf-forum_hdfgroup.org
