Eric Iverson writes: > I am not familiar with HDF5. For big text type files I would use J64 and > memory map the big file to a noun. That is a start.
Lettow, Kenneth writes: > I have not worked with HDF5 files, but you should take a look at using > memory mapped files in J. > http://www.jsoftware.com/jwiki/Studio/Mapped%20Files I had a look at this and it looks quite simple and powerful - as long as you stay in the J universe. I also suspect that portability of those files between machines is safe to assume only for text-based versions. So assume you have 60 GB of binary data you need to work with and archive for ten years - what do you do? That's not an academic question but my very real situation. Right now I keep such data either in netCDF or in HDF5 files. Both are platform-neutral and stable binary file formats the let me store arrays of any data type. My 60 GB are single-precision floats, for example. I used netCDF in the past, but I am currently transitioning to HDF5 because of better performance and more flexible storage options. Konrad. ---------------------------------------------------------------------- For information about J forums see http://www.jsoftware.com/forums.htm
