The library implements most of the functionality needed for reading and writing HDF5 files. It is actively maintained and the goal is to support all the features of HDF5.

Also provided is a fast way of working with large arrays of records, much faster than OCaml arrays of records. See examples/bench/bench_struct.ml`.

Building

./configure
make

lib/caml - simplified HDF5 wrapper

Store an array

open Hdf5_caml

let () =
  let a = [| 0.; 1.; 2. |] in

  let output = H5.create_trunc "file.h5" in
  H5.write_float_array output "a" a;
  H5.close output;

  let input = H5.open_rdonly "file.h5" in
  let b = H5.read_float_array "a" in
  H5.close input;

  assert (a = b)

Store a table

open Hdf5_caml

module Temperature = struct
  [%%h5struct
    time      "Time"      Int;
    latitude  "Latitude"  Float;
    longitude "Longitude" Float;
    temp      "Temp"      Float]
end

let () =
  let a = Record.Vector.create () in
  Record.(set (Vector.append a) ~time:10 ~latitude:45.2 ~longitude:0.2 ~temp:15.3);
  Record.(set (Vector.append a) ~time:11 ~latitude:45.2 ~longitude:0.2 ~temp:15.5);
  Record.(set (Vector.append a) ~time:12 ~latitude:45.3 ~longitude:0.5 ~temp:16.2);
  let a = Record.Vector.to_array a in

  let output = H5.create_trunc "file.h5" in
  Record.Array.make_table a output "Temperature";
  H5.close output

lib/raw - raw HDF5 wrapper

open Bigarray
open Hdf5_raw

let _FILE        = "SDS.h5"
let _DATASETNAME = "IntArray"
let _NX          = 5
let _NY          = 6

let () =
  let data = Array2.create int32 c_layout _NX _NY in
  for j = 0 to _NX - 1 do
    for i = 0 to _NY - 1 do
      data.{j, i} <- Int32.of_int (i + j)
    done
  done;
  let file = H5f.create _FILE [ H5f.Acc.TRUNC ] in
  let dataspace = H5s.create_simple [| _NX; _NY |] in
  let datatype = H5t.copy H5t.native_int in
  H5t.set_order datatype H5t.Order.LE;
  let dataset = H5d.create file _DATASETNAME datatype dataspace in
  H5d.write dataset H5t.native_int H5s.all H5s.all (genarray_of_array2 data);
  H5t.close datatype;
  H5d.close dataset;
  H5s.close dataspace;
  H5f.close file