HDF5Logger

This package creates a logger for storing individual frames of data over time. The frames can be scalars or arrays of any dimension, and the size must be fixed from sample to sample. Further, the total number of samples to log for each source must be known in advance. This keeps the logging very fast. It's useful, for instance, when one is running a simulation, some value in the sim needs to be logged every X seconds, and the end time of the simulation is known, so the total number of samples that will be needed is also known.

Create a logger. This actually creates and opens the HDF5 file.

using HDF5Logger
log = Log("my_log.h5")

Add a couple of streams. Suppose we'll log a 3-element gyro reading and a 3-element accelerometer signal each a total of 100 times.

num_samples = 100
example_gyro_reading = [0., 0., 0.]
example_accel_reading = [0., 0., 0.]

# Preallocate space for these signals.
add!(log, "/sensors/gyro",  example_gyro_reading, num_samples)
add!(log, "/sensors/accel", example_accel_reading, num_samples)

Log the first sample of each.

log!(log, "/sensors/gyro",  [1., 2., 3.])
log!(log, "/sensors/accel", [4., 5., 6.])
# We can now log to each of these signals 99 more times.

Always clean up.

close!(log);

Did that work?

using HDF5 # Use the regular HDF5 package to load what we logged.
h5open("my_log.h5", "r") do file
    gyro_data  = read(file, "/sensors/gyro")
    accel_data = read(file, "/sensors/accel")
    display(gyro_data[:,1])
    display(accel_data[:,1])
end

3-element Array{Float64,1}:
1.00
2.00
3.00
3-element Array{Float64,1}:
4.00
5.00
6.00

Yep!

The same process works with scalars, matrices, integers, etc.