Analysis for experiment result file with H5 extension

[BECATRUE]

After an submitted experiment is done, the result file is created in ARTIQ master directory.
In this discussion, I will let you know about this result file.

Introduction

Path

The result file is located in ${artiq-master}/{results}/{submission_date}/{submission_time}.
And its name is {RID}-{experiment_class_name}.py.

For example, ~/artiq/results/2023-08-09/19/000001432-TTLResolution.h5.

H5 format

The "H5" or "HDF5" format is for huge amounts of numerical data, and easily manipulating that data from NumPy.

How to access

In Python, we can access this format using h5py package.
If you installed ARTIQ package, this has already been installed with ARTIQ.

Also, if you use VS Code, some extensions e.g., H5Web may help you to show h5 file visually.

Key point

The "H5" use a directory scheme. So we can handle data like a directory structure easily.

For details, see also h5py docs.

Result file

h5 example.zip

The above is an example of the actual result file.

Structure

Its keys and values are as follows:

archive: <HDF5 group "/archive" (0 members)>
artiq_version: <HDF5 dataset "artiq_version": shape (), type "|O">
datasets: <HDF5 group "/datasets" (1 members)>
expid: <HDF5 dataset "expid": shape (), type "|O">
rid: <HDF5 dataset "rid": shape (), type "<i8">
run_time: <HDF5 dataset "run_time": shape (), type "<f8">
start_time: <HDF5 dataset "start_time": shape (), type "<f8">
test: <HDF5 dataset "test": shape (10,), type "<f4">

datasets

It contains all datasets created in this experiment.
Since the h5 format supports NumPy arrays, the dataset is also the same type.

expid

It is the same as expid in artiq_client.py.
It contains the class name, the arguments, the schedule options, etc.

rid

It is just the RID.

How to access the file using h5py

Fetch the data

import h5py

with h5py.File(file_path, 'r') as f:
        expid = f["expid"]  # data path like a directory path
        print(expid[()])

Add the data

import h5py

with h5py.File(file_path, 'w') as f:
        f.create_dataset("test", (10,))  # add a NumPy array with length 10

[jaeunkim]

Hi, thank you for the useful tutorial!
I think there's a tiny typo in the first tutorial ("Fetch the data"): it should be print(expid[()]), instead of print(rid[()])

[BECATRUE]

Oh, you're right!
Thanks for letting me know😃