Skip to main content

A binary timeseries storage format, where the time axis is given via an expression.

Project description

BinaryTimeseries

A binary timeseries storage format, where the time axis is given via an expression.

Scope

This is the specification for a really simple binary file format for storing a regularly-spaced sequence of single-channel measurement data in an efficiently writeable and readable format. The basic assumption is that the time axis t_i of a series of N measurements can be computed on the fly from the array indices:

for (int i=0; i<N; i++) {
	t_i = t_0 + i * Delta_t;
}

where t_0 is the (reference) timestamp of the first sample and Delta_t is the sampling interval.

The data values y_i are stored as raw values y_i_raw, optionally with an offset scalingOffset and a scaling factor scalingFactor:

for (int i=0; i<N; i++) {
	if (hasScaling) {
		y_i = y_i_raw;
	} else {
		y_i = scalingOffset + y_i_raw * scalingFactor;
	}
}

The maximum number of samples that can be stored inside this file format is limited by the maximum value of the (signed) int type, which is

Integer.MAX_VALUE == 2^31 - 1 == 2_147_483_647 \approx 2.1e9

This corresponds to a total duration of T_max = (2^31-1) * Delta_t.

In the case of raw double values as y_i_raw, the corresponding maximum file size that can occur is

(64 + Double.BYTES * 2_147_483_647) == (64+8*(2^31-1)) \approx 16 GB

where 64 bytes are reserved for the file header information.

Suppose an ADC samples at a frequency f = 1 MHz. Then, the sampling interval is Delta_t = 1/f = 1 µs and the maximum time series length that can be stored in one file in this file format is T_max \approx 2147 s.

The recommended file name extension for this file format is *.bts for Binary Time Series.

Fast subset reading

The main goal of this file format is to allow easy and fast reading of subsets of the whole time series data. Having an equally spaced time axis allows to compute the data indices inside a given time interval and using the definitions in Sec. 3 of the documentation (see below), the offsets in the file can be computed for seeking to the computed position in the file and reading only the required data from there on.

Documentation

The specification of this file format is available as a PDF in this repository: Binary Timeseries File Format Specification.

The LaTeX source code and the compiled PDF of this specification are also embedded (as resources) in the jar of the Java implementation on Maven Central.

Implementation

A Java implementation of this file format using a ByteBuffer as the file abstraction layer is available in this repository.

The latest release is available on Maven Central:

<dependency>
	<groupId>de.labathome</groupId>
	<artifactId>BinaryTimeseries</artifactId>
	<version>1.0.3</version>
</dependency>

A (currently read-only) Python implementation of this file format is available on PyPI:

pip install BinaryTimeseries

Useage

A starting point on how to use these classes is given in the following example files:

src/main/java/examples/Examples.java

BinaryTimeseries/Examples.py

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Files for BinaryTimeseries, version 1.0.3
Filename, size File type Python version Upload date Hashes
Filename, size BinaryTimeseries-1.0.3-py3-none-any.whl (13.4 kB) File type Wheel Python version py3 Upload date Hashes View hashes
Filename, size BinaryTimeseries-1.0.3.tar.gz (8.9 kB) File type Source Python version None Upload date Hashes View hashes

Supported by

Elastic Elastic Search Pingdom Pingdom Monitoring Google Google BigQuery Sentry Sentry Error logging AWS AWS Cloud computing DataDog DataDog Monitoring Fastly Fastly CDN DigiCert DigiCert EV certificate StatusPage StatusPage Status page