The wt5 File Format

WrightTools stores data in binary wt5 files.

wt5 is a sub-format of HDF5.


wt5 files are hdf5 files with particular structure and attributes defined. wt5 objects may appear embedded within a larger hdf5 file or vise-versa, however this is untested. At the root of a wt5 file, a Collection or Data object is found. Collection and Data are hdf5 groups. A Collection may have children consisting of Collection and/or Data. A Data may have children consisting of Variable and/or Channel. Variable and Channel are hdf5 datasets.


The following metadata is handled within WrightTools and define the necessary attributes to be a wt5 file. It is recommended not to write over these attributes manually except at import time (e.g. from_<x> function).

name Collection Data Variable Channel description/notes
name yes yes yes yes Usually matches the last component of the path, except for root, /, which does not have a path with it’s name
class yes yes yes yes Identifies which kind of WrightTools object it is.
created yes yes     Timestamp of when the object was made, can be overwritten with source file creation time by from_<x> functions.
__version__ yes yes     wt5 version identifier
item_names yes yes     Ordered list of the children
variable_names   yes     Ordered list of all Variables
channel_names   yes     Ordered list of all Channels
axes   yes     Ordered list of axes expressions which define how a Data object is represented
constants   yes     Ordered list of expressions for values which are constant
kind   yes     Short description of what type of file it originated from, usually the instrument
source   yes     File path/url to the original file as read in
label     yes yes Identifier used to create more complex labels in Axes or Constants, which are used to plot
units     yes yes Units assigned to the dataset
min     yes yes Cached minimum value
max     yes yes Cached maximum value
argmin     yes yes Cached index of minimum value
argmax     yes yes Cached index of maximum value
signed       yes Boolean for treating channel as signed/unsigned


The HDF5 data model contains two primary objects: the group and the dataset. Groups are used to hierarchically organize content within the file. Each group is a container for datasets and other groups. Think of groups like folders in your computers file system. Every HDF5 file contains a top-level root group, signified by /.

Datasets are specialty containers for raw data values. Think of datasets like multidimensional arrays, similar to the numpy ndarray. Each dataset has a specific data type, such as integer, float, or character.

Groups and datasets can contain additional metadata. This metadata is stored in a key: value pair system called attrs, similar to a python dictionary.

Much more information can be found on the HDF5 tutorial.

WrightTools relies upon the h5py package, a Pythonic interface to HDF5.


wt5 is a binary format, so it cannot be interpreted with traditional text editors. Since wt5 is a sub-format of HDF5, WrightTools benefits from the ecosystem of HDF5 tools that already exists. This means that it is possible to import and interact with wt5 files without WrightTools, or even without python.


Export an HDF5 file to a human-readable ASCII file using h5dump.

See also HDF to Excel.


Use the official HDF5 Fortran Library.


HDF COMPASS, a simple tool for navigating and viewing data within HDF5 files (no editing functionality).

HDF VIEW, a visual tool for browsing and editing HDF5 files.


MATLAB offers built-in high-level HDF5 functions including h5disp, h5read, and h5readatt.

Python (without WrightTools)

We reccomend the amazing h5py package.


h5cli: bash-like interface to interacting with HDF5 files.

h5diff: compare two HDF5 files, reporting the differences.

h5ls: print information about one or more HDF5 files.

Complete list of official HDF5 tools


Version 1.0.0

Initial release of the format.

Version 1.0.1

Changes internal handling of strings. Bare strings are no longer required to call encode() before storing.

Version 1.0.2

Adds “constants” as a stored attribute in the attrs dictionary, a list of strings just like axes.