Reading and writing Table and EventTable objects¶
Note
This document complements the upstream Astropy documentation on
reading/writing Table objects, please refer to
Reading and Writing Table Objects.
Astropy provides an excellent unified input/output
system for the Table object, and GWpy extends upon that to
include common gravitational-wave file types, as well as providing
event-specific input/output registrations for event data.
In the most general case you can read a table of data as follows:
EventTable from an ASCII file¶>>> from gwpy.table import EventTable
>>> table = EventTable.read('mydata.txt')
EventTable.read() will attempt to automatically identify the file
format based on the file extension and/or the contents of the file, however,
the format keyword argument can be used to manually identify the input
file-format.
The read() and write() methods take
different arguments and keywords based on the input/output file format,
see Built-in file formats for details on reading/writing for
each of the built-in formats.
Accessing Open Data event catalogues¶
The Gravitational-Wave Open Science Centre (GWOSC) publishes the Gravitational-Wave Transient Event Catalogues (GWTCs), allowing public access to tables of gravitational wave events and their parameters.
GWpy’s EventTable class comes with a
fetch_open_data() method that queries the GWOSC database
to download data for the relevant catalogue.
Simple queries¶
The simplest query just requires the catalogue name, and will return all parameters for all events in the catalogue:
EventTable.fetch_open_data() (output correct as of Dec 17, 2021)¶>>> from gwpy.table import EventTable
>>> events = EventTable.fetch_open_data("GWTC-2")
>>> print(events)
name final_mass_source_lower ... mass_2_source_upper mass_1_source_upper
...
------------------ ----------------------- ... ------------------- -------------------
GW190408_181802-v1 -2.8 ... 3.3 5.1
GW190412-v3 -3.8 ... 1.6 4.7
GW190413_052954-v1 -9.2 ... 7.3 12.6
GW190413_134308-v1 -11.4 ... 11.7 13.5
GW190421_213856-v1 -8.7 ... 8.0 10.4
GW190424_180648-v1 -10.1 ... 7.6 11.1
GW190425-v2 None ... 0.3 0.6
GW190426_152155-v1 None ... 0.8 3.9
GW190503_185404-v1 -7.7 ... 7.7 9.2
GW190512_180714-v1 -3.5 ... 3.6 5.3
GW190513_205428-v1 -5.8 ... 7.7 9.5
GW190514_065416-v1 -10.4 ... 9.3 14.7
GW190517_055101-v1 -8.9 ... 7.0 11.7
GW190519_153544-v1 -13.8 ... 11.0 10.7
GW190521-v3 -22.4 ... 22.7 28.7
GW190521_074359-v1 -4.4 ... 5.4 5.9
GW190527_092055-v1 -9.3 ... 10.5 16.4
GW190602_175927-v1 -14.9 ... 14.3 15.7
GW190620_030421-v1 -12.1 ... 12.2 16.0
GW190630_185205-v1 -4.6 ... 5.2 6.9
GW190701_203306-v1 -8.9 ... 8.7 11.8
GW190706_222641-v1 -13.5 ... 14.6 14.6
GW190707_093326-v1 -1.3 ... 1.4 3.3
GW190708_232457-v1 -1.8 ... 2.0 4.7
GW190719_215514-v1 -10.2 ... 9.0 18.0
GW190720_000836-v1 -2.2 ... 2.3 6.7
GW190727_060333-v1 -7.5 ... 7.1 9.5
GW190728_064510-v1 -1.3 ... 1.7 7.2
GW190731_140936-v1 -10.8 ... 9.7 12.2
GW190803_022701-v1 -8.5 ... 7.8 10.6
GW190814-v2 -0.9 ... 0.08 1.1
GW190828_063405-v1 -4.3 ... 4.6 5.8
GW190828_065509-v1 -4.5 ... 3.6 7.0
GW190909_114149-v1 -16.8 ... 13.4 52.7
GW190910_112807-v1 -8.6 ... 6.3 7.6
GW190915_235702-v1 -6.0 ... 5.6 9.5
GW190924_021846-v1 -1.0 ... 1.4 7.0
GW190929_012149-v1 -25.3 ... 19.3 33.0
GW190930_133541-v1 -1.5 ... 1.7 12.4
The full list of available columns can be queried as follows:
EventTable (output correct as of Dec 17, 2021)¶>>> print(events.info)
<EventTable masked=True length=39>
name dtype unit description
-------------------------------- ------- ------- --------------------------------
name str18
final_mass_source_lower object final_mass_source_lower
chirp_mass_source_lower float64 chirp_mass_source_lower
total_mass_source_upper float64 total_mass_source_upper
chirp_mass_lower float64 chirp_mass_lower
...
Filtered queries¶
The columns returned can be selected using the column keyword,
and mathematical selection filters can be applied on-the-fly
using the selection keyword:
>>> t = EventTable.fetch_open_data(
... "GWTC-2",
... selection="mass_1_source < 4",
... columns=["name", "mass_1_source", "mass_2_source", "luminosity_distance"],
... )
>>> print(t)
name mass_1_source mass_2_source luminosity_distance
solMass solMass Mpc
----------- ------------- ------------- -------------------
GW190425-v2 2.0 1.4 160.0
More on filtering an EventTable
For more information on filtering, see Filtering tables.
Built-in file formats¶
GWpy extends the Astropy functionality with readers for the following file formats:
Each of the sub-sections below outlines how to read and write in these
file formats, include the custom keyword arguments to pass to
EventTable.read() and EventTable.write().
Listing all available formats
To list all available formats, consult the documentation for
EventTable.read().
Multi-processed file reading¶
The EventTable.read() method accepts the nproc
keyword argument, allowing multi-processed reading of lists of files.
This argument can be used with any file-format, not just those defined below,
but is not backported to for use with Table.read().
LIGO_LW XML¶
Additional dependencies: python-ligo-lw
The LIGO Scientific Collaboration uses a custom scheme of XML in which to
store tabular data, called LIGO_LW.
Complementing the scheme is a python library - python-ligo-lw - which
allows users to read and write all of the different types of tabular data
produced by gravitational-wave searches.
Reading and writing tables in LIGO_LW XML format is supported with
format='ligolw', tablename=<tablename>' where <tablename> can be
any of the supported LSC table names (see ligo.lw.lsctables).
Reading¶
When reading, the tablename keyword argument should be given to identify
the table in the file, as follows:
EventTable from LIGO_LW XML.¶>>> t = EventTable.read('H1-LDAS_STRAIN-968654552-10.xml.gz', tablename='sngl_burst')
The result should be something similar to this:
EventTable from LIGO_LW XML.¶>>> print(t)
ifo peak_time peak_time_ns start_time ... confidence chisq chisq_dof bandwidth
--- --------- ------------ ---------- ... ---------- ----- --------- ---------
H1 968654557 783203126 968654557 ... 16.811825 0.0 512.0 256.0
H1 968654557 781250001 968654557 ... 16.816761 0.0 512.0 256.0
H1 968654557 779296876 968654557 ... 16.696106 0.0 512.0 256.0
H1 968654557 777343751 968654557 ... 16.739489 0.0 512.0 256.0
H1 968654557 775390626 968654557 ... 16.802326 0.0 512.0 256.0
H1 968654557 773437501 968654557 ... 16.30731 0.0 512.0 256.0
H1 968654557 771484376 968654557 ... 16.307253 0.0 512.0 256.0
H1 968654557 769531251 968654557 ... 16.35647 0.0 512.0 256.0
H1 968654557 767578126 968654557 ... 16.561176 0.0 512.0 256.0
H1 968654557 765625001 968654557 ... 16.393112 0.0 512.0 256.0
H1 968654557 763671876 968654557 ... 16.404041 0.0 512.0 256.0
H1 968654557 761718751 968654557 ... 16.405825 0.0 512.0 256.0
H1 968654557 759765626 968654557 ... 16.715092 0.0 512.0 256.0
H1 968654557 757812501 968654557 ... 17.512749 0.0 512.0 256.0
H1 968654557 755859376 968654557 ... 17.347675 0.0 512.0 256.0
H1 968654557 753906251 968654557 ... 17.077478 0.0 512.0 256.0
H1 968654557 751953126 968654557 ... 16.742907 0.0 512.0 256.0
... ... ... ... ... ... ... ... ...
H1 968654560 342773438 968654559 ... 11.029792 0.0 16.0 8.0
H1 968654559 280273438 968654558 ... 12.363036 0.0 16.0 8.0
H1 968654559 217773438 968654558 ... 13.985101 0.0 16.0 8.0
H1 968654559 155273438 968654558 ... 14.662391 0.0 16.0 8.0
H1 968654559 92773438 968654558 ... 15.864924 0.0 16.0 8.0
H1 968654559 30273438 968654558 ... 16.321821 0.0 16.0 8.0
H1 968654558 967773438 968654558 ... 16.975931 0.0 16.0 8.0
H1 968654558 905273438 968654558 ... 19.160393 0.0 16.0 8.0
H1 968654560 811523438 968654560 ... 11.270205 0.0 8.0 4.0
H1 968654560 686523438 968654560 ... 15.839205 0.0 8.0 4.0
H1 968654560 561523438 968654560 ... 15.944695 0.0 8.0 4.0
H1 968654560 436523438 968654559 ... 14.384432 0.0 8.0 4.0
H1 968654560 311523438 968654559 ... 14.465309 0.0 8.0 4.0
H1 968654560 186523438 968654559 ... 13.045853 0.0 8.0 4.0
H1 968654560 561523438 968654560 ... 11.636543 0.0 4.0 4.0
H1 968654560 436523438 968654560 ... 15.344837 0.0 4.0 4.0
H1 968654560 311523438 968654560 ... 11.367717 0.0 4.0 4.0
Length = 2052 rows
Single-table files
The tablename keyword can be omitted if there is only a single table
in the file.
To restrict the columns returned in the new EventTable, use the columns
keyword argument:
EventTable from LIGO_LW XML.¶>>> t = EventTable.read(
... "H1-LDAS_STRAIN-968654552-10.xml.gz",
... tablename="sngl_burst",
... columns=["peak_time", "peak_time_ns", "snr", "peak_frequency"],
... )
Many LIGO_LW table objects (as defined in ligo.lw.lsctables) include
utility functions to create new columns by combining others,
e.g. to calculate the Q of a sine-Gaussian pulse from the duration and
central frequency.
These ‘columns’ can be requested directly, providing the
ligo.lw.table.Table representation of the data has a
get_ method for that name:
get_ columns into an EventTable from LIGO_LW XML.¶>>> t = EventTable.read(
... "H1-LDAS_STRAIN-968654552-10.xml.gz",
... tablename="sngl_burst",
... columns=["snr", "q", "duration", "central_freq"],
... )
Note
When reading a processed column in this manner, all required input columns
for a processed column must be included in the columns keyword list.
To exclude these columns from the returned data, use the
ligolw_columns= keyword to specify the columns required to provide
the output columns:
>>> t = EventTable.read(
... "H1-LDAS_STRAIN-968654552-10.xml.gz",
... tablename="sngl_burst",
... columns=["snr", "q"],
... ligolw_columns=["snr", "duration", "central_freq"],
... )
By default, the returned EventTable uses the dtypes returned by the
ligol.lw library, and functions therein, which often end up as
numpy.object_ arrays in the table.
To force all columns to have real numpy data types, use the
use_numpy_dtypes=True keyword, which will cast (known) custom object
types to a standard numpy.dtype, e.g:
use_numpy_dtypes=True when reading an EventTable from LIGO_LW XML.¶>>> t = EventTable.read(
... "H1-LDAS_STRAIN-968654552-10.xml.gz",
... tablename="sngl_burst",
... columns=["peak"],
... ligolw_columns=["peak_time", "peak_time_ns"])
>>> print(type(t[0]["peak"]))
<type 'lal.LIGOTimeGPS'>
>>> t = EventTable.read(
... "H1-LDAS_STRAIN-968654552-10.xml.gz",
... tablename="sngl_burst",
... columns=["peak"],
... ligolw_columns=["peak_time", "peak_time_ns"],
... use_numpy_dtypes=True)
>>> print(type(t[0]["peak"]))
<type 'numpy.float64'>
Writing¶
A table can be written as follows:
>>> t.write('new-table.xml', format='ligolw', tablename='sngl_burst')
Because LIGO_LW isn’t the only scheme of XML, the format keyword is
required for all Table.write() operations.
If the target file already exists, an IOError will be
raised, use overwrite=True to force a new file to be written.
To write a table to an existing file, use append=True:
>>> t.write('new-table.xml', format='ligolw', tablename='sngl_burst', append=True)
To replace an existing table of the given type in an existing file, while
preserving other tables, use both append=True and overwrite=True:
>>> t.write('new-table.xml', format='ligolw', tablename='sngl_burst', append=True, overwrite=True)
Note
The python-ligo-lw library reads and writes files using an updated
version of the LIGO_LW format compared to glue.ligolw used to.
GWpy should support both format versions natively when _reading_, but
when _writing_ you may need to explicitly pass the
ilwdchar_compat=True keyword in order to write using the old
format:
>>> t.write('new-table.xml', format='ligolw', tablename='sngl_burst',
... ilwdchar_compat=True)
Coherence WaveBurst ASCII (aka EVENTS.txt)¶
coherent WaveBurst (cWB) is an analysis pipeline is used to detect generic gravitational-wave
bursts, without using a signal model to restrict the analysis, and runs in
both low-latency (online) and offline modes over current GWO data.
The analysis uses the ROOT framework for most data products, but also
produces ASCII data in a custom format commonly written in a file called
EVENTS.txt.
Reading¶
To read a cWB ASCII file:
EventTable from cWB-format ASCII.¶>>> t = EventTable.read('EVENTS.txt', format='ascii.cwb')
See the astropy.io.ascii.read() documentation for full details on
keyword arguments when reading ascii.cwb files.
Writing¶
To write a table using the cWB ASCII format:
EventTable to a cWB-format ASCII file.¶>>> t.write('EVENTS.txt', format='ascii.cwb')
[the output file name is not required to be 'EVENTS.txt', this is
simply the convention used in the cWB analysis.]
ROOT¶
Additional dependencies: uproot
Note
uproot version 4 (at time of writing) cannot handle writing ROOT files,
so if wish to use this version to read files, you may need to also install
uproot3 as well.
Reading¶
To read a ROOT tree into a Table
(or EventTable), specify the relevant tree via the treename
keyword argument:
EventTable from a ROOT file.¶>>> t = Table.read('my-data.root', treename='triggers')
If treename=None is given (default), a single tree will be read if only one exists in the file, otherwise a ValueError will be raised.
Any other keyword arguments will be passed directly to uproot.tree.TTreeMethods.arrays().
Writing¶
To write a Table as a ROOT tree:
EventTable to a ROOT file.¶>>> t.write('new-table.root')
As with reading, the treename keyword argument can be used to specify
the tree, the default is treename='tree'.
By default, an existing file with an existing tree of the given name will be
appended to, to overwrite use the mode='recreate' keyword argument:
EventTable over an existing tree in a ROOT file.¶>>> t.write('new-table.root', treename='triggers', mode='recreate')
Any other keyword arguments will be passed directly to uproot.newtree.
PyCBC Live (HDF5)¶
PyCBC Live is a low-latency search for gravitational waves from compact binary coalescences, built from the PyCBC analysis package. This search writes files on the LIGO Data Grid (LIGO.ORG-authenticated users only) in HDF5 format, containing tables of events; each column in the table is recorded as a separate HDF5 Dataset.
Reading¶
To read an EventTable from a pycbc_live format HDF5 file, use the
format='hdf5.pycbc_live' keyword:
EventTable from a PyCBC-Live HDF5 file.¶>>> t = EventTable.read("H1-Live-1234567890-4.hdf", format="hdf5.pycbc_live")
To restrict the returned columns, use the columns keyword argument:
EventTable from PyCBC-Live HDF5.¶>>> t = EventTable.read(
... "H1-Live-1234567890-4.hdf",
... format="hdf5.pycbc_live",
... columns=["end_time", "snr", "chisq"],
... )
Similarly to the LIGO_LW XML format, some processed columns can be specified that are not included in the HDF5 files, but are created on-the-fly. Supported processed columns are:
mchirpnew_snr
These can be specified without having to specify any of the input columns.
Additionally, PyCBC HDF5 table Groups include extra datasets that aren’t
part of the table, e.g. 'psd'.
These can be included in the returned EventTable.meta dict via the
keyword extended_metadata=True (default), or excluded with
extended_metadata=False.
Writing¶
Writing tables in PyCBC Live HDF5 format is not supported at this time.
SNAX (HDF5)¶
The SNAX (Signal-based Noise Acquisition and eXtraction) analysis pipeline is a low-latency search for identifying glitches in h(t) and auxiliary channel data using glitch waveforms, operating in low-latency (online) and offline modes. This search writes files on the LIGO Data Grid (LIGO.ORG-authenticated users only) in HDF5 format containing regularly-sampled features; each channel in the table is recorded as a separate HDF5 Dataset.
Reading¶
To read an EventTable from a snax format HDF5 file,
use the format='hdf5.snax' keyword:
EventTable from a SNAX-format HDF5 file.¶>>> t = EventTable.read("H-SNAX_FEATURES-1255853400-20.h5", format="hdf5.snax")
By default, all channels contained in the file are read in.
To restrict the returned channels, use the channels keyword argument:
EventTable from a SNAX-format HDF5 file.¶>>> t = EventTable.read(
... "H-SNAX_FEATURES-1255853400-20.h5",
... format="hdf5.snax",
... channels="H1:CAL-DELTAL_EXTERNAL_DQ",
... )
To restrict the returned columns, use the columns keyword argument:
EventTable from SNAX-format HDF5.¶>>> t = EventTable.read(
... "H-GSTLAL_IDQ_FEATURES-1255853400-20.h5",
... format="hdf5.snax",
... columns=["channel", "time", "snr"],
... )
Writing¶
Writing tables in SNAX HDF5 format is not supported at this time.
GWF¶
Additional dependencies: LDAStools.frameCPP
The Gravitational-Wave Frame file format supports tabular data via
FrEvent structures, which allow storage of arbitrary event information.
Reading¶
To read an EventTable from a GWF-format file, specify the filename and
the name of the FrEvent structures to read:
EventTable from GWF.¶>>> t = EventTable.read('events.gwf', 'my-event-name')
To restrict the returned columns, use the columns keyword argument:
EventTable from GWF.¶>>> t = EventTable.read('events.gwf', 'my-event-name', columns=['time', 'amplitude'])
All FrEvent structures contain the following columns, any other
columns are use-specific:
Column name |
Description (from LIGO-T970130) |
|---|---|
|
Reference time of event, as defined by the search algorithm |
|
Continuous output amplitude returned by event |
|
Likelihood estimate of event, if available (probability = -1 if cannot be estimated) |
|
Signal duration before reference time (seconds) |
|
Signal duration after reference time (seconds) |
|
Descriptor of event |
Writing¶
Writing tables in GWF format is not supported at this time.