NestIO compatible with Nest 3.2 #1097

morales-gregorio · 2022-03-09T12:26:41Z

Hi all,

@rgutzen @jasperalbers and I were looking into the NestIO today and we came up with this edits that make the current NestIO compatible with both Nest 2.x and 3.x versions.

We incorporated all the changes suggested by @essink into this PR.

Essentially we force the IO to ignore all the headers (which raises a warning that it did so) as suggested by @rgutzen, which I also wrote into PR #739

Additionally, we found that the current NestIO decides whether to read neo.AnalogSignals or neo.SpikeTrain based on the file name alone. This is problematic since Nest 3.2 (maybe others too) saves spikes to the .dat format by default, which the IO expected to be analogsignals. I fixed this issue by adding a new keyword target_object which then determines what kind of object will be read. This does not allow to read analogsignals and spikes into the same segment (two different calls are needed), but this seems a rather good practice in my opinion, since spiketrains and analogsignales are rarely combined into the same segment.

I tested the code to load a .dat file generated with Nest 3.2 and it works, I can add a proper test somewhere if you want.

Best,
Aitor

morales-gregorio · 2022-03-09T14:31:54Z

Looking into the capability to deal with lists it only seems to complicate things and it is not capable of dealing with multiple files of the same type, which is something I would expect. I would suggest getting rid of the capability to deal with lists altogether and stick to single files as most other IOs do. This then allows to remove code further and if someone really wants to have the analogsignals and spikes together it is just a matter of running this two times.

Moritz-Alexander-Kern · 2022-08-12T10:36:54Z

Hi all,
thanks for taking care of this.
We created a NEST-elephant tutorial and we ran into this issue with the nestio and NEST 3.3.

Since this was tested with NEST 3.2, maybe add this info to the docstring(?):

python-neo/neo/io/nestio.py

Line 4 in 50abfbc

Tested with NEST2.10.0

JuliaSprenger · 2023-02-16T14:14:14Z

Hi @morales-gregorio Do you have a small test file for this format that could be added to gin?

morales-gregorio · 2023-02-17T14:25:52Z

Just rebased the branch and some chaos has followed. My local history looks as one would expect, with this branch now splitting from the current master. I am not sure why Github interprets that all the previous commits are part of this PR :/

morales-gregorio · 2023-02-17T17:04:58Z

@JuliaSprenger I created a small NEST (3.3) simulation where I record both spikes and analogsignals. The results and simulation script are here: test_NEST_simulation_and_results.zip.

The current intended use is therefore something like this:

import neo
import quantities as pq
import matplotlib.pyplot as plt
# Load spikes
io = neo.NestIO(['sr1-142-0.dat', 'sr2-154-0.dat'], target_object='SpikeTrain')
bl = io.read_block([], t_start=0*pq.s, t_stop=20*pq.s)

# Plot spikes
plt.figure(figsize=(4, 3))
for i, st in enumerate(bl.segments[0].spiketrains):
    if st.annotations['source_file'] == 'sr1-142-0.dat':
        color = 'b'
    elif st.annotations['source_file'] == 'sr2-154-0.dat':
        color = 'r'
    plt.scatter(st.times, [i]*len(st.times), marker='|', c=color)
plt.xlabel('Time (ms)')
plt.ylabel('Neurons')
plt.savefig('nest_spikes.png', dpi=300)

# Load anasigs
io = neo.NestIO(['vm1-141-0.dat', 'vm2-153-0.dat'], target_object='AnalogSignal')
bl = io.read_block([], t_start=0*pq.s, t_stop=20*pq.s)

# Plot anasigs
plt.figure(figsize=(4, 3))
for i, ana in enumerate(bl.segments[0].analogsignals):
    if ana.annotations['source_file'] == 'vm1-141-0.dat':
        color = 'b'
    elif ana.annotations['source_file'] == 'vm2-153-0.dat':
        color = 'r'

    plt.plot(ana.times, 8*i + ana.magnitude, c=color, alpha=0.5)

plt.xlabel('Time (ms)')
plt.ylabel('Neurons')
plt.savefig('nest_anasigs.png', dpi=300)

Which creates these two plots in this case:

Co-Authored-By: Simon Essink <[email protected]>

Co-Authored-By: Robin Gutzen <[email protected]>

Co-Authored-By: Robin Gutzen <[email protected]> Co-Authored-By: Simon Essink <[email protected]> Co-Authored-By: Jasper Albers <[email protected]>

… longer dependent on the file format. Previous behaviour lead to the IO trying to always read spikes from .gdf files and analogsignals from .dat files, which is not necessarily the case with Nest 3.2 (possibly earlier)

Co-Authored-By: Simon Essink <[email protected]>

Moritz-Alexander-Kern · 2023-08-03T11:53:34Z

@JuliaSprenger I just fixed the history, are there any open issues with the current version of the NestIO?

or @apdavison

mdenker

This is really great, and I was able to successfully read in a couple of gdf from NEST 3.3. I marked a few points in this code that caught my eye when going through the changes.

neo/io/nestio.py

mdenker · 2024-01-31T21:24:19Z

neo/io/nestio.py

@@ -482,8 +500,8 @@ def read_segment(self, gid_list=None, time_unit=pq.ms, t_start=None,
        gid_list : list, default: None
            A list of GDF IDs of which to return SpikeTrain(s). gid_list must
            be specified if the GDF file contains neuron IDs, the default None
-            then raises an error. Specify an empty list [] to retrieve the spike
-            trains of all neurons.
+            then raises an error. Specify an empty list [] to retrieve the


This help text is not complete: The code also accepts tuples of the form (a,b), which are expanded to range(a,b). I am not sure this is particularly helpful, I would suggest to drop this code below (line 538), rather than adapt this help text to include this case.

mdenker · 2024-01-31T21:28:42Z

neo/io/nestio.py

-            gid_ids = np.array([np.searchsorted(data[:, 0], gid, side='left'),
-                                np.searchsorted(data[:, 0], gid, side='right')])
-        gid_data = data[gid_ids[0]:gid_ids[1], :]
+            gids = np.array([np.searchsorted(data[:, 0], gid, side='left'),


Should not this be the searched in the ID column, i.e.,

Suggested change

gids = np.array([np.searchsorted(data[:, 0], gid, side='left'),

gids = np.array([np.searchsorted(data[:, id_column], gid, side='left'),

A similar change is probably needed below for id_shifts, i.e.,

id_shifts[0] = np.searchsorted(gid_data[:, time_column],

mdenker · 2024-01-31T21:34:39Z

neo/io/nestio.py

-            spiketrain_list = [SpikeTrain(train, units=time_unit,
-                                          t_start=t_start, t_stop=t_stop,
-                                          id=None, **args)]
+        spiketrain_list = []


The spiketrains are read here as SpikeTrain object, however, meanwhile, a SpikeTrainList object is probably more appropriate. Perhaps this would be a separate PR.

Then again, maybe one could just

Suggested change

spiketrain_list = []

spiketrain_list = SpikeTrainList

The NestIO was written before SpikeTrainList even existed, so yes, we should use SpikeTrainList, since this is also the superior way to load many spiketrains (which simulations will create)

mdenker · 2024-01-31T21:46:17Z

neo/io/nestio.py

-        # extracting complete gid list for anasig generation
-        if (gid_list == []) and id_column is not None:
-            gid_list = np.unique(data[:, id_column])
+        analogsignal_list = []


If I understand the NEST analogsignal outputs correctly, one could consider creating one AnalogSignal per file (i.e., for each entry in self.IOs), and then add all channels (gids) into columns of that AnalogSignal. Currently, separate AnalogSignals are created for each gid in each file.

This probably needs discussion.

I agree that a single analogsignal object should be created per source file

I think for multimeter files it should be one AnalogSignal per recorded variable (i.e. per column in the source file, not including the time and id columns), but each signal object should contain channels for all neuron ids.

mdenker · 2024-01-31T21:49:32Z

neo/io/nestio.py

-            then raises an error. Specify an empty list [] to retrieve the spike
-            trains of all neurons.
+            then raises an error. Specify an empty list [] to retrieve the
+            spike trains of all neurons.
        time_unit : Quantity (time), optional, default: quantities.ms
            The time unit of recorded time stamps in DAT as well as GDF files.
        t_start : Quantity (time), optional, default: 0 * pq.ms


This is not quite true, the default is None. The identical discrepancy is found in a few doc strings.

heplesser

Dear @morales-gregorio!

Thank you for preparing this PR, the lack of support for NEST3 in NEO has been holding back use of many useful tools for some time. I must admit that I do not know NEO in detail, so some of my comments below are probably just due to my ignorance.

In addition to the comments in the code, I have some points that concern importing from NEST overall. I focus here on NEST 3. For backward compatibility with NEST2, I would suggest to check if the first line of file begins with a numeric value. If so, it must be NEST2 and can be read with a NEST2 reader.

File header

If I understand your code correctly, it simply ignores the three header lines of NEST ASCII files. This is okay to just get things to work again, but I think it would make much more sense to make use of this information. The header looks like this:

# NEST version: 3.6.0-post0.dev0
# RecordingBackendASCII version: 2
sender	time_ms
1	9.900
1	21.800

It provides several bits of useful information which could be extracted with regexes:

The first line can be used to check that it really is NEST output and to record the NEST version in metadata (if supported by NEO)
The second line can be used to check NEST output format version, in case we later introduce different versions; this would allow explicit selection of the compatible reader
One could actually parse those first two lines and if they do not match conclude that it must be a NEST 2 file.
The third line identifies columns. This becomes important in particular when recording with the multimeter or if time is not recorded in milliseconds (see examples below). I think it would be advantageous if the column headers were read and used to label the signals recorded.

Data formats

The following script produces output in all formats NEST currently supports. Note that the voltmeter is just a multimeter pre-configured to record V_m, so the formats are redundant. I still provide both examples, since voltmeter is widely used.

Notes

In the examples below, all offsets are zero. If one simulates with precise-spiking neurons, they can differ from zero.
Where time is given in steps, the actual spike time is calculated as steps * resolution - offset.
The current NEST data format using steps/offset is actually incomplete, because the resolution, i.e., the duration of the time step is not given in the file. We need to fix that in the next version of the recording file format. Therefore, I think it is okay for NEO not to support that format a present.
Recording time in steps is technically possible for voltmeter/multimeter, but makes little sense, so it is okay not to support that format.
In multimeter recordings, the quantities recorded depend on the neuron model and the choices made by the experimenter. I am not sure if any order is imposed on the columns (e.g., alphabetical).
File rows are not guaranteed to be sorted either by time or GID.

Sample script

nest.ResetKernel()
nest.overwrite_files = True

n = nest.Create('iaf_psc_alpha', params={'I_e': 600})
s1 = nest.Create('spike_recorder', params={'record_to': 'ascii'})
s2 = nest.Create('spike_recorder', params={'record_to': 'ascii', 'time_in_steps': True})
v1 = nest.Create('voltmeter', params={'record_to': 'ascii', 'interval': nest.resolution})
v2 = nest.Create('voltmeter', params={'record_to': 'ascii', 'interval': nest.resolution, 'time_in_steps': True})
m1 = nest.Create('multimeter', params={'record_to': 'ascii', 'interval': nest.resolution,
                                       'record_from': n.recordables})
m2 = nest.Create('multimeter', params={'record_to': 'ascii', 'interval': nest.resolution,
                                       'record_from': n.recordables, 'time_in_steps': True})

nest.Connect(n, s1 + s2)
nest.Connect(v1 + v2 + m1 + m2, n)

nest.Simulate(100)

Plain spike recording

# NEST version: 3.6.0-post0.dev0
# RecordingBackendASCII version: 2
sender	time_ms
1	9.900
1	21.800

Spike recording with time in steps

# NEST version: 3.6.0-post0.dev0
# RecordingBackendASCII version: 2
sender	time_step	time_offset
1	99	0.000
1	218	0.000

Plain voltmeter

# NEST version: 3.6.0-post0.dev0
# RecordingBackendASCII version: 2
sender	time_ms	V_m
1	0.100	-69.761
1	0.200	-69.525

Voltmeter with time in steps

time_offset will always be zero.

# NEST version: 3.6.0-post0.dev0
# RecordingBackendASCII version: 2
sender	time_step	time_offset	V_m
1	1	0.000	-69.761
1	2	0.000	-69.525

Multimeter

Also possible with time in steps
Columns beyond time_ms depend on model and user choice

# NEST version: 3.6.0-post0.dev0
# RecordingBackendASCII version: 2
sender	time_ms	I_syn_ex	I_syn_in	V_m
1	0.100	0.000	0.000	-69.761
1	0.200	0.000	0.000	-69.525

Data from parallel simulations

When recording from parallel simulations, file names will have the form <label>-<gid>-<vp>.dat, with each virtual process recording to a single file
Data needs to be collected for all files matching <label>-<gid>-[0-9]+.dat
This can be done with the current interface by providing a list of all file names matching the pattern.
For a better user interface, it might make sense to allow the user to specify a pattern and then do the globbing internally

Support for pathlib

Currently, file names can only be provided as strings. In the longer term, one could provide a more powerful interface by supporting also Python's pathlib, maybe even including globs.

heplesser · 2024-01-31T18:39:49Z

neo/io/nestio.py

    mode = 'file'

-    def __init__(self, filenames=None):
+    def __init__(self, filenames=None, target_object='SpikeTrain',
+                 additional_parameters={}):


Never use a mutable as default argument value. Use None instead and then create the empty dictionary inside the constructor if None is passed. Otherwise, the dict is created once when the class is defined, not every time the constructor is called.

The additional_paramters parameter is also not described in the docstring.

heplesser · 2024-01-31T18:46:39Z

neo/io/nestio.py

+        self.IOs = []
        for filename in filenames:
-            path, ext = os.path.splitext(filename)
-            ext = ext.strip('.')
-            if ext in self.extensions:
-                if ext in self.avail_IOs:
-                    raise ValueError('Received multiple files with "%s" '
-                                     'extension. Can only load single file of '
-                                     'this type.' % ext)
-                self.avail_IOs[ext] = ColumnIO(filename)
-            self.avail_formats[ext] = path
+            self.IOs.append(ColumnIO(filename, additional_parameters))


You could turn this into a list comprehension

self.IOs = [ColumnIO(filename, additional_parameters) for filename in filenames]

heplesser · 2024-01-31T18:49:21Z

neo/io/nestio.py

-        column_ids = [id_column, time_column] + value_columns
+        column_ids = [id_column, time_column]
+        if value_columns is not None:
+            column_ids += value_columns


Does this work? value_columns is an int from above and here you add it to a list? Should it be .append(value_columns)?

And how could value_columns be None?

heplesser · 2024-02-01T08:21:07Z

neo/io/nestio.py

+        column_list = [id_column, time_column]
+        if value_columns is not None:
+            column_list += value_columns


How does this differ from the column_ids above?

heplesser · 2024-02-01T08:21:23Z

neo/io/nestio.py

-        column_ids = [id_column, time_column] + value_columns
+        column_ids = [id_column, time_column]
+        if value_columns is not None:
+            column_ids += value_columns
        for i, cid in enumerate(column_ids):
            if cid is None:


How could cid be None?

neo/io/nestio.py

heplesser · 2024-02-01T08:28:22Z

neo/io/nestio.py

        if '.' not in line:
            additional_parameters['dtype'] = np.int32
+        else:
+            additional_parameters['dtype'] = np.float32


Why do you use 32-bit types here? In the "full K" simulation we had 1.86e9 neurons, so the largest GIDs got very close to the upper end of signed 32 bit ints. For coming brainscale models, GIDs will need the 64 bit range. For long-running simulations, there is also a risk of loss of time precision if we only have 7-8 digits in the mantissa of floats.

mdenker · 2024-02-01T13:19:44Z

Hi @heplesser , thanks for the comments and the detailed description of NEST outputs. I think this information is super helpful in driving this IO forwards. Also, thanks for your insights in problems that may arise at the "full scale" level.

…stIO_3.2

Co-authored-by: Michael Denker <[email protected]>

heplesser · 2024-04-05T10:19:25Z

Hello everyone!

After discussion with @mdenker and @apdavison yesterday, I provide here a little script that creates a network consisting of

aeif_cond_alpha neurons (to show multimeter recording)
iaf_psc_alpha_ps neurons (to show precise time recording)
weight_recorder (to show weight recording)

Running the script will generate a number of spike recorder, voltmeter, multimeter and weight recorder output files with different settings. Note in particular the effect of "time_in_steps": If false (default), time is recorded in ms under heading "time_ms", if true time will be recorded in columns "time_step" and "time_offset", where the actual spike time is given by time_step * resolution - time_offset. Unfortunately, the resolution is not stored in the file, so I suggest to not support these files for now.

Do not rely on ordering of columns or rows. In parallel simulations, there will be one file per virtual process, enumerated by the last part of the file name.

NEST 3 files can be safely identified by the

# NEST version: 3.7.0-rc1
# RecordingBackendASCII version: 2

header. NEST 2 files do not have these headers. I would suggest to use the header for identification, also to be prepared for future versions of the ASCII backend.

nest.ResetKernel()
nest.overwrite_files = True

g = nest.Create('poisson_generator', params={'rate': 100000})
n = nest.Create('aeif_cond_alpha', 10)
p = nest.Create('iaf_psc_alpha_ps', 5)

rnt = nest.Create('spike_recorder', params={'record_to': 'ascii',
                                            'label': 'aeif_spikes_times'})
rns = nest.Create('spike_recorder', params={'record_to': 'ascii',
                                           'time_in_steps': True,
                                            'label': 'aeif_spikes_steps'})

rpt = nest.Create('spike_recorder', params={'record_to': 'ascii',
                                            'label': 'precise_spikes_times'})
rps = nest.Create('spike_recorder', params={'record_to': 'ascii',
                                           'time_in_steps': True,
                                            'label': 'precise_spikes_steps'})

vt = nest.Create('voltmeter', params={'record_to': 'ascii',
                                      'label': 'voltmeter_times'})
vs = nest.Create('voltmeter', params={'time_in_steps': True,
                                      'label': 'voltmeter_steps'})

mm1t = nest.Create('multimeter', params={'record_from': ('g_ex', 'g_in', 'V_m', 'w'),
                                         'record_to': 'ascii',
                                            'label': 'multimeter_1ms'})
mmrt = nest.Create('multimeter', params={'record_from': ('g_ex', 'g_in', 'V_m', 'w'),
                                         'record_to': 'ascii',
                                         'interval': nest.resolution,
                                            'label': 'multimeter_res'})

w = nest.Create('weight_recorder', params={'record_to': 'ascii',
                                            'label': 'weight_rec'})

nest.SetDefaults('stdp_synapse', {'weight_recorder': w})

nest.Connect(g, n)
nest.Connect(n, n, syn_spec={'synapse_model': 'stdp_synapse', 'weight': 100.})
nest.Connect(n, p, syn_spec={'weight': 1000.})
nest.Connect(n, rnt + rns)
nest.Connect(p, rpt + rps)
nest.Connect(vt + vs + mm1t + mmrt, n)

nest.Simulate(100)

morales-gregorio · 2024-04-05T10:37:57Z

Hi @heplesser ! Thanks for pushing this forward and coming up with good test examples.

I think we should not introduce a NEST dependency to the neo tests, I would instead suggest to create several (small) files for multiple NEST versions. The files should have descriptive names like "nest3.7_voltmeter_ascii.dat". What do you think?

heplesser · 2024-04-05T11:13:24Z

Hi @heplesser ! Thanks for pushing this forward and coming up with good test examples.

I think we should not introduce a NEST dependency to the neo tests, I would instead suggest to create several (small) files for multiple NEST versions. The files should have descriptive names like "nest3.7_voltmeter_ascii.dat". What do you think?

@morales-gregorio I agree, otherwise we could easily end up with circular dependencies and all-to-heavy test setups. Frozen example files for testing should work nicely. It should suffice to base the versioning on the ASCII Recording Backend version, which changes more slowly than the NEST version.

morales-gregorio · 2024-04-05T11:27:20Z

Hi @heplesser ! Thanks for pushing this forward and coming up with good test examples.
I think we should not introduce a NEST dependency to the neo tests, I would instead suggest to create several (small) files for multiple NEST versions. The files should have descriptive names like "nest3.7_voltmeter_ascii.dat". What do you think?

@morales-gregorio I agree, otherwise we could easily end up with circular dependencies and all-to-heavy test setups. Frozen example files for testing should work nicely. It should suffice to base the versioning on the ASCII Recording Backend version, which changes more slowly than the NEST version.

@heplesser could you create those files? 😇

Moritz-Alexander-Kern · 2024-04-05T11:32:53Z

Hey @morales-gregorio ,

the current test files for testing NEST-IO are located on NeuralEnsemble /
ephy_testing_data in nest here: https://gin.g-node.org/NeuralEnsemble/ephy_testing_data/src/master/nest

The contents are:

0gid-1time-1256-0.gdf
0gid-1time-2gex-1262-0.dat
0gid-1time-2gex-3Vm-1261-0.dat
0gid-1time-2Vm-1259-0.dat
0gid-1time-2Vm-3gex-4gin-1260-0.dat
0gid-1time-2Vm-3Iex-4Iin-1264-0.dat
0gid-1time_in_steps-1258-0.gdf
0gid-1time_in_steps-2Vm-1263-0.dat
0time-1255-0.gdf
0time_in_steps-1257-0.gdf
brunel-delta-nest_mod.py  # <--- script that generated the files
N1-0gid-1time-2Vm-1265-0.dat
N1-0time-1Vm-1266-0.dat
N1-0Vm-1267-0.dat

I have copied and formatted the above script into generate_data.py:

"""
Script to generate test data for the nestIO.

This script creates a network simulation using the NEST simulator (https://www.nest-simulator.org/).
It generates output files for spike recorders, voltmeters, multimeters and weight recorders with
different settings. Note the effect of "time_in_steps": If false (default), time is recorded in ms
under heading "time_ms"; if true, time will be recorded in columns "time_step" and "time_offset",
where the actual spike time is given by time_step * resolution - time_offset. The resolution is not
stored in the file. In parallel simulations, there will be one file per virtual process, enumerated
by the last part of the file name.

Usage:
    python generate_data.py


"""

import nest


def main():
    """
    Executes the main logic of the script.

    This function sets up a simulation using the NEST simulator (https://www.nest-simulator.org/)
    for simulating neural networks. It creates a network consisting of aeif_cond_alpha neurons
    (to show multimeter recording), iaf_psc_alpha_ps neurons (to show precise time recording),
    and weight_recorder (to show weight recording). The simulation is run for a specified duration.

    Note: Do not rely on ordering of columns or rows. In parallel simulations, there will be
    one file per virtual process, enumerated by the last part of the file name.

    Parameters:
        None

    Returns:
        None

    """
    nest.ResetKernel()
    nest.overwrite_files = True

    g = nest.Create("poisson_generator", params={"rate": 100000})
    n = nest.Create("aeif_cond_alpha", 10)
    p = nest.Create("iaf_psc_alpha_ps", 5)

    rnt = nest.Create(
        "spike_recorder",
        params={"record_to": "ascii", "label": "aeif_spikes_times"},
    )
    rns = nest.Create(
        "spike_recorder",
        params={
            "record_to": "ascii",
            "time_in_steps": True,
            "label": "aeif_spikes_steps",
        },
    )

    rpt = nest.Create(
        "spike_recorder",
        params={"record_to": "ascii", "label": "precise_spikes_times"},
    )
    rps = nest.Create(
        "spike_recorder",
        params={
            "record_to": "ascii",
            "time_in_steps": True,
            "label": "precise_spikes_steps",
        },
    )

    vt = nest.Create("voltmeter", params={"record_to": "ascii", "label": "voltmeter_times"})
    vs = nest.Create("voltmeter", params={"time_in_steps": True, "label": "voltmeter_steps"})

    mm1t = nest.Create(
        "multimeter",
        params={
            "record_from": ("g_ex", "g_in", "V_m", "w"),
            "record_to": "ascii",
            "label": "multimeter_1ms",
        },
    )
    mmrt = nest.Create(
        "multimeter",
        params={
            "record_from": ("g_ex", "g_in", "V_m", "w"),
            "record_to": "ascii",
            "interval": nest.resolution,
            "label": "multimeter_res",
        },
    )

    w = nest.Create("weight_recorder", params={"record_to": "ascii", "label": "weight_rec"})

    nest.SetDefaults("stdp_synapse", {"weight_recorder": w})

    nest.Connect(g, n)
    nest.Connect(n, n, syn_spec={"synapse_model": "stdp_synapse", "weight": 100.0})
    nest.Connect(n, p, syn_spec={"weight": 1000.0})
    nest.Connect(n, rnt + rns)
    nest.Connect(p, rpt + rps)
    nest.Connect(vt + vs + mm1t + mmrt, n)

    nest.Simulate(100)


if __name__ == "__main__":
    main()

which generates: (.zip attached)

aeif_spikes_steps-18-0.dat
aeif_spikes_times-17-0.dat
generate_data.py
multimeter_1ms-23-0.dat
multimeter_res-24-0.dat
precise_spikes_steps-20-0.dat
precise_spikes_times-19-0.dat
voltmeter_times-21-0.dat
weight_rec-25-0.dat

nestIO_test_data.zip

Refer to this readme here: https://gin.g-node.org/NeuralEnsemble/ephy_testing_data to add the files.

…nto NestIO_3.2

Co-authored-by: Hans Ekkehard Plesser <[email protected]>

apdavison · 2024-05-06T14:20:13Z

There are 8 failing tests. I've looked into the first of these, and it seems to be that the case of id_column=None is not being correctly handled.

JuliaSprenger added this to the 0.11.0 milestone Mar 31, 2022

JuliaSprenger requested a review from apdavison June 23, 2022 14:15

apdavison modified the milestones: 0.10.3, 0.11.0 Aug 30, 2022

apdavison modified the milestones: 0.11.0, 0.12.0 Sep 29, 2022

JuliaSprenger modified the milestones: 0.12.0, 0.12.1, 0.13.0 Apr 2, 2023

JuliaSprenger self-assigned this Apr 3, 2023

morales-gregorio and others added 10 commits August 3, 2023 11:56

Simon's changes

0f7639b

Co-Authored-By: Simon Essink <[email protected]>

Ignore file header if there is one

1cbf80c

Co-Authored-By: Robin Gutzen <[email protected]>

change author list

b52282b

Co-Authored-By: Robin Gutzen <[email protected]> Co-Authored-By: Simon Essink <[email protected]> Co-Authored-By: Jasper Albers <[email protected]>

turn additional_parameters into a kwarg

42137a8

remove support for lists

de0b512

pass the correct kwargs in the tests

f73f38d

minor fixes

5a9abcf

Explicitly read as floats if a dot is found

496552c

Simon's changes

cf65570

Co-Authored-By: Simon Essink <[email protected]>

fix columnIO kwarg

5a2b9db

apdavison assigned apdavison and mdenker Jan 26, 2024

mdenker requested changes Jan 31, 2024

View reviewed changes

heplesser reviewed Feb 1, 2024

View reviewed changes

heplesser mentioned this pull request Feb 1, 2024

Documentation for recording spike times is obtuse nest/nest-simulator#3070

Closed

apdavison modified the milestones: 0.13.0, 0.13.1 Feb 2, 2024

morales-gregorio and others added 4 commits April 4, 2024 15:19

Merge branch 'master' of github.com:NeuralEnsemble/python-neo into Ne…

f3a7484

…stIO_3.2

rename additional_parameters to kwargs

61fd293

Update neo/io/nestio.py

342a943

Co-authored-by: Michael Denker <[email protected]>

Update neo/io/nestio.py

8c5e62e

Co-authored-by: Michael Denker <[email protected]>

morales-gregorio and others added 6 commits April 8, 2024 20:15

turn loop to compact list comprehension

d1e97a0

fix hardcoded column indices

78bb2ab

fix too long line

d65a104

Merge branch 'NestIO_3.2' of github.com:morales-gregorio/python-neo i…

01294c8

…nto NestIO_3.2

Update neo/io/nestio.py

7530b84

Co-authored-by: Hans Ekkehard Plesser <[email protected]>

Update neo/io/nestio.py

a96794b

Co-authored-by: Hans Ekkehard Plesser <[email protected]>

zm711 modified the milestones: 0.13.1, 0.14.0 May 3, 2024

zm711 modified the milestones: 0.14.0, 0.15.0 Dec 13, 2024

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

NestIO compatible with Nest 3.2 #1097

NestIO compatible with Nest 3.2 #1097

morales-gregorio commented Mar 9, 2022

morales-gregorio commented Mar 9, 2022

Moritz-Alexander-Kern commented Aug 12, 2022

JuliaSprenger commented Feb 16, 2023

morales-gregorio commented Feb 17, 2023

morales-gregorio commented Feb 17, 2023

Moritz-Alexander-Kern commented Aug 3, 2023

mdenker left a comment

mdenker Jan 31, 2024

mdenker Jan 31, 2024

mdenker Jan 31, 2024

mdenker Jan 31, 2024

morales-gregorio Apr 4, 2024

mdenker Jan 31, 2024

mdenker Jan 31, 2024

morales-gregorio Apr 5, 2024

apdavison Apr 5, 2024

mdenker Jan 31, 2024

heplesser left a comment

heplesser Jan 31, 2024

heplesser Jan 31, 2024

morales-gregorio Apr 8, 2024

heplesser Jan 31, 2024

heplesser Feb 1, 2024

heplesser Feb 1, 2024

heplesser Feb 1, 2024

mdenker commented Feb 1, 2024

heplesser commented Apr 5, 2024

morales-gregorio commented Apr 5, 2024

heplesser commented Apr 5, 2024

morales-gregorio commented Apr 5, 2024 •

edited

Loading

Moritz-Alexander-Kern commented Apr 5, 2024 •

edited

Loading

apdavison commented May 6, 2024

	gids = np.array([np.searchsorted(data[:, 0], gid, side='left'),
	gids = np.array([np.searchsorted(data[:, id_column], gid, side='left'),

NestIO compatible with Nest 3.2 #1097

Are you sure you want to change the base?

NestIO compatible with Nest 3.2 #1097

Conversation

morales-gregorio commented Mar 9, 2022

morales-gregorio commented Mar 9, 2022

Moritz-Alexander-Kern commented Aug 12, 2022

JuliaSprenger commented Feb 16, 2023

morales-gregorio commented Feb 17, 2023

morales-gregorio commented Feb 17, 2023

Moritz-Alexander-Kern commented Aug 3, 2023

mdenker left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

heplesser left a comment

Choose a reason for hiding this comment

File header

Data formats

Notes

Sample script

Plain spike recording

Spike recording with time in steps

Plain voltmeter

Voltmeter with time in steps

Multimeter

Data from parallel simulations

Support for pathlib

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

mdenker commented Feb 1, 2024

heplesser commented Apr 5, 2024

morales-gregorio commented Apr 5, 2024

heplesser commented Apr 5, 2024

morales-gregorio commented Apr 5, 2024 • edited Loading

Moritz-Alexander-Kern commented Apr 5, 2024 • edited Loading

apdavison commented May 6, 2024

morales-gregorio commented Apr 5, 2024 •

edited

Loading

Moritz-Alexander-Kern commented Apr 5, 2024 •

edited

Loading