Unexpected merging behavior in utils.connect()

[VENOM314]

Hello Jaxley community:

I was connecting pre and post synaptic sites and wrote code like this

pre = net.cell(range(len(synapse_locations))).branch(0).loc(0.5)

branches, locs = zip(*synapse_locations)

# all connect to last cell of net
post = net.cell(len(synapse_locations)).branch(list(branches)).loc(list(locs))

connect(pre, post, AlphaSynapse())

But got error saying
ValueError: Length of values (47) does not match length of index (50)

I am wondering if there is any merging behavior of pre/post views that caused this error as the code seems quite self-contained.

[michaeldeistler]

Hey! Thanks for starting this discussion!

TL;DR: you can fix it as follows:

import pandas as pd

filtered_nodes = pd.concat(
    [net.cell(len(synapse_locations)).branch(b).loc(l).nodes for b, l in zip(branches, locs)]
)
post = net.select(nodes=filtered_nodes.index)

Longer explanation

The reason that this is failing is that you are trying to match pairs of branches and locs at the same time. This is, by default, not supported. To see what is going wrong, consider the following:

ncomp = 2
comp = jx.Compartment()
branch = jx.Branch(comp, ncomp)
cell = jx.Cell(branch, [-1, 0, 0, 1, 1])

branches = [0, 1]
locs = [0.1, 0.9]
cell.branch(branches).loc(locs).nodes  # Display the nodes in a jupyter notebook.

This will show four rows. Why? cell.branch(branches).loc(locs) finds all compartments which have
branch = 0 && loc = 0.1 or branch = 1 && loc = 0.1 or branch = 0 && loc = 0.9 or branch = 1 && loc = 0.9.

If you want to match pairs of branches and locs (as you want to do), you have to manually find the corresponding rows in net.nodes. In your case, you can do this:

filtered_nodes = pd.concat(
    [net.cell(len(synapse_locations)).branch(b).loc(l).nodes for b, l in zip(branches, locs)]
)

Does this make sense?

Self-contained example

For future reference, here is a self-contained example:

import pandas as pd

import jaxley as jx
from jaxley.synapses import IonotropicSynapse
from jaxley.connect import connect

ncomp = 2
comp = jx.Compartment()
branch = jx.Branch(comp, ncomp)
cell = jx.Cell(branch, [-1, 0, 0, 1, 1])
net = jx.Network([cell, cell, cell, cell])

branches = [0, 1, 2]
locs = [0.1, 0.9, 0.9]

pre = net.cell(range(len(branches))).branch(0).loc(0.5)

filtered_nodes = pd.concat(
    [net.cell(len(branches)).branch(b).loc(l).nodes for b, l in zip(branches, locs)]
)
post = net.select(nodes=filtered_nodes.index)

connect(pre, post, IonotropicSynapse())

Michael

[VENOM314]

I see, thank you for the solution and explanation of the dimension mismatch!