Repo/version: gplugins 2.0.1 on gdsfactory 9.34.2
What happened: In gplugins.gmeep.get_simulation, the vertical extent is sized using
layers_thickness = [layer_to_thickness[layer] for layer in component.layers
if layer in layer_to_thickness]
t_core = sum(layers_thickness)component.layers yields GDS tuples such as (3, 0), while the keys of layer_to_thickness are layer objects (LogicalLayer, and derived ones like WG - GRA). They never compare equal, so layers_thickness is empty and t_core is 0 for every Cornerstone component tried. The 220 nm core contributes nothing to the domain it is supposed to size.
Steps to reproduce:
import gdsfactory as gf
from cspdk.si220.cband import PDK
PDK.activate()
stack = PDK.get_layer_stack()
stack.get_layer_to_thickness().keys() # LogicalLayer/DerivedLayer objects
gf.get_component("mmi2x2").layers # [(3, 0)]Expected: t_core should capture the 220 nm core thickness (and any other layers the component touches) to set the vertical cell extent correctly.
Actual: layers_thickness comes back empty and t_core is 0. This compounds with issue 3 in the same file's account: the ignored margin makes the cell too large, while the empty t_core makes it too small. Both are silent.
Suggested fix: Normalise both sides to the same layer representation before the membership test, and raise rather than continue when the result is empty — a zero core thickness is never a legitimate answer. Independently, sum is the wrong reduction for stacked layers that overlap in z; max(zmax) - min(zmin) over the levels the component touches would be the geometric answer.
Repo/version: gplugins 2.0.1 on gdsfactory 9.34.2
What happened: In
gplugins.gmeep.get_simulation, the vertical extent is sized usingcomponent.layersyields GDS tuples such as(3, 0), while the keys oflayer_to_thicknessare layer objects (LogicalLayer, and derived ones likeWG - GRA). They never compare equal, solayers_thicknessis empty andt_coreis 0 for every Cornerstone component tried. The 220 nm core contributes nothing to the domain it is supposed to size.Steps to reproduce:
Expected:
t_coreshould capture the 220 nm core thickness (and any other layers the component touches) to set the vertical cell extent correctly.Actual:
layers_thicknesscomes back empty andt_coreis 0. This compounds with issue 3 in the same file's account: the ignored margin makes the cell too large, while the emptyt_coremakes it too small. Both are silent.Suggested fix: Normalise both sides to the same layer representation before the membership test, and raise rather than continue when the result is empty — a zero core thickness is never a legitimate answer. Independently,
sumis the wrong reduction for stacked layers that overlap in z;max(zmax) - min(zmin)over the levels the component touches would be the geometric answer.