Correctly handle 2D/1D custom medium when exporting to GDS #1247
Conversation
lucas-flexcompute
force-pushed
the
lucas/custom-medium-to-gds-fix
branch
from
7ed7d02
to
6ffa06c
Compare
There was a problem hiding this comment.
Thanks @lucas-flexcompute, just a minor comment suggesting some more explanation so it's more clear what this fixes but otherwise looks good!
| scale = max(b - a for a, b in zip(bb_min, bb_max)) | ||
| for coord in (eps.x, eps.y, eps.z): | ||
| if len(coord) > 1: | ||
| scale = min(scale, np.diff(coord).min()) |
There was a problem hiding this comment.
what exactly does this do? My reading is that for the scale it takes the minimum of
- the custom medium bounding box size
- the custom medium pixel size?
What edge case does this fix? I guess before were custom medium not being resolved properly with the provided scale?
Maybe a comment here and also adding more detail in the changelog could be helpful down the line.
- Properly take pixel size into account in the length scale when converting a custom medium structure to GDS.
or something.
There was a problem hiding this comment.
The pixel size was already taken into account before. The scale is the minimal distance between custom medium points with respect to each direction (x, y, z). The problem was that, if any of the directions had a single point, np.diff returned an empty array and the min function would raise an error. Now we initialize the scale with the largest possible value (from the bounding box) and, for each direction with at least 2 coordinates, keep only the minimum. Does it make sense to you? If so, I'll add those comments.
There was a problem hiding this comment.
oh I see, yea if you could just add a comment in the source and maybe make the changelog item more specific to that effect, it would help a lot, thanks for explaining.
- Properly handle length sale in exporting custom medium structures to GDS when the geometry size is very small along some dimensions.
or something
The 0D case is also handled just to avoid an obscure error. Signed-off-by: Lucas Heitzmann Gabrielli <lucas@flexcompute.com>
lucas-flexcompute
force-pushed
the
lucas/custom-medium-to-gds-fix
branch
from
6ffa06c
to
3e71ab5
Compare
The 0D case is also handled just to avoid an obscure error.