Repository navigation
Cloud from numpy doesn't save las with shift. #218
Description
Hello -
We have upgraded to the latest release of CloudComPy (February 20, 2024 -> June 19, 2025) and we have trouble saving LAS/LAZ clouds created from numpy arrays - The shift is not used when saving. The same issue does not occur for bin/xyz.
Without respecting the shift, the points overflow on save, with LAS header offset of saved file as [0,0,0].
What we want is to save out the file using the shift that was set.
To reproduce:
import numpy as np
import cloudComPy as cc
print("=== Create cloud ===")
points = np.array([[0.0, 0.0, 0.0]], dtype=np.float32)
cloud = cc.ccPointCloud("test")
cloud.coordsFromNPArray_copy(points)
cloud.setGlobalShift(0.0, -2_500_000.0, 0.0)
print("Shift:", cloud.getGlobalShift())
print("isShifted:", cloud.isShifted())
print("Points:", cloud.toNpArrayCopy())
print("=== Save ===")
cc.SavePointCloud(cloud, "test.las")
print("=== Load ===")
loaded = cc.loadPointCloud("test.las", mode=cc.CC_SHIFT_MODE.XYZ, x=0.0, y=-2_500_000.0, z=0.0)
print("Shift:", loaded.getGlobalShift())
print("isShifted:", loaded.isShifted())
print("Points:", loaded.toNpArrayCopy())
Output:
=== Create cloud ===
Shift: (0.0, -2500000.0, 0.0)
isShifted: True
Points: [[0. 0. 0.]]
=== Save ===
=== Load ===
Shift: (0.0, -2500000.0, 0.0)
isShifted: True
Points: [[ 0. -4647483.5 0. ]] ## Should be [0,0,0]
LAS saving works when I load/save a cloud that already has an appropriate shift in the header.
Possibly related to the expanded options in CloudCompare that suggests can look at the LAS header instead of the set shift when saving:
https://www.cloudcompare.org/forum/viewtopic.php?p=30162#p30162
Appreciate any ideas or short-term fixes.
Activity
prascle commentedon Nov 5, 2025
Hi,
Sorry for the late reply.
I can reproduce the problem. Thanks for the link to the CloudCompare forum. I think you're right, I'll try with the extended settings. Unfortunately, I don't see any way to set these parameters without development: I'll try to release a fix as soon as possible...
Regards,
Paul
bthetford commentedon Jan 13, 2026
I think I'm experiencing the same problem. I do notice a difference in the output if I use
setGlobalShift()or not. If I have a float64 ndarraypointsthat looks like this:And I create a
ccPointCloudinstance and write it to las like this:Dragging the result into CloudCompare, this is what I see:
There's no "input" option, and the X and Y coords are the same and the numbers are strange (they should all be around 6782000 and -1994000). Viewing the file, it appears transformed in 3D.