Migration guides¶
From laspy 1.7.x to laspy 2.0.0¶
laspy 2.0 is essentially a complete overhaul of the code base so you will probably have a few changes to make to the parts of your code that uses laspy 1.7.
However there should not be that many, and should hopefully be worth.
- The benefits of laspy 2.0 are:
Better LAZ support: reading and writing of LAZ 1.1 to 1.4 (See the Installation section)
Support for Chunked / Streaming reading and writing of LAS/LAZ files.
Support for reading data coming from other sources than files on disk (bytes or file-objects)
The biggest changes between 1.7 and 2.0 are how files are read and written.
laspy 1.7 had the concept of laspy.File with an open mode.
laspy 2.0 does not have the laspy.File class anymore but a LasData
class instead which provide access to the header, vlrs and fields/dimensions.
The get_* and set_* (eg get_classification) of laspy.File are not available in the new LasData.
The following sections should hopefully get you started
Reading a file¶
import laspy
# Opening a file in laspy 1.7
file = laspy.file.File("somepath.las", mode ="r")
# Reading a file in laspy 2.0
las = laspy.read("somepath.las")
Accessing the point fields / dimensions¶
# accessing a field in laspy 1.7:
classification = file.classification
# accessing a field in laspy 2.0 (names from 1.7 should be compatible with 2.0):
classification = las.classification
Writing¶
# laspy 1.7:
file.pt_src_id[:] = 2
file.close()
# laspy 2.0
las.pt_src_id[:] = 2
las.write("somepath.laz")
Creating a file¶
import laspy
# laspy 1.7
new_file = laspy.file.File("new_path.las", header=file.header, mode="w")
new_file.X = ...
new_file.Y = ...
...
new_file.close()
# laspy 2.0
new_las = laspy.LasData(las.header)
new_las.X = ...
new_las.Y = ...
...
new_las.write("new_las.las")
# if you do not have an existing header:
new_las = laspy.create(file_version="1.2", point_format=3)
new_las.X = ...
new_las.Y = ...
...
new_las.write("new_las.las")
# or
new_header = laspy.LasHeader(version="1.2", point_format=3)
new_las = laspy.LasData(las.header)
new_las.X = ...
new_las.Y = ...
...
new_las.write("new_las.las")
Header change¶
The Header (LasHeader) class was modernized from laspy 1.7 to laspy 2.0,
a few of the field names in the new header class do not have the same name.
This list is non exhaustive, and for the items in this list the backward compatibility could be kept (meaning the 1.7 name is still usable)
1.7 name |
2.0 name |
|---|---|
max |
maxs |
min |
mins |
scale |
scales |
offset |
offsets |
filesource_id |
file_source_id |
major_version |
version.major |
minor_version |
version.minor |
system_id |
system_identifier |
date |
creation_date |
point_return_count |
number_of_points_by_return |
software_id |
generating_software |
From pylas 0.4.x to laspy 2.0.0¶
laspy 2.0 is essentially pylas, so the core of the library is the same.
Changes in LAZ backend¶
With laspy 2.0.0, the lazperf backend support was dropped, and the laszip backend changed from using the laszip executable to using laszip python bindings.
If you used lazperf or relied on the laszip executable you’ll have to choose between the available backends. (see Installation section).
Changes in bit fields¶
Some fields in LAS are ‘bit fields’.
with laspy 0.4.x, there was a inconsistency between ‘normal’ fields and ‘bit’ fields, when getting a bit field, laspy returned a copy of the values in a new numpy array whereas when getting a normal field, the array you got acted as a ‘view’ on the real array where the values where stored.
That meant that modifying the values of the array you got from a bit field would no propagate to the real array.
import laspy
import numpy as np
las = laspy.read("tests/data/simple.las")
# return number is a bit field
print(las.return_number)
# array([1, 1, 1, ..., 1, 1, 1], dtype=uint8)
ascending_order = np.argsort(las.return_number)[::-1]
print(las.return_number[ascending_order])
# array([4, 4, 4, ..., 1, 1, 1], dtype=uint8)
las.return_number[:] = las.return_number[ascending_order]
print(las.return_number)
# array([1, 1, 1, ..., 1, 1, 1], dtype=uint8) # bif oof
las.return_number[0] = 7
print(las.return_number)
# array([1, 1, 1, ..., 1, 1, 1], dtype=uint8) # again value not updated
# To actually update you have to do
las.return_number = las.return_number[ascending_order]
print(las.return_number)
# array([4, 4, 4, ..., 1, 1, 1], dtype=uint8)
rn = las.return_number[ascending_order]
rn[0] = 7
las.return_number = rn
print(las.return_number)
# array([7, 4, 4, ..., 1, 1, 1], dtype=uint8)
In order to try to solve this inconsistency, laspy >= 0.5.0
introduced the SubFieldView class that is meant to propagate
modifications to the real array, and tries to act like a real numpy array.
import laspy
import numpy as np
las = laspy.read("tests/data/simple.las")
print(las.return_number)
# <SubFieldView([1 1 1 ... 1 1 1])>
ascending_order = np.argsort(las.return_number)[::-1]
las.return_number[:] = las.return_number[ascending_order]
print(las.return_number)
# <SubFieldView([4 4 4 ... 1 1 1])>
las.return_number[0] = 7
print(las.return_number)
# <SubFieldView([7 4 4 ... 1 1 1])>
It may be possible that some operation on SubFieldView fail, in that case it is easy to copy them to numpy arrays:
import laspy
import numpy as np
las = laspy.read("tests/data/simple.las")
print(las.return_number)
# <SubFieldView([1 1 1 ... 1 1 1])>
array = np.array(las.return_number)
# array([1, 1, 1, ..., 1, 1, 1], dtype=uint8)
The logic is also the same for ‘Scaled dimensions’ such as x, y, z and scaled extra bytes, where a ScaledArrayView class has been introduced
import laspy
import numpy as np
las = laspy.read("tests/data/simple.las")
print(las.x)
# <ScaledArrayView([637012.24 636896.33 636784.74 ... 637501.67 637433.27 637342.85])>>
# ScaledArray view should behave as much as possible as a numpy array
# However if something breaks in your code when upgrading, and / or
# you need a true numpy array you can get one by doing
array = np.array(las.x)
# array([637012.24, 636896.33, 636784.74, ..., 637501.67, 637433.27,
# 637342.85])
Changes in extra bytes creation¶
The API to create extra bytes changed slightly, now the parameters needed
(and the optional ones) are coupled into ExtraBytesParams
Other changes¶
The points attribute of as LasData used to return a numpy array
it now returns a PackedPointRecord to get the same array as before,
use the array property of the point record.
# laspy <= 0.4.3
las = laspy.read("somefile.las")
array = las.points
# laspy 1.0.0
las = laspy.read("somefile.las")
array = las.points.array