Migrating from 0.21 to 0.22
Previously deprecated DisconnectedSpace
archetype/component got now removed previously-deprecated-disconnectedspace-archetypecomponent-got-now-removed
The deprecated DisconnectedSpace
archetype and DisconnectedSpace
component have been removed.
To achieve the same effect, you can log any of the following "invalid" transforms:
- zeroed 3x3 matrix
- zero scale
- zeroed quaternion
- zero axis on axis-angle rotation
Previously, the DisconnectedSpace
archetype played a double role by governing view spawn heuristics & being used as a transform placeholder.
This led to a lot of complexity and often broke or caused confusion (see https://github.com/rerun-io/rerun/issues/6817, https://github.com/rerun-io/rerun/issues/4465, https://github.com/rerun-io/rerun/issues/4221).
By now, explicit blueprints offer a better way to express which views should be spawned and what content they should query.
(you can learn more about blueprints here).
Removed num_instances
keyword argument to rr.log_components()
removed-numinstances-keyword-argument-to-rrlogcomponents
For historical reasons, the rr.log_components()
function of the Python SDK accepts an optional, keyword-only argument num_instances
.
It was no longer used for several releases, so we removed it.
Note: although rr.log_components()
is technically a public API, it is undocumented, and we discourage using it.
For logging custom components, use rr.AnyValue
and rr.AnyBatchValue
.
Rust's ViewCoordinates
archetype now has static methods instead of constants rusts-viewcoordinates-archetype-now-has-static-methods-instead-of-constants
As part of the switch to "eager archetype serialization" (serialization of archetype components now occurs at time of archetype instantiation rather than logging), we can no longer offer constants for the ViewCoordinates
archetype like ViewCoordinates::RUB
.
Instead, there's now methods with the same name, i.e. ViewCoordinates::RUB()
.
Rust's Tensor
archetype can no longer access tensor data as ndarray
view directly rusts-tensor-archetype-can-no-longer-access-tensor-data-as-ndarray-view-directly
As part of the switch to "eager archetype serialization" (serialization of archetype components now occurs at time of archetype instantiation rather than logging), we can no longer offer exposing the Tensor
archetype as ndarray::ArrayView
directly.
However, it is still possible to do so with the TensorData
component.