Rerun

Migrating from 0.24 to 0.25

Removed deprecated --serve CLI argument removed-deprecated-serve-cli-argument

Use --web-viewer instead.

Removed the --drop-at-latency CLI argument removed-the-dropatlatency-cli-argument

This feature has been defunct for a while. A better replacement can be tracked in this issue.

Removed the -o CLI argument shorthand for --stdout in script_add_args removed-the-o-cli-argument-shorthand-for-stdout-in-scriptaddargs

Use --stdout directly instead.

Flush takes an optional timeout, and returns errors flush-takes-an-optional-timeout-and-returns-errors

When flushing a recording stream you can now give it a maximum time for how long it should block. The flush will block until either it completes, fails (e.g. because of connection loss), or the timeout is reached.

Previously this could only be configured for gRPC sinks, and it was configured once upon setting up the connection.

In the C++ and Python APIs, negative timeouts used to have special meaning. Now they are no longer permitted.

The Python flush calls now raises an error if the flushing did not complete successfully.

❗ Deprecations -deprecations

Python 3.9 python-39

Support for Python 3.9 is being deprecated. Python 3.9 is past end-of-life. See: https://devguide.python.org/versions/ In the next release, we will fully drop support and switch to Python 3.10 as the minimum supported version.

See an overview for supported python versions here.

archetype specification in AnyValues archetype-specification-in-anyvalues

Previously, logging two AnyValues with the same field name but different archetype name under the same entity would lead to an inconsistency where the viewer would disambiguate them, but not the dataframe API.

arbitrary_int = 10
example = AnyValues()
example.with_field(
    ComponentDescriptor("component_name", "archetype_name"), arbitrary_int
)
example.with_field(
    ComponentDescriptor("component_name", "different_archetype"), arbitrary_int+1
)
rr.log("/path", example)

In the viewer we would see two component_name entries under different archetypes but they would not be uniquely queryable.

from rerun.dataframe import load_recording
rec = load_recording("<path_to_logs_above>.rrd")
rec.view(index="log_time", contents="/path").select().schema
# Only shows one `component_name` component

To address that, we split this functionality into two utilities:

  • AnyValues, which has no archetype name
  • DynamicArchetype, which requires an archetype name

When using DynamicArchetype, the dataframe API will include the archetype the column names (similar to how built-in components are handled), which reduces the possibility for ambiguity. In the next release we will remove the ability to specify an archetype when creating AnyValues to finalize the transition.

arbitrary_int = 10
example = DynamicArchetype("archetype_name")
example.with_component_from_data(
    "component_name", arbitrary_int
)
another_example = DynamicArchetype("another_archetype")
another_example.with_field(
    "component_name", arbitrary_int+1
)
rr.log("/path", example)
rr.log("/path", another_example)
from rerun.dataframe import load_recording
rec = load_recording("<path_to_logs_above>.rrd")
rec.view(index="log_time", contents="/path").select().schema
# Only shows two `component_name` components deduplicated by archetype!
Privacy policy