Send dataframes to a recording

If your data is already in a table, you can log the whole table in a single call instead of logging it row by row or column by column. In Python, rr.send_dataframe logs a table into the active recording, so it's saved with the rest of your data and can be shown next to your other views.

It accepts:

a pyarrow Table, RecordBatch, or RecordBatchReader,
a datafusion DataFrame,
any other object that implements the Arrow C stream interface (__arrow_c_stream__). Recent versions of pandas and polars implement it, so you can pass their dataframes directly. With older versions, convert first, e.g. with pa.Table.from_pandas(df) or df.to_arrow() (polars).

Pass the timeline columns as index and the desired entity path for the remaining columns as entity_path:

"""Log an entire table, in a single operation."""

from __future__ import annotations

import numpy as np
import pandas as pd

import rerun as rr

rr.init("rerun_example_send_dataframe", spawn=True)

steps = np.arange(0, 64)
df = pd.DataFrame({
    "step": steps,
    "sin": np.sin(steps / 10.0),
    "cos": np.cos(steps / 10.0),
})

rr.send_dataframe(df, index="step", entity_path="/trig")

Each remaining column becomes a component of that entity, which you can display in a DataframeView.

Like send_columns, this ignores any timeline values set with rr.set_time and bypasses the micro-batcher: only the timelines in the table are added. To control component types or spread columns across several entities, see Chunks.

Reference

Python `send_dataframe`