Schedules and spans¶

Timeline.intervals draws a bar per row from its start to its end, on lanes: jobs on machines, shifts of people, the time a request spends in each service. gantt() is this with a lane per task.

Bars that overlap¶

Bars of one lane that overlap in time are stacked in sub-rows, the first free row first, so none hides another:

import numpy as np
from anytimeline import Timeline

rng = np.random.default_rng(5)
n = 160
jobs = {
    "machine": rng.choice([f"press {i}" for i in range(1, 7)], n),
    "start": np.datetime64("2026-10-12T06:00", "m")
    + np.sort(rng.integers(0, 3 * 24 * 60, n)).astype("timedelta64[m]"),
    "minutes": rng.integers(20, 240, n),
    "product": rng.choice(["A", "B", "C", "D"], n),
}
plot = Timeline(title="Jobs per machine", height=380)
plot.intervals(jobs, end=None, duration="minutes", duration_unit="min", lane="machine",
               color="product", text="{product}")
print(plot._embed_html())

overlap="overlay" draws them on top of each other instead; height= sets the bars' height as a fraction of the lane (default 0.7).

Start and end, or duration¶

• start= and end=: two columns of times of the same kind.
• end=None, duration=: a column of durations (timedelta), or of numbers with duration_unit= ("d", "h", "min", "s", "ms", "us", "ns"). With plain-number times, durations are in the same numbers.

A row whose end equals its start is a milestone, drawn as a diamond. A row without an end is open: it runs to the plot's last time and fades out, like a job that is still running. A row that ends before it starts is an error that names the first such row.

Bars and events together¶

Layers share the lanes and the time axis, so bars and events line up: here the time each request spends in a service, with the request's events on top.

import polars as pl
from anytimeline.synthetic import service_log

log = pl.DataFrame(service_log(8)).sort("time")
stays = (
    log.with_columns(end=pl.col("time").shift(-1).over("request"))
    .drop_nulls("end")
    .rename({"time": "start"})
)
plot = Timeline(lane_order=["gateway", "auth", "api", "cache", "db"], height=340)
plot.intervals(stays, lane="service", color="request", height=0.45, opacity=0.5, name="spans")
plot.events(log, time="time", lane="service", symbol="event", order="seq", name="events")
one = log.filter(pl.col("request") == "r0002")["time"]
plot.zoom_to(one.min(), one.max())
print(plot._embed_html())

Everything else¶

Colors (color=, palette=, colormap=, clim=, opacity=), text=, tooltip= and selection work as for events; the selection gives rows per layer: plot.selection == {"intervals": [3, 17]}. On the Order scale each bar's start and end take a place in the order of all events and bars.