What an event chart is for¶

An event chart draws instants: one mark per event at the moment it happened, without a duration. It answers three questions at a glance: what happened (color, symbol), when (position along time), and in what order (which mark comes first).

The idea is old and has many names. In survival analysis an event chart draws a line per patient with a symbol for each event (Lee, Hess and Dubin, The American Statistician, 2000). Process mining calls it a dotted chart: a row per case, a dot per activity. Neuroscience calls it a raster plot: a row per neuron, a tick per spike. Log viewers and trace viewers draw the same picture with a lane per host, thread or service.

Lanes¶

Lanes are rows of a category: a service, a host, a patient, a request. They turn a cloud of events into rows you can compare: when did this lane get busy, which lane failed first? Their order matters: put lanes in the order work flows through them (lane_order=), or by when they started (sort="first").

Time or order¶

On the Time scale, distance is elapsed time: gaps and bursts are visible, but events a microsecond apart sit on top of each other. On the Order scale every event gets the same width, one after another: the order of a burst is readable, at the cost of the real gaps, which the axis marks with a break and the time skipped. The two answer different questions, so the plot has a switch rather than a choice made for you.

Timelines researchers call these the chronological and sequential scales; the break marks make the Order scale sequential with interim durations (Brehmer et al., Timelines Revisited, 2017).

Routes¶

When events belong to a group that moves between lanes, such as a request passing through services, a line through the group's events in order shows its path. On the Order scale, crossing lines show where one request overtook another.