filter

Keeps the rows a predicate accepts, and drops the rest. It sits in .flow() beside derive, which can do the same thing with an arbitrary callback — filter just says what it does.

gf.chart(seafood, { axes: true })
  .flow(
    gf.filter((row) => row.species === "Salmon"),
    gf.spread({ by: "lake", dir: "x" })
  )
  .mark(gf.rect({ h: "count", fill: "steelblue" }))
  .render(root, { w: 400, h: 250 });

Signature

filter(pred);

Parameters

Parameter	Type	Description
pred	(row: T) => boolean	Kept when it returns true

The predicate can be written by hand, or built from a field expression — field(name).between(lo, hi) returns exactly this shape. Its bounds are plain numbers, and it throws if the expression carries pipeline ops (field("x").bin(10).between(...)), which would otherwise test the raw field: a predicate is not a value slot.

Examples

// A plain predicate
.flow(filter((d) => d.year === 2020), spread({ by: "category", dir: "x" }))

// A value window, with polars' `closed` ends
.flow(filter(field("day").between(100, 120, { closed: "right" })))

// A window that follows a clock. `between`'s bounds are plain numbers, so the
// clock read goes in the predicate: the rows kept change as the timer advances
// (see the reactivity guide). `between(v, lo, hi)` is the bare value test behind
// `field(name).between(...)`.
const day = timer({ domain: [1, 365], step: 1, duration: 10000 });
.flow(filter((d) => between(d.day, day() - 20, day(), { closed: "right" })))

// A window on a CYCLIC field. Day-of-year wraps, so the window is on the
// distance back from the playhead, modulo the year — which keeps the trail 20
// days long at the loop boundary instead of letting it collapse to the few
// days of the new year so far.
.flow(filter((d) => between((day() - d.day + 365) % 365, 0, 20, { closed: "left" })))

Domains are inferred from what survives

filter runs before domain inference, so the scales see only the rows it kept. In an animated chart whose filter follows a clock, that means the axes rescale every frame: with one day's rows in hand, the x domain is that day's extent and nothing else. Two ways to hold the scales still:

• Give the chart an explicit domain. Under geo with a lon/lat window, the window replaces the data extent entirely, so filtering cannot move the map — which is why the bird migration examples stay put.
• Keep the full data in the chart and vary something paint-only instead, e.g. opacity: live((d) => (d.day === day() ? 1 : 0)).

Python

filter is JavaScript only. Its predicate is a live JavaScript function with nothing to put on the wire, the same reason derive's callback does not cross the bridge.