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 function — filter just says what it does.

from gofish import chart, circle, field, filter, scatter

chart(penguins, axes=True).flow(
    filter(field("Body Mass (g)").between(3500, 4500, closed="left")),
    scatter(x="Flipper Length (mm)", y="Beak Length (mm)"),
).mark(circle(r=3, fill="Species"))

Signature

filter(predicate) -> Operator

Parameters

predicate is a field predicate or a Python function of one row.

A field predicate comes from field(name).between(lo, hi, closed=None). Its bounds are plain numbers, and it raises if the expression carries pipeline ops (field("x").bin(Bin.step(10)).between(...)), which would otherwise test the raw field: a predicate is not a value slot. It is data, so it crosses to JavaScript as these fields:

Keep the rows a field predicate accepts (filter(field(name).between(lo, hi, { closed }))). A filter over a hand-written predicate has no wire form and serializes as an opaque derive.

Option	Type	Default	Description
predicate	FieldPredicate		Required. The field predicate field(name).between(lo, hi, { closed }) builds: { field, between: [lo, hi], closed? }.

A function, such as lambda row: row["year"] == 2020, runs in your kernel. The filter is then a derive, so the chart calls back into Python every time it renders.

Returns an Operator for use inside .flow().

Examples

# A value window, with polars' `closed` ends
.flow(filter(field("day").between(100, 120, closed="right")))

# A plain predicate
.flow(filter(lambda row: row["year"] == 2020), spread(by="category", dir="x"))

Domains are inferred from what survives

filter runs before domain inference, so the scales see only the rows it kept. To hold a scale to the full data, give the chart an explicit domain.