Vega-Lite Filter Transform

Use filter inside transform to keep only rows that match a condition.

{
  "transform": [
    {"filter": ...}
  ]
}

This page covers the most common vega lite filter and altair transform filter patterns.

Filter Syntax (Quick Reference)

You can also use or and not inside logical predicates.

Example 1: Basic Vega Filter Expression

Keep only cars with MPG above 30.

{
  "data": {"url": "data/cars.json"},
  "transform": [
    {"filter": "datum.Miles_per_Gallon > 30"}
  ],
  "mark": "point",
  "encoding": {
    "x": {"field": "Horsepower", "type": "quantitative"},
    "y": {"field": "Miles_per_Gallon", "type": "quantitative"}
  }
}

Example 2: Transform Filter with Predicate Object

This version avoids raw expression strings and is easier to maintain.

{
  "data": {"url": "data/cars.json"},
  "transform": [
    {
      "filter": {
        "and": [
          {"field": "Miles_per_Gallon", "gte": 30},
          {"field": "Weight_in_lbs", "lt": 3000}
        ]
      }
    }
  ],
  "mark": "point",
  "encoding": {
    "x": {"field": "Horsepower", "type": "quantitative"},
    "y": {"field": "Miles_per_Gallon", "type": "quantitative"},
    "color": {"field": "Origin", "type": "nominal"}
  }
}

Example 3: Category Filter (oneOf)

Use oneOf when you need an explicit allow-list.

{
  "data": {"url": "data/cars.json"},
  "transform": [
    {"filter": {"field": "Origin", "oneOf": ["USA", "Japan"]}}
  ],
  "mark": "bar",
  "encoding": {
    "x": {"field": "Origin", "type": "nominal"},
    "y": {"aggregate": "count", "type": "quantitative"}
  }
}

Altair transform_filter Patterns

# 1) Expression string
chart.transform_filter("datum.Origin == 'USA'")
 
# 2) Datum expression
chart.transform_filter(
  (alt.datum.Horsepower > 120) &
  (alt.datum.Miles_per_Gallon > 20)
)
 
# 3) Predicate object
chart.transform_filter(
  alt.FieldOneOfPredicate(field="Origin", oneOf=["USA", "Japan"])
)

Common Mistakes

1. Filtering with a missing field (for example Price in cars.json) returns empty or unexpected results.
2. Not handling null values for numeric fields. Use {"valid": true} when needed.
3. Mixing string and number comparisons ("30" vs 30) in expressions.

FAQ

What is the difference between Vega filter and Vega-Lite filter?

The idea is the same: remove rows before drawing marks. Vega-Lite provides a concise declarative syntax in the transform array.

How do I write transform filter for multiple conditions?

Use either an expression string (&&, ||) or a logical predicate object with and / or / not.

Why does my chart show no points after filtering?

Check field names, data types, and null values first. In this page's examples, use Weight_in_lbs instead of non-existent fields like Price.