Number Inference

Figma has no native number type. Numeric values like maxCharacterCount, minRows, and tabIndex are stored as TEXT component properties with string content. Without intervention, these appear in your spec as StringProp — technically faithful to Figma but semantically wrong. The inferNumberProps feature detects numeric text values and emits them as NumberProp instead.

The Problem

A component spec that types maxCharacterCount: "24" as a string forces every consumer to parse it themselves. Code generators produce string parameters, documentation shows string types, and linters can’t validate numeric ranges. The type information is lost at the source.

What It Does

When inferNumberProps is enabled, the processing engine examines TEXT code-only props and applies a deterministic inference guard. If a prop’s default value (and all example values) pass the guard, the prop is emitted as NumberProp (type: 'number') with numeric values instead of strings.

The Inference Guard

A value is treated as numeric when all of these conditions hold:

Not empty ("") or a bare minus sign ("-")
No leading zero before another digit — rejects "007", "0800", "01" (likely identifiers or formatted codes) but allows "0" and "0.5"
Parses to a finite number (not NaN or Infinity)

Before (without `inferNumberProps`)

maxCharacterCount:
  type: string
  default: "24"

After (with `inferNumberProps: true`)

maxCharacterCount:
  type: number
  default: 24

The value moves from a quoted string to a native number. Downstream consumers get the correct type without additional parsing.

Known Edge Cases

The inference guard is a heuristic. Some values pass all conditions but may not be intended as numbers:

Value	Inferred?	Why
`"24"`	Yes	Standard integer
`"3.14"`	Yes	Standard decimal
`"-42"`	Yes	Negative number
`"0"`, `"0.5"`	Yes	Zero and zero-prefixed decimals
`"007"`	No	Leading zero — likely an identifier
`"0800"`	No	Leading zero — likely a formatted code
`""`	No	Empty string
`"90210"`	Yes	Passes all guards, but could be a zip code
`"1.0"`	Yes	Passes all guards, but could be a version string

The last two rows are why the feature is opt-in. Enabling it means accepting responsibility for occasional false positives where a numeric-looking string is actually a semantic identifier.

When to Use It

Enable inferNumberProps when your Figma library uses TEXT code-only props for genuinely numeric values — counts, indices, thresholds, dimensions. This is most valuable when:

Code generators consume your specs — they produce correctly typed parameters instead of strings
Documentation is auto-generated — readers see number types, not ambiguous strings
Validation tooling checks ranges — numeric min/max checks only work on number types

Leave it disabled if your library uses TEXT props for values that look numeric but aren’t (version strings, formatted codes, identifiers with leading zeros).

Configuration

Add inferNumberProps: true under model.processing in your config file:

model:
  processing:
    inferNumberProps: true

Default: false (absent). Existing specs are unaffected until you opt in.

Design Rationale

Number inference follows the same pattern as boolean inference, which already exists in the processing engine. A VARIANT property with exactly "true" / "false" values is emitted as BooleanProp — that inference is deterministic and always on. Number inference from TEXT props is probabilistic (the guard is a heuristic), so it requires an explicit opt-in flag.

NumberProp is a first-class member of the AnyProp union with its own type: 'number' discriminant, consistent with the discriminated-union pattern used by all other prop types.