Transfer missing typed values with a validity mask

Transfers an R vector as a branded JavaScript object containing contiguous values and validity typed arrays. The validity byte is zero for a missing value and one otherwise.

Usage

js_masked_typed_array(x)

Arguments

x

A logical, integer, or double vector without factor or date classes.

Value

A lightweight wrapper around the input vector.

Details

Logical values use a Uint8Array, integers use an Int32Array, and doubles use a Float64Array. Every representation uses a Uint8Array validity mask and exposes the original length as length.

R integer and logical NA values have validity zero and a placeholder value of zero. Double NA is treated the same way, while ordinary NaN remains a valid Float64 value. This preserves the R distinction between NA_real_ and NaN.

JavaScript can modify the contents of both typed arrays. Returning the branded object to R reconstructs the original R vector type using the current values and validity bytes. A nonzero validity byte is treated as valid. Removing the brand by constructing a plain object returns the two buffers separately under the ordinary object conversion rules.

The mask uses one byte per element. Use js_typed_array() when a mask is not required. Factors and dates are rejected because their class semantics cannot be represented by the numeric values and validity buffers alone.

Examples

ctx <- JSContext$new(profile = "bare")
ctx$source(code = paste(
  "function fillMissing(x) {",
  "  x.values[1] = 20;",
  "  x.validity[1] = 1;",
  "  return x;",
  "}"
))
values <- js_masked_typed_array(c(10L, NA_integer_, 30L))
ctx$call("fillMissing", values)
#> [1] 10 20 30