Skip to contents

Stable SipHash-1-3 of an R object

Source: R/hashing.R
st_hash_obj.Rd

Computes a stable hash of an R object by serializing it with base::serialize(version = 3) and hashing the resulting bytes via secretbase::siphash13(). The hash is stable across R sessions (given the same R version and object structure) and suitable for change detection.

Usage

st_hash_obj(x)

Arguments

x

Any R object (data.frame, data.table, list, vector, etc.)

Value

Lowercase hex string (16 hex characters) from siphash13().

Attribute Normalization

Before hashing, this function normalizes the order of object attributes to ensure consistent hashes even when operations (like collapse::rowbind() + collapse::funique()) leave attributes in different orders.

The normalization reorders attributes to a canonical form:

  1. Priority attributes: names, row.names, class, .internal.selfref

  2. Other attributes: alphabetically sorted

This ensures that logically identical objects produce identical hashes regardless of their attribute creation history.

Why This Matters

Without normalization, two data.frames that are identical() can produce different hashes if their internal attributes are in different orders. This breaks change detection in stamp, causing false positives where objects are incorrectly flagged as changed.

Performance

  • For small to medium objects: negligible overhead

  • For large objects: creates a shallow copy (data is referenced, not duplicated)

  • The normalization cost is typically < 1% of total hashing time

Examples

if (FALSE) { # \dontrun{
# Two ways to create the "same" data
dt_a <- data.table(x = 1:5)
dt_b <- data.table(x = 1:5) |> collapse::rowbind(data.table(x = 1:5)) |> collapse::funique()

# They're identical in content
identical(dt_a, dt_b)  # TRUE

# And now they hash the same too!
st_hash_obj(dt_a) == st_hash_obj(dt_b)  # TRUE
} # }