zer-adapters

Type adapters for the zer entity-resolution library, bridging Polars DataFrames and Arrow RecordBatches to zer-core Records without a string round-trip.

Documentation: docs.zal-analytics.ch
Website: www.zal-analytics.ch
Support & feedback: info@zal-analytics.ch

Feature flags

Flag	Adds
`polars`	`PolarsIngest` extension trait for Polars `DataFrame`
`arrow`	`ArrowIngest` extension trait for Arrow `RecordBatch`

Enable only the features you need to keep compile times low.

Usage

use zer_adapters::PolarsIngest;

// Convert a Polars DataFrame into zer Records directly.
let records = df.into_zer_records(&schema)?;

This crate also provides BenchResultWriter, AccuracyMetrics, and timestamp utilities used by zer-bench (internal tooling, not part of the public API).

Breaking changes

v1.1

into_records signature changed, DatasetConfig replaces id_start

Both PolarsIngest::into_records and ArrowIngest::into_records now take &DatasetConfig instead of an integer id_start. DatasetConfig names the source label and the column to use as each record's natural key; IDs are derived via FNV-1a(source:key).

// v1.0
let records_a = df_a.into_records(1);
let records_b = df_b.into_records(n_a + 1);  // manual offset to avoid collisions

// v1.1
let records_a = df_a.into_records(&DatasetConfig::new("A", "bsn"));
let records_b = df_b.into_records(&DatasetConfig::new("B", "record_id"));
// no offset needed, source label is part of the hash

DatasetConfig is a new public struct in this crate.

License