Integrates polytropos open records with the high-performance PostgreSQL driver hasql.

Architectural Overview & Rationale

Bypassing Tuple Limits & Intermediate Allocations

Conventional database libraries map SQL rows to Haskell data types via intermediate tuples (e.g. (a, b, c)). This imposes GHC's 62-column tuple limit and generates large numbers of short-lived heap objects.

recordRow builds a Row (Record r) that consumes one column per row field, in row order, and packs the decoded values straight into the contiguous SmallArray payload of the record. The column decoders are combined by the AllRowDecoder constraint walk in a single pass, appending one slot at a time. Rows of arbitrary arity (50, 100, 200+ columns) decode with zero tuple overhead.

Each column picks its value decoder from the Column class at that field's own type, allowing heterogeneous column rows.

When to Use This Module

• When executing high-throughput PostgreSQL queries with hasql where rows have many columns or where open records are used as query results.

{-# LANGUAGE DataKinds #-}
{-# LANGUAGE TypeApplications #-}

import qualified Hasql.Decoders as HD
import Polytropos
import Polytropos.Hasql

type UserRow = Record '["id" ':= Int64, "age" ':= Int32]

userDecoder :: HD.Row UserRow
userDecoder = recordRow @UserRow