Update relation_id_from_string to use the 128-bits of the Sha256 has rather than

just 64-bits.

Author: nystrom

meta/src/meta/templates/parser.go.template

@@ -10,6 +10,7 @@ package lqp
 
 import (
 	"crypto/sha256"
+	"encoding/binary"
 	"fmt"
 	"math"
 	"math/big"
@@ -362,16 +363,9 @@ func (p *Parser) startFragment(fragmentID *pb.FragmentId) *pb.FragmentId {{
 }}
 
 func (p *Parser) relationIdFromString(name string) *pb.RelationId {{
-	// Create RelationId from string hash (matching Python implementation)
-	// Python uses: int(hashlib.sha256(name.encode()).hexdigest()[:16], 16)
-	// This takes only first 8 bytes (16 hex chars) as id_low, id_high is always 0
-	// Python interprets the hex as big-endian, so we read bytes in big-endian order
 	hash := sha256.Sum256([]byte(name))
-	var low uint64
-	for i := 0; i < 8; i++ {{
-		low = (low << 8) | uint64(hash[i])
-	}}
-	high := uint64(0)
+	low := binary.LittleEndian.Uint64(hash[:8])
+	high := binary.LittleEndian.Uint64(hash[8:16])
 	relationId := &pb.RelationId{{IdLow: low, IdHigh: high}}
 
 	// Store the mapping for the current fragment if we're inside one

meta/src/meta/templates/parser.jl.template

@@ -228,8 +228,9 @@ end
 function relation_id_from_string(parser::ParserState, name::String)
     # Create RelationId from string and track mapping for debug info
     hash_bytes = sha256(name)
-    id_low = Base.parse(UInt64, bytes2hex(hash_bytes[1:8]), base=16)
-    id_high = UInt64(0)
+    # Match the convention in backir-to-lqp.jl
+    id_low = reinterpret(UInt64, hash_bytes[1:8])[1]
+    id_high = reinterpret(UInt64, hash_bytes[9:16])[1]
     relation_id = Proto.RelationId(id_low, id_high)
 
     # Store the mapping for the current fragment if we're inside one

meta/src/meta/templates/parser.py.template

@@ -213,8 +213,9 @@ class Parser:
 
     def relation_id_from_string(self, name: str) -> Any:
         """Create RelationId from string and track mapping for debug info."""
-        id_low = int(hashlib.sha256(name.encode()).hexdigest()[:16], 16)
-        id_high = 0
+        hash_bytes = hashlib.sha256(name.encode()).digest()
+        id_low = int.from_bytes(hash_bytes[:8], byteorder='little')
+        id_high = int.from_bytes(hash_bytes[8:16], byteorder='little')
         relation_id = logic_pb2.RelationId(id_low=id_low, id_high=id_high)
 
         # Store the mapping for the current fragment if we're inside one

sdks/go/src/parser.go

@@ -10,6 +10,7 @@ package lqp
 
 import (
 	"crypto/sha256"
+	"encoding/binary"
 	"fmt"
 	"math"
 	"math/big"
@@ -388,16 +389,9 @@ func (p *Parser) startFragment(fragmentID *pb.FragmentId) *pb.FragmentId {
 }
 
 func (p *Parser) relationIdFromString(name string) *pb.RelationId {
-	// Create RelationId from string hash (matching Python implementation)
-	// Python uses: int(hashlib.sha256(name.encode()).hexdigest()[:16], 16)
-	// This takes only first 8 bytes (16 hex chars) as id_low, id_high is always 0
-	// Python interprets the hex as big-endian, so we read bytes in big-endian order
 	hash := sha256.Sum256([]byte(name))
-	var low uint64
-	for i := 0; i < 8; i++ {
-		low = (low << 8) | uint64(hash[i])
-	}
-	high := uint64(0)
+	low := binary.LittleEndian.Uint64(hash[:8])
+	high := binary.LittleEndian.Uint64(hash[8:16])
 	relationId := &pb.RelationId{IdLow: low, IdHigh: high}
 
 	// Store the mapping for the current fragment if we're inside one

sdks/julia/LogicalQueryProtocol.jl/src/parser.jl

@@ -254,8 +254,9 @@ end
 function relation_id_from_string(parser::ParserState, name::String)
     # Create RelationId from string and track mapping for debug info
     hash_bytes = sha256(name)
-    id_low = Base.parse(UInt64, bytes2hex(hash_bytes[1:8]), base=16)
-    id_high = UInt64(0)
+    # Match the convention in backir-to-lqp.jl
+    id_low = reinterpret(UInt64, hash_bytes[1:8])[1]
+    id_high = reinterpret(UInt64, hash_bytes[9:16])[1]
     relation_id = Proto.RelationId(id_low, id_high)
 
     # Store the mapping for the current fragment if we're inside one

sdks/python/src/lqp/gen/parser.py

@@ -241,8 +241,9 @@ def start_fragment(
 
     def relation_id_from_string(self, name: str) -> Any:
         """Create RelationId from string and track mapping for debug info."""
-        id_low = int(hashlib.sha256(name.encode()).hexdigest()[:16], 16)
-        id_high = 0
+        hash_bytes = hashlib.sha256(name.encode()).digest()
+        id_low = int.from_bytes(hash_bytes[:8], byteorder='little')
+        id_high = int.from_bytes(hash_bytes[8:16], byteorder='little')
         relation_id = logic_pb2.RelationId(id_low=id_low, id_high=id_high)
 
         # Store the mapping for the current fragment if we're inside one