Deprecate <?> operator, remove post_parse_analyze_hook and `Executo…

…rStart_hook` hooks and related codes.

Update tests to not use <?> operator

README.md

@@ -106,19 +106,9 @@ FROM small_world ORDER BY vector <-> ARRAY[0,0,0] LIMIT 1;
 
 ### A note on operators and operator classes
 
-Lantern supports several distance functions in the index and it has 2 modes for operators:
+Lantern supports several distance functions in the index
 
-1. `lantern.pgvector_compat=TRUE` (default)
-   In this mode there are 3 operators available `<->` (l2sq), `<=>` (cosine), `<+>` (hamming).
-
-   Note that in this mode, you need to use right operator in order to trigger an index scan.
-
-2. `lantern.pgvector_compat=FALSE`
-   In this mode you only need to specify the distance function used for a column at index creation time. Lantern will automatically infer the distance function to use for search so you always use `<?>` operator in search queries.
-
-   Note that in this mode, the operator `<?>` is intended exclusively for use with index lookups. If you expect to not use the index in a query, use the distance function directly (e.g. `l2sq_dist(v1, v2)`)
-
-> To switch between modes set `lantern.pgvector_compat` variable to `TRUE` or `FALSE`.
+There are 3 operators available `<->` (l2sq), `<=>` (cosine), `<+>` (hamming).
 
 There are four defined operator classes that can be employed during index creation:
 

ci/scripts/run-tests-linux.sh

@@ -20,17 +20,6 @@ function run_pgvector_tests(){
     pushd /tmp/pgvector
       # Add lantern to load-extension in pgregress
       sed -i '/REGRESS_OPTS \=/ s/$/ --load-extension lantern/'  Makefile
-
-      # Set pgvector_compat flag in test files
-      for file in ./test/sql/*; do
-        echo 'SET lantern.pgvector_compat=TRUE;' | cat - $file > temp && mv temp $file
-      done
-
-      # Set pgvector_compat flag in result files
-      for file in ./test/expected/*.out; do
-        echo 'SET lantern.pgvector_compat=TRUE;' | cat - $file > temp && mv temp $file
-      done
-
       # Run tests
       make installcheck
     popd

lantern_hnsw/scripts/integration_tests.py

@@ -46,7 +46,6 @@ def primary():
     node.init()
     node.append_conf("enable_seqscan = off")
     node.append_conf("maintenance_work_mem = '1GB'")
-    node.append_conf("lantern.pgvector_compat=FALSE")
     node.append_conf("checkpoint_timeout = '100min'")
     node.append_conf("min_wal_size = '1GB'")
     node.append_conf("checkpoint_completion_target = '0.9'")
@@ -158,7 +157,6 @@ def generic_vector_query(
 
     dist_with_function = f"{distance_metric}_dist(v, ({query_vector}))"
     dist_with_concrete_op = f"v {DIST_OPS[distance_metric]} ({query_vector})"
-    dist_with_generic_op = f"v <?> ({query_vector})"
 
     query_generator = (
         lambda order_by: f"""
@@ -173,8 +171,6 @@ def generic_vector_query(
         return query_generator(dist_with_function)
     elif kind == "concrete":
         return query_generator(dist_with_concrete_op)
-    elif kind == "generic":
-        return query_generator(dist_with_generic_op)
 
 
 @pytest.mark.parametrize("distance_metric", ["l2sq", "cos"], scope="session")
@@ -197,9 +193,6 @@ def test_selects(db, setup_copy_table_with_index, distance_metric, quant_bits, r
         concrete_op_query = generic_vector_query(
             table_name, distance_metric, "concrete", query_vector_id=q_vec_id
         )
-        generic_op_query = generic_vector_query(
-            table_name, distance_metric, "generic", query_vector_id=q_vec_id
-        )
 
         exact_explain_query = f"EXPLAIN {exact_query}"
         exact_plan = primary.execute("testdb", exact_explain_query)
@@ -214,7 +207,7 @@ def test_selects(db, setup_copy_table_with_index, distance_metric, quant_bits, r
             q_vec_id == exact_res[0][0]
         ), "First result in exact query result should be the query vector"
 
-        for query in [generic_op_query, concrete_op_query]:
+        for query in [concrete_op_query]:
             explain_query = f"EXPLAIN {query}"
             plan = primary.execute("testdb", explain_query)
             assert f"Index Scan using idx_{table_name}" in str(
@@ -348,9 +341,6 @@ def test_inserts(setup_copy_table_with_index, distance_metric, quant_bits, reque
         concrete_op_query = generic_vector_query(
             table_name, distance_metric, "concrete", query_vector_id=q_vec_id
         )
-        generic_op_query = generic_vector_query(
-            table_name, distance_metric, "generic", query_vector_id=q_vec_id
-        )
 
         exact_explain_query = f"EXPLAIN {exact_query}"
         for db in [primary, replica]:
@@ -367,7 +357,7 @@ def test_inserts(setup_copy_table_with_index, distance_metric, quant_bits, reque
                 exact_res[0][0] in inserted_vector_orig_ids[q_vec_id]
             ), "First result in exact query result should be the query vector"
 
-            for query in [generic_op_query, concrete_op_query]:
+            for query in [concrete_op_query]:
                 explain_query = f"EXPLAIN {query}"
                 plan = db.execute("testdb", explain_query)
                 assert f"Index Scan using idx_{table_name}" in str(

lantern_hnsw/src/hnsw.c

@@ -347,9 +347,7 @@ static float8 vector_dist(Vector *a, Vector *b, usearch_metric_kind_t metric_kin
 PGDLLEXPORT PG_FUNCTION_INFO_V1(ldb_generic_dist);
 Datum       ldb_generic_dist(PG_FUNCTION_ARGS)
 {
-    if(ldb_pgvector_compat) {
-        elog(ERROR, "Operator can only be used when lantern.pgvector_compat=FALSE");
-    }
+    elog(ERROR, "Operator <?> is deprecated. Please explicitly use the operator that matches your distance function.");
     PG_RETURN_NULL();
 }
 

lantern_hnsw/src/hnsw/options.c

@@ -15,9 +15,6 @@
 #include <utils/rel.h>  // RelationData
 #include <utils/syscache.h>
 
-#include "../hooks/executor_start.h"
-#include "../hooks/post_parse.h"
-
 // We import this header file
 // to access the op class support function pointers
 #include "../hnsw.h"
@@ -54,10 +51,6 @@ int   ldb_external_index_port;
 char *ldb_external_index_host;
 bool  ldb_external_index_secure;
 
-// if this variable is set to true
-// our operator rewriting hooks will be disabled
-bool ldb_pgvector_compat;
-
 // this variable is only set during testing and controls whether
 // certain elog() calls are made
 // see ldb_dlog() definition and callsites for details
@@ -366,17 +359,6 @@ void _PG_init(void)
                              NULL,
                              NULL);
 
-    DefineCustomBoolVariable("lantern.pgvector_compat",
-                             "Whether or not the operator <-> should automatically detect the right distance function",
-                             "set this to 1 to disable operator rewriting hooks",
-                             &ldb_pgvector_compat,
-                             true,
-                             PGC_USERSET,
-                             0,
-                             NULL,
-                             NULL,
-                             NULL);
-
     DefineCustomIntVariable("lantern.external_index_port",
                             "Port for external indexing",
                             "Change this value if you run lantern daemon on different port",
@@ -417,30 +399,8 @@ void _PG_init(void)
     MarkGUCPrefixReserved("_lantern_internal");
 #endif
 
-    original_post_parse_analyze_hook = post_parse_analyze_hook;
-    original_ExecutorStart_hook = ExecutorStart_hook;
-
-    post_parse_analyze_hook = post_parse_analyze_hook_with_operator_check;
-    ExecutorStart_hook = ExecutorStart_hook_with_operator_check;
-
 #ifndef NDEBUG
     signal(SIGSEGV, ldb_wait_for_gdb);
     signal(SIGABRT, ldb_wait_for_gdb);
 #endif
 }
-
-// Called with extension unload.
-void _PG_fini(void)
-{
-    // Return back the original hook value.
-    // This check is because there might be case if while we stop the hooks (in pgvector_compat mode)
-    // Another extension will be loaded and it will overwrite the hooks
-    // And when lantern extension will be unloaded it will set the hooks to original values
-    // Overwriting the current changed hooks set by another extension
-    if(ExecutorStart_hook == ExecutorStart_hook_with_operator_check) {
-        ExecutorStart_hook = original_ExecutorStart_hook;
-    }
-    if(post_parse_analyze_hook == post_parse_analyze_hook_with_operator_check) {
-        post_parse_analyze_hook = original_post_parse_analyze_hook;
-    }
-}

lantern_hnsw/src/hnsw/options.h

@@ -78,7 +78,6 @@ bytea* ldb_amoptions(Datum reloptions, bool validate);
 extern int   ldb_hnsw_init_k;
 extern int   ldb_hnsw_ef_search;
 extern bool  ldb_is_test;
-extern bool  ldb_pgvector_compat;
 extern int   ldb_external_index_port;
 extern char* ldb_external_index_host;
 extern bool  ldb_external_index_secure;