Merge pull request #1323 from Enferlain/fix/1196-unlabeled-candidate-limit

fix(cypher): scan all unlabeled query candidates

src/cypher/cypher.c

@@ -3071,31 +3071,14 @@ static void scan_alternation_labels(cbm_store_t *store, const char *project, con
     free(copy);
 }
 
-static void scan_pattern_nodes(cbm_store_t *store, const char *project, int max_rows,
-                               cbm_node_pattern_t *first, cbm_node_t **out_nodes, int *out_count) {
+static void scan_pattern_nodes(cbm_store_t *store, const char *project, cbm_node_pattern_t *first,
+                               cbm_node_t **out_nodes, int *out_count) {
     if (first->label && strchr(first->label, '|')) {
         scan_alternation_labels(store, project, first->label, out_nodes, out_count);
     } else if (first->label) {
         cbm_store_find_nodes_by_label(store, project, first->label, out_nodes, out_count);
     } else {
-        cbm_search_params_t params = {.project = project,
-                                      .min_degree = CYP_FOUND_NONE,
-                                      .max_degree = CYP_FOUND_NONE,
-                                      .limit = max_rows * CYP_GROWTH_10};
-        cbm_search_output_t sout = {0};
-        cbm_store_search(store, &params, &sout);
-        *out_count = sout.count;
-        *out_nodes = malloc(sout.count * sizeof(cbm_node_t));
-        for (int i = 0; i < sout.count; i++) {
-            (*out_nodes)[i] = sout.results[i].node;
-            sout.results[i].node.name = NULL;
-            sout.results[i].node.project = NULL;
-            sout.results[i].node.label = NULL;
-            sout.results[i].node.qualified_name = NULL;
-            sout.results[i].node.file_path = NULL;
-            sout.results[i].node.properties_json = NULL;
-        }
-        cbm_store_search_free(&sout);
+        cbm_store_find_nodes(store, project, out_nodes, out_count);
     }
     /* Apply inline property filters — free rejected nodes' strings */
     if (first->prop_count > 0) {
@@ -4779,7 +4762,7 @@ static int expand_additional_patterns(cbm_store_t *store, cbm_query_t *q, const
 
         cbm_node_t *extra_nodes = NULL;
         int extra_count = 0;
-        scan_pattern_nodes(store, project, max_rows, &patn->nodes[0], &extra_nodes, &extra_count);
+        scan_pattern_nodes(store, project, &patn->nodes[0], &extra_nodes, &extra_count);
         int rc = 0;
         if (patn->rel_count == 0) {
             rc = cross_join_nodes(bindings, bind_count, extra_nodes, extra_count, nvar, opt);
@@ -4831,7 +4814,7 @@ static int execute_single(cbm_store_t *store, cbm_query_t *q, const char *projec
     /* Step 1: Scan initial nodes */
     cbm_node_t *scanned = NULL;
     int scan_count = 0;
-    scan_pattern_nodes(store, project, max_rows, &pat0->nodes[0], &scanned, &scan_count);
+    scan_pattern_nodes(store, project, &pat0->nodes[0], &scanned, &scan_count);
 
     /* Build initial bindings with early WHERE */
     int bind_cap = scan_count > max_rows ? scan_count : (max_rows > 0 ? max_rows : SKIP_ONE);

src/store/store.c

@@ -117,6 +117,7 @@ struct cbm_store {
     sqlite3_stmt *stmt_find_node_by_id;
     sqlite3_stmt *stmt_find_node_by_qn;
     sqlite3_stmt *stmt_find_node_by_qn_any; /* QN lookup without project filter */
+    sqlite3_stmt *stmt_find_nodes;
     sqlite3_stmt *stmt_find_nodes_by_name;
     sqlite3_stmt *stmt_find_nodes_by_name_any; /* name lookup without project filter */
     sqlite3_stmt *stmt_find_nodes_by_label;
@@ -1140,6 +1141,7 @@ void cbm_store_close(cbm_store_t *s) {
     finalize_stmt(&s->stmt_find_node_by_id);
     finalize_stmt(&s->stmt_find_node_by_qn);
     finalize_stmt(&s->stmt_find_node_by_qn_any);
+    finalize_stmt(&s->stmt_find_nodes);
     finalize_stmt(&s->stmt_find_nodes_by_name);
     finalize_stmt(&s->stmt_find_nodes_by_name_any);
     finalize_stmt(&s->stmt_find_nodes_by_label);
@@ -1978,7 +1980,9 @@ static int find_nodes_generic(cbm_store_t *s, sqlite3_stmt **slot, const char *s
     }
 
     bind_text(stmt, SKIP_ONE, project);
-    bind_text(stmt, ST_COL_2, val);
+    if (val) {
+        bind_text(stmt, ST_COL_2, val);
+    }
 
     int cap = ST_INIT_CAP_16;
     int n = 0;
@@ -2006,6 +2010,19 @@ static int find_nodes_generic(cbm_store_t *s, sqlite3_stmt **slot, const char *s
     return CBM_STORE_OK;
 }
 
+int cbm_store_find_nodes(cbm_store_t *s, const char *project, cbm_node_t **out, int *count) {
+    if (!s) {
+        *out = NULL;
+        *count = 0;
+        return CBM_STORE_ERR;
+    }
+    return find_nodes_generic(s, &s->stmt_find_nodes,
+                              "SELECT id, project, label, name, qualified_name, file_path, "
+                              "start_line, end_line, properties FROM nodes "
+                              "WHERE project = ?1;",
+                              project, NULL, out, count);
+}
+
 int cbm_store_find_nodes_by_name(cbm_store_t *s, const char *project, const char *name,
                                  cbm_node_t **out, int *count) {
     if (!s) {

src/store/store.h

@@ -403,6 +403,9 @@ int cbm_store_find_node_by_qn(cbm_store_t *s, const char *project, const char *q
 /* Find node by qualified_name only (no project filter — QNs are globally unique). */
 int cbm_store_find_node_by_qn_any(cbm_store_t *s, const char *qn, cbm_node_t *out);
 
+/* Find all nodes in a project. Returns allocated array, caller frees. */
+int cbm_store_find_nodes(cbm_store_t *s, const char *project, cbm_node_t **out, int *count);
+
 /* Find nodes by name (exact match). Returns allocated array, caller frees. */
 int cbm_store_find_nodes_by_name(cbm_store_t *s, const char *project, const char *name,
                                  cbm_node_t **out, int *count);

tests/test_cypher.c

@@ -1005,6 +1005,46 @@ TEST(cypher_exec_where_eq) {
     PASS();
 }
 
+/* #1196: max_rows limits projected results, not the source-node candidates
+ * considered before WHERE. The old unlabeled scan searched only
+ * 10 * max_rows nodes, so a valid match later in search order disappeared. */
+TEST(cypher_exec_unlabeled_where_beyond_result_limit_issue1196) {
+    cbm_store_t *s = cbm_store_open_memory();
+    ASSERT_NOT_NULL(s);
+    ASSERT_EQ(cbm_store_upsert_project(s, "test", "/tmp/test"), CBM_STORE_OK);
+
+    for (int i = 0; i < 11; i++) {
+        char name[32];
+        char qn[64];
+        snprintf(name, sizeof(name), "early_%02d", i);
+        snprintf(qn, sizeof(qn), "test.%s", name);
+        cbm_node_t distractor = {.project = "test",
+                                 .label = "Function",
+                                 .name = name,
+                                 .qualified_name = qn,
+                                 .file_path = "early.py"};
+        ASSERT_GT(cbm_store_upsert_node(s, &distractor), 0);
+    }
+
+    cbm_node_t late = {.project = "test",
+                       .label = "Function",
+                       .name = "zz_late_match",
+                       .qualified_name = "test.zz_late_match",
+                       .file_path = "late.py"};
+    ASSERT_GT(cbm_store_upsert_node(s, &late), 0);
+
+    cbm_cypher_result_t r = {0};
+    int rc = cbm_cypher_execute(
+        s, "MATCH (n) WHERE n.name = \"zz_late_match\" RETURN n.name", "test", 1, &r);
+    ASSERT_EQ(rc, 0);
+    ASSERT_EQ(r.row_count, 1);
+    ASSERT_STR_EQ(r.rows[0][0], "zz_late_match");
+
+    cbm_cypher_result_free(&r);
+    cbm_store_close(s);
+    PASS();
+}
+
 /* #874: coalesce(var.prop, literal) in WHERE — null-safe numeric filters
  * for audit queries over OPTIONAL graph properties. The parser rejected the
  * call outright ("unexpected operator"); RETURN-side coalesce already
@@ -3897,6 +3937,7 @@ SUITE(cypher) {
     RUN_TEST(cypher_issue252_tointeger);
     RUN_TEST(cypher_issue305_count_star_alias);
     RUN_TEST(cypher_exec_where_eq);
+    RUN_TEST(cypher_exec_unlabeled_where_beyond_result_limit_issue1196);
     RUN_TEST(cypher_exec_varlength_path_semantics_issue797);
     RUN_TEST(cypher_exec_where_coalesce_issue874);
     RUN_TEST(cypher_exec_where_regex);