feat: unlimited size cube

3 files changed, 12 insertions, 3 deletions

diff --git a/include/fieldstat/fieldstat.h b/include/fieldstat/fieldstat.h
index 8667d27..f634b00 100644
--- a/include/fieldstat/fieldstat.h
+++ b/include/fieldstat/fieldstat.h
@@ -68,7 +68,7 @@ struct fieldstat *fieldstat_dup(const struct fieldstat *instance);
     * @param shared_tags: tags that are shared by all cubes in this instance. This is the key of the cube. Cannot be NULL. Must be unique. shared_tags are ordered.
     * @param n_tag: number of shared tags.
     * @param mode: sampling mode. Refer to enum sampling_mode.
-    * @param max_n_cell: max number of cells in each cube.
+    * @param max_n_cell: max number of cells in each cube. When mode is TOPK, max_n_cell > 0, while in COMPREHENSIVE mode, max_n_cell can be 0, meaning that there is no limit.
     * @return cube id, if success; otherwise, return -1. Fail only when `instance` == NULL. the uniqueness of shared_tags is not checked, user should guarantee that shared_tags are unique.
 */
 int fieldstat_register_cube(struct fieldstat *instance, const struct fieldstat_tag *shared_tags, size_t n_tag, enum sampling_mode mode, size_t max_n_cell);
diff --git a/src/fieldstat.c b/src/fieldstat.c
index 07142a3..a0ccdbb 100644
--- a/src/fieldstat.c
+++ b/src/fieldstat.c
@@ -273,6 +273,13 @@ int fieldstat_register_cube(struct fieldstat *instance, const struct fieldstat_t
         shared_tags = NULL;
         n_tag = 0;
     }
+    if (mode == SAMPLING_MODE_TOPK && max_n_cell == 0) {
+        return -1;
+    }
+    if (max_n_cell == 0) {
+        max_n_cell = INTMAX_MAX;
+    }
+
     struct fs_cube *cube = fieldstat_cube_new(shared_tags, n_tag, mode, max_n_cell);
 
     int cube_id = fieldstat_append_cube_to_instance(instance, cube);
diff --git a/src/tags/cell_manager.c b/src/tags/cell_manager.c
index a10c5e2..9dfe01d 100644
--- a/src/tags/cell_manager.c
+++ b/src/tags/cell_manager.c
@@ -44,8 +44,10 @@ struct cell_manager *cell_manager_new_without_map(enum sampling_mode sampling_mo
     cell_manager->next_cell_id = 0;
 
     cell_manager->id_tag_array_len = max_cell_num;
-    cell_manager->id_tag_array = calloc(max_cell_num, sizeof(struct tag_hash_key *));
-    assert(cell_manager->id_tag_array != NULL);
+    if (cell_manager->id_tag_array_len > 4096) { // support dynamic size cube (especially when max_max_num is unlimited)
+        cell_manager->id_tag_array_len = 4096;
+    }
+    cell_manager->id_tag_array = calloc(cell_manager->id_tag_array_len, sizeof(struct tag_hash_key *));
     cell_manager->max_cell_id = -1;
 
     return cell_manager;