Implement bool compression (#7233)

.unreleased/pr_7701

@@ -0,0 +1 @@
+Implements: #7701 Implement a custom compression algorithm for bool columns. It is experimental and can undergo backwards-incompatible changes. For testing, enable it using timescaledb.enable_bool_compression = on.

sql/updates/latest-dev.sql

@@ -6,3 +6,6 @@ CREATE FUNCTION _timescaledb_functions.compressed_data_has_nulls(_timescaledb_in
     RETURNS BOOL
     LANGUAGE C STRICT IMMUTABLE
     AS '@MODULE_PATHNAME@', 'ts_update_placeholder';
+
+INSERT INTO _timescaledb_catalog.compression_algorithm( id, version, name, description) values
+( 5, 1, 'COMPRESSION_ALGORITHM_BOOL', 'bool');

sql/updates/reverse-dev.sql

@@ -6,3 +6,4 @@ ALTER TABLE _timescaledb_internal.bgw_job_stat_history
 
 DROP FUNCTION IF EXISTS _timescaledb_functions.compressed_data_has_nulls(_timescaledb_internal.compressed_data);
 
+DELETE FROM _timescaledb_catalog.compression_algorithm WHERE id = 5 AND version = 1 AND name = 'COMPRESSION_ALGORITHM_BOOL';

src/cross_module_fn.c

@@ -77,6 +77,8 @@ CROSSMODULE_WRAPPER(dictionary_compressor_append);
 CROSSMODULE_WRAPPER(dictionary_compressor_finish);
 CROSSMODULE_WRAPPER(array_compressor_append);
 CROSSMODULE_WRAPPER(array_compressor_finish);
+CROSSMODULE_WRAPPER(bool_compressor_append);
+CROSSMODULE_WRAPPER(bool_compressor_finish);
 CROSSMODULE_WRAPPER(create_compressed_chunk);
 CROSSMODULE_WRAPPER(compress_chunk);
 CROSSMODULE_WRAPPER(decompress_chunk);
@@ -419,6 +421,8 @@ TSDLLEXPORT CrossModuleFunctions ts_cm_functions_default = {
 	.dictionary_compressor_finish = error_no_default_fn_pg_community,
 	.array_compressor_append = error_no_default_fn_pg_community,
 	.array_compressor_finish = error_no_default_fn_pg_community,
+	.bool_compressor_append = error_no_default_fn_pg_community,
+	.bool_compressor_finish = error_no_default_fn_pg_community,
 	.hypercore_handler = process_hypercore_handler,
 	.hypercore_proxy_handler = process_hypercore_proxy_handler,
 	.is_compressed_tid = error_no_default_fn_pg_community,

src/cross_module_fn.h

@@ -150,6 +150,8 @@ typedef struct CrossModuleFunctions
 	PGFunction dictionary_compressor_finish;
 	PGFunction array_compressor_append;
 	PGFunction array_compressor_finish;
+	PGFunction bool_compressor_append;
+	PGFunction bool_compressor_finish;
 	PGFunction hypercore_handler;
 	PGFunction hypercore_proxy_handler;
 	PGFunction is_compressed_tid;

src/guc.c

@@ -149,6 +149,7 @@ TSDLLEXPORT bool ts_guc_auto_sparse_indexes = true;
 TSDLLEXPORT bool ts_guc_default_hypercore_use_access_method = false;
 bool ts_guc_enable_chunk_skipping = false;
 TSDLLEXPORT bool ts_guc_enable_segmentwise_recompression = true;
+TSDLLEXPORT bool ts_guc_enable_bool_compression = false;
 
 /* Enable of disable columnar scans for columnar-oriented storage engines. If
  * disabled, regular sequence scans will be used instead. */
@@ -746,6 +747,17 @@ _guc_init(void)
 							 NULL,
 							 NULL);
 
+	DefineCustomBoolVariable(MAKE_EXTOPTION("enable_bool_compression"),
+							 "Enable experimental bool compression functionality",
+							 "Enable bool compression",
+							 &ts_guc_enable_bool_compression,
+							 false,
+							 PGC_USERSET,
+							 0,
+							 NULL,
+							 NULL,
+							 NULL);
+
 	/*
 	 * Define the limit on number of invalidation-based refreshes we allow per
 	 * refresh call. If this limit is exceeded, fall back to a single refresh that

src/guc.h

@@ -69,6 +69,7 @@ extern TSDLLEXPORT bool ts_guc_enable_delete_after_compression;
 extern TSDLLEXPORT bool ts_guc_enable_merge_on_cagg_refresh;
 extern bool ts_guc_enable_chunk_skipping;
 extern TSDLLEXPORT bool ts_guc_enable_segmentwise_recompression;
+extern TSDLLEXPORT bool ts_guc_enable_bool_compression;
 
 #ifdef USE_TELEMETRY
 typedef enum TelemetryLevel

tsl/src/compression/README.md

@@ -67,6 +67,15 @@ structure and does not actually compress it (though TOAST-based compression
 can be applied on top). It is the compression mechanism used when no other
 compression mechanism works. It can store any type of data.
 
+### Bool Compressor
+
+The bool compressor is a simple compression algorithm that stores boolean values
+using the simple8b_rle algorithm only, without any additional processing. During
+decompression it decompresses the data and stores it in memory as a bitmap. The
+row based iterators then walk through the bitmap. The bool compressor differs from
+the other compressors in that it stores the last non-value as a place holder for
+the null values. This is done to make vectorization easier.
+
 # Merging chunks while compressing #
 
 ## Setup ##

tsl/src/compression/algorithms/CMakeLists.txt

@@ -3,5 +3,6 @@ set(SOURCES
     ${CMAKE_CURRENT_SOURCE_DIR}/datum_serialize.c
     ${CMAKE_CURRENT_SOURCE_DIR}/deltadelta.c
     ${CMAKE_CURRENT_SOURCE_DIR}/dictionary.c
-    ${CMAKE_CURRENT_SOURCE_DIR}/gorilla.c)
+    ${CMAKE_CURRENT_SOURCE_DIR}/gorilla.c
+    ${CMAKE_CURRENT_SOURCE_DIR}/bool_compress.c)
 target_sources(${TSL_LIBRARY_NAME} PRIVATE ${SOURCES})