杭州易景数通科技有限公司开源了其羲和数据库openHalo,github地址: https://github.com/HaloTech-Co-Ltd/openHalo.git, openHalo基于PostgreSQL开发,其最大的特点是兼容MySQL,支持MySQL,PostgreSQL两种协议。其支持以通过MySQL客户端连接openHalo数据库实例。另外,其同一数据库实例可同时支持MySQL客户端与PostgreSQL客户端连接,openHalo会自动识别客户端类型,选择合适的协议进行通信,选择合适的解析器、优化器、执行器进行处理。

安装与使用
- 下载源码:
git clone https://github.com/HaloTech-Co-Ltd/openHalo.git
- 编译安装
./configure --prefix=/home/postgres/openhalo --enable-debug --with-uuid=ossp --with-icu CFLAGS=-O0 make && make install cd contrib make && make install
- 初始化数据库实例
initdb -D openhalo-data/
# 查看数据库实例目录,PG内核版本为14
postgres@slpc:~/openhalo/openhalo-data$ ls
base pg_hba.conf pg_notify pg_stat pg_twophase postgresql.auto.conf
global pg_ident.conf pg_replslot pg_stat_tmp PG_VERSION postgresql.conf
pg_commit_ts pg_logical pg_serial pg_subtrans pg_wal
pg_dynshmem pg_multixact pg_snapshots pg_tblspc pg_xact
postgres@slpc:~/openhalo/openhalo-data$ cat PG_VERSION
14
- 修改postgresql.conf配置文件
开源版本兼容MySQL。Oralce,DB2,SQLServer,Sybase等数据库的兼容性仅企业版支持。
# MySQL相关GUC参数
#mysql.listener_on = true # (enable MySQL listener; change requires restart)
#mysql.port = 3306 # (second_port is for MySQL mode; change requires restart)
#mysql.halo_mysql_version = '5.7.32-log' # (change requires restart)
#mysql.ci_collation = true # (change requires restart)
#mysql.explicit_defaults_for_timestamp = false # (change requires restart)
#mysql.auto_rollback_tx_on_error = false # (change requires restart)
#mysql.halo_mysql_ci_collation = 0 # (change requires restart)
#mysql.column_name_case_control = 0 # (change requires restart)
#mysql.max_allowed_packet = 536870912B # (512MB, change requires restart)
#mysql.support_multiple_table_update = true # (change requires restart)
database_compat_mode = 'mysql' # database compat mode, values can be:
# postgresql (default)
# mysql (MySQL mode)
# oracle (support in Enterprise Edition)
# db2 (support in Enterprise Edition)
# sqlserver (support in Enterprise Edition)
# sybase (support in Enterprise Edition)
standard_parserengine_auxiliary = 'on' # on, off, yes, no, true or false
- 配置环境变量
sudo mkdir /var/run/openhalo sudo chown postgres:postgres /var/run/openhalo export PGDATA=/home/postgres/openhalo/openhalo-data export PATH=/home/postgres/openhalo/bin:$PATH export LD_LIBRARY_PATH=/home/postgres/openhalo/lib:$LD_LIBRARY_PATH
- 启动数据库实例
pg_ctl start -D $PGDATA
- 连接数据库实例,可以看到其PostgreSQL内核版本为14.10
postgres@slpc:~/openhalo$ psql
psql (1.0.14.10)
Type "help" for help.
halo0root=# show database_compat_mode ;
database_compat_mode
----------------------
mysql
(1 row)
- 安装
aux_mysql扩展,该扩展为MySQL兼容辅助扩展模块
-- 安装aux_mysql扩展
halo0root=# create extension aux_mysql cascade;
NOTICE: installing required extension "uuid-ossp"
CREATE EXTENSION
halo0root=# \dx
List of installed extensions
Name | Version | Schema | Description
-----------+---------+------------+-------------------------------------------------
aux_mysql | 1.4 | public | MySQL Supplementary Extension
plpgsql | 1.0 | pg_catalog | PL/pgSQL procedural language
uuid-ossp | 1.1 | public | generate universally unique identifiers (UUIDs)
(3 rows)
- 安装mysql客户端,通过MySQL客户端连接openHalo数据库实例
sudo apt-get install mysql-client
# 通过mysql客户端连接openHalo数据库实例
postgres@slpc:~/openhalo/openhalo-data$ mysql -P 3306 -h 127.0.0.1
Welcome to the MySQL monitor. Commands end with ; or \g.
Your MySQL connection id is 1
Server version: 5.7.32-log MySQL Server (GPL)
Copyright (c) 2000, 2025, Oracle and/or its affiliates.
Oracle is a registered trademark of Oracle Corporation and/or its
affiliates. Other names may be trademarks of their respective
owners.
Type 'help;' or '\h' for help. Type '\c' to clear the current input statement.
mysql> select version();
+------------+
| version |
+------------+
| 5.7.32-log |
+------------+
1 row in set (0.00 sec)
mysql> show databases;
+--------------------+
| Database |
+--------------------+
| information_schema |
| mys_sys |
| mysql |
| performance_schema |
| sys |
+--------------------+
5 rows in set (0.00 sec)
- 创建数据库
mysql> create database halo0;
Query OK, 1 row affected (0.00 sec)
初步源码分析
openHalo最大的特色就是兼容MySQL,我们初步分析一下openHalo是如何兼容MySQL的。首先是aux_mysql扩展,该扩展中包含了兼容MySQL所需的函数、运算符等,这部分大概有三万多行SQL代码。
typedef enum
{
POSTGRESQL_COMPAT_MODE,
MYSQL_COMPAT_MODE,
} DatabaseCompatModeType;
另外,实现了MySQL语法解析器,兼容MySQL语法,以及其他SQL命令相关的实现。
List *raw_parser(const char *str, RawParseMode mode)
{
List *raw_parsetree = NIL;
MemoryContext oldctx = CurrentMemoryContext;
if (parserengine == NULL) // 如果没有指定解析引擎,则使用标准解析引擎
parserengine = GetStandardParserEngine();
Assert(parserengine != NULL);
Assert(parserengine->raw_parser != NULL);
PG_TRY();
{
raw_parsetree = parserengine->raw_parser(str, mode); // 调用解析引擎的解析函数
}
PG_CATCH();
{
if (raw_parsetree == NIL && parserengine->is_standard_parser == false
&& (standard_parserengine_auxiliary == STANDARDARD_PARSERENGINE_AUXILIARY_ON
&& parserengine->need_standard_parser == true))
{
MemoryContextSwitchTo(oldctx);
FlushErrorState();
MemoryContextSwitchTo(oldctx);
raw_parsetree = GetStandardParserEngine()->raw_parser(str, mode);
}
else
PG_RE_THROW();
}
PG_END_TRY();
return raw_parsetree;
}
为了支持多款解析器引擎,openHalo定义了ParserRoutine结构体,该结构体中包含了多种接口函数,用于处理不同解析器的解析、语法分析等。实现时需要把原先PostgreSQL解析相关的接口抽象出来。
/* API for parser engine */
typedef struct ParserRoutine
{
NodeTag type;
bool is_standard_parser;
bool need_standard_parser;
const ScanKeywordList *keywordlist;
const uint16 *keyword_tokens;
const uint8 *keyword_categories;
raw_parser_function raw_parser;
transformStmt_function transformStmt;
transformSelectStmt_function transformSelectStmt;
transformInsertStmt_function transformInsertStmt;
transformDeleteStmt_function transformDeleteStmt;
transformUpdateStmt_function transformUpdateStmt;
transformSetOperationStmt_function transformSetOperationStmt;
transformCallStmt_function transformCallStmt;
transformOptionalSelectInto_function transformOptionalSelectInto;
transformSetOperationTree_function transformSetOperationTree;
analyze_requires_snapshot_function analyze_requires_snapshot;
parse_sub_analyze_function parse_sub_analyze;
transformGroupClause_function transformGroupClause;
transformDistinctClause_function transformDistinctClause;
transformOnConflictArbiter_function transformOnConflictArbiter;
transformExpr_function transformExpr;
transformCreateStmt_function transformCreateStmt;
transformAlterTableStmt_function transformAlterTableStmt;
assign_query_collations_function assign_query_collations;
assign_list_collations_function assign_list_collations;
assign_expr_collations_function assign_expr_collations;
select_common_collation_function select_common_collation;
preTransformWindowAgg_fn preTransformWindowAgg;
finalTransformColumnRef_fn finalTransformColumnRef;
ParseFuncOrColumn_function ParseFuncOrColumn;
func_get_detail_function func_get_detail;
make_fn_arguments_function make_fn_arguments;
RewriteInterface rewrite;
} ParserRoutine;
MySQL语法解析器调用栈如下:
mys_yyparse(core_yyscan_t yyscanner) (src\backend\parser\mysql\mys_gram.c:35413)
mys_raw_parser(const char * str, RawParseMode mode) (src\backend\parser\mysql\mys_parser.c:353)
raw_parser(const char * str, RawParseMode mode) (src\backend\parser\parser.c:66)
pg_parse_query(const char * query_string) (src\backend\tcop\postgres.c:641)
standard_exec_simple_query(const char * query_string) (src\backend\tcop\postgres.c:1115)
exec_simple_query(const char * query_string) (src\backend\tcop\postgres2.c:65)
PostgresMain(int argc, char ** argv, const char * dbname, const char * username) (src\backend\tcop\postgres.c:4802)
mainFunc(Port * port, int argc, char ** argv) (src\backend\adapter\mysql\adapter.c:808)
BackendRun(Port * port) (src\backend\postmaster\postmaster.c:4572)
BackendStartup(Port * port) (src\backend\postmaster\postmaster.c:4297)
ServerLoop() (src\backend\postmaster\postmaster.c:1773)
PostmasterMain(int argc, char ** argv) (src\backend\postmaster\postmaster.c:1442)
main(int argc, char ** argv) (src\backend\main\main.c:209)
我们实际调试一条INSERT语句,看一下其调用栈:
mys_transformInsertStmt(ParseState * pstate, InsertStmt * stmt) (src\backend\parser\mysql\mys_analyze.c:956)
transformInsertStmt(ParseState * pstate, InsertStmt * stmt) (src\backend\parser\analyze2.c:70)
mys_transformStmt(ParseState * pstate, Node * parseTree) (src\backend\parser\mysql\mys_analyze.c:191)
transformStmt(ParseState * pstate, Node * parseTree) (src\backend\parser\parser_ep.c:77)
mys_transformOptionalSelectInto(ParseState * pstate, Node * parseTree) (src\backend\parser\mysql\mys_analyze.c:149)
transformOptionalSelectInto(ParseState * pstate, Node * parseTree) (src\backend\parser\analyze2.c:257)
transformTopLevelStmt(ParseState * pstate, RawStmt * parseTree) (src\backend\parser\analyze.c:253)
standard_parse_analyze(RawStmt * parseTree, const char * sourceText, Oid * paramTypes, int numParams, QueryEnvironment * queryEnv) (src\backend\parser\analyze.c:162)
parse_analyze(RawStmt * parseTree, const char * sourceText, Oid * paramTypes, int numParams, QueryEnvironment * queryEnv) (src\backend\parser\parser_ep.c:45)
pg_analyze_and_rewrite(RawStmt * parsetree, const char * query_string, Oid * paramTypes, int numParams, QueryEnvironment * queryEnv) (src\backend\tcop\postgres.c:695)
standard_exec_simple_query(const char * query_string) (src\backend\tcop\postgres.c:1257)
exec_simple_query(const char * query_string) (src\backend\tcop\postgres2.c:65)
PostgresMain(int argc, char ** argv, const char * dbname, const char * username) (src\backend\tcop\postgres.c:4802)
mainFunc(Port * port, int argc, char ** argv) (src\backend\adapter\mysql\adapter.c:808)
BackendRun(Port * port) (src\backend\postmaster\postmaster.c:4572)
BackendStartup(Port * port) (src\backend\postmaster\postmaster.c:4297)
ServerLoop() (src\backend\postmaster\postmaster.c:1773)
PostmasterMain(int argc, char ** argv) (src\backend\postmaster\postmaster.c:1442)
main(int argc, char ** argv) (src\backend\main\main.c:209)
openHalo单个数据库实例可同时支持多种数据库模式,包括MySQL、PostgreSQL等,因此需要支持多种解析器引擎。当客户端连接时,openHalo会根据客户端的请求选择对应的解析引擎。每当客户端连接时,会fork一个新的进程,用于处理客户端的请求,该进程在初始化时会调用InitPostgres函数,在该过程中会调用InitParserEngine()选择合适的解析器。同时在该函数中,会进行一系列的初始化操作,包括执行InitExecutorEngine()选择合适的执行引擎,执行InitPlannerEngine()选择合适的优化器引擎。
我们看一下INSERT语句执行器相关调用栈:
ExecModifyTable(PlanState * pstate) (src\backend\executor\mys_nodeModifyTable.c:3063)
ExecProcNodeFirst(PlanState * node) (src\backend\executor\execProcnode.c:464)
ExecProcNode(PlanState * node) (src\include\executor\executor.h:260)
standard_ExecutePlan(EState * estate, PlanState * planstate, _Bool use_parallel_mode, CmdType operation, _Bool sendTuples, uint64 numberTuples, ScanDirection direction, DestReceiver * dest, _Bool execute_once) (src\backend\executor\execMain.c:1567)
ExecutePlan(EState * estate, PlanState * planstate, _Bool use_parallel_mode, CmdType operation, _Bool sendTuples, uint64 numberTuples, ScanDirection direction, DestReceiver * dest, _Bool execute_once) (src\backend\executor\execMain.c:2989)
mys_ExecutorRun(QueryDesc * queryDesc, ScanDirection direction, uint64 count, _Bool execute_once) (src\backend\executor\mys_execMain.c:220)
ExecutorRun(QueryDesc * queryDesc, ScanDirection direction, uint64 count, _Bool execute_once) (src\backend\executor\execMain.c:315)
ProcessQuery(PlannedStmt * plan, const char * sourceText, ParamListInfo params, QueryEnvironment * queryEnv, DestReceiver * dest, QueryCompletion * qc) (src\backend\tcop\pquery.c:160)
PortalRunMulti(Portal portal, _Bool isTopLevel, _Bool setHoldSnapshot, DestReceiver * dest, DestReceiver * altdest, QueryCompletion * qc) (src\backend\tcop\pquery.c:1294)
PortalRun(Portal portal, long count, _Bool isTopLevel, _Bool run_once, DestReceiver * dest, DestReceiver * altdest, QueryCompletion * qc) (src\backend\tcop\pquery.c:793)
standard_exec_simple_query(const char * query_string) (src\backend\tcop\postgres.c:1341)
exec_simple_query(const char * query_string) (src\backend\tcop\postgres2.c:65)
PostgresMain(int argc, char ** argv, const char * dbname, const char * username) (src\backend\tcop\postgres.c:4802)
mainFunc(Port * port, int argc, char ** argv) (src\backend\adapter\mysql\adapter.c:808)
BackendRun(Port * port) (src\backend\postmaster\postmaster.c:4572)
BackendStartup(Port * port) (src\backend\postmaster\postmaster.c:4297)
ServerLoop() (src\backend\postmaster\postmaster.c:1773)
PostmasterMain(int argc, char ** argv) (src\backend\postmaster\postmaster.c:1442)
main(int argc, char ** argv) (src\backend\main\main.c:209)
除了语法方面,openHalo还支持MySQL协议。为了支持多种协议,openHalo将原有PostgreSQL协议中相关接口进行了抽象,定义了ProtocolInterface结构体,该结构体中包含了多种接口函数,用于处理不同协议的连接、认证、消息处理等。
typedef struct ProtocolRoutine ProtocolInterface;
/*
* ProtocolRoutine
*
* All the callbacks implementing a specific wire protocol
* AWS Babelfish compatible
*/
typedef void (*fn_listen_init)(void);
typedef int (*fn_accept)(pgsocket server_fd, struct Port *port);
typedef void (*fn_close)(pgsocket server_fd);
typedef void (*fn_init)(void);
typedef int (*fn_start)(struct Port *port);
typedef void (*fn_authenticate)(struct Port *port, const char **username);
typedef void (*fn_mainfunc)(struct Port *port, int argc, char *argv[]) pg_attribute_noreturn();
typedef void (*fn_send_message)(ErrorData *edata);
typedef void (*fn_send_cancel_key)(int pid, int32 key);
typedef void (*fn_comm_reset)(void);
typedef bool (*fn_is_reading_msg)(void);
typedef void (*fn_send_ready_for_query)(CommandDest dest);
typedef int (*fn_read_command)(StringInfo inBuf);
typedef void (*fn_end_command)(QueryCompletion *qc, CommandDest dest);
typedef bool (*fn_printtup)(TupleTableSlot *slot, DestReceiver *self, CommandTag commandTag);
typedef void (*fn_printtup_startup)(DestReceiver *self, int operation, TupleDesc typeinfo, CommandTag commandTag);
typedef void (*fn_printtup_shutdown)(DestReceiver *self);
typedef void (*fn_printtup_destroy)(DestReceiver *self);
typedef int (*fn_process_command)(int *first_char, StringInfo inBuf);
typedef void (*fn_report_param_status)(const char *name, char *val);
typedef struct ProtocolRoutine {
NodeTag type;
fn_listen_init listen_init;
fn_accept accept;
fn_close close;
fn_init init;
fn_start start;
fn_authenticate authenticate;
fn_mainfunc mainfunc;
fn_send_message send_message;
fn_send_cancel_key send_cancel_key;
fn_comm_reset comm_reset;
fn_is_reading_msg is_reading_msg;
fn_send_ready_for_query send_ready_for_query;
fn_read_command read_command;
fn_end_command end_command;
fn_printtup printtup;
fn_printtup_startup printtup_startup;
fn_printtup_shutdown printtup_shutdown;
fn_printtup_destroy printtup_destroy;
fn_process_command process_command;
fn_report_param_status report_param_status;
} ProtocolRoutine;
其中MySQL协议接口定义如下:
static ProtocolInterface protocolHandler = {
.type = T_MySQLProtocol,
.listen_init = initListen,
.accept = acceptConn,
.close = closeListen,
.init = initServer,
.start = startServer,
.authenticate = authenticate,
.mainfunc = mainFunc,
.send_message = sendErrorMessage,
.send_cancel_key = mysqlSendCancelKey,
.comm_reset = NULL,
.is_reading_msg = NULL,
.send_ready_for_query = sendReadyForQuery,
.read_command = readCommand,
.end_command = endCommand,
.printtup = printTup,
.printtup_startup = printTupStartup,
.printtup_shutdown = printTupShutdown,
.printtup_destroy = printTupDestroy,
.process_command = processCommand,
.report_param_status = reportParamStatus
};
其他方面,可通过代码调试来查看具体的实现逻辑。这里不再详细介绍。
源码结构分析
在openHalo的源码中,为兼容MySQL涉及如下代码,当前目前整理的肯定不够全面,后续再不断进行补充完善:
openHalo
--> contrib/aux_mysql // MySQL兼容辅助扩展模块, 大概有三万多行SQL代码
--> aux_mysql--1.1--1.2.sql
--> aux_mysql--1.1.sql
--> aux_mysql--1.2--1.3.sql
--> aux_mysql--1.3--1.4.sql
--> src/backend/parser/parsereng.c // 解析引擎相关代码
--> src/backend/optimizer/plan/planner_engine.c // 优化器引擎相关代码
--> src/backend/executor/executor_engine.c // 执行引擎相关代码
--> src/backend/executor/mys_execMain.c
--> src/backend/executor/mys_execPartition.c
--> src/backend/executor/mys_execProcnode.c
--> src/backend/executor/mys_executor.c // mysql执行引擎相关代码
--> src/backend/executor/mys_nodeModifyTable.c
--> src/backend/parser/mysql // MySQL语法解析器相关代码, 大概有三万多行代码
--> check_myskeywords.pl
--> mys_analyze.c
--> mys_gram.y // MySQL语法解析器
--> mys_keywords.c
--> mys_parse_agg.c
--> mys_parse_clause.c
--> mys_parse_expr.c
--> mys_parse_func.c
--> mys_parse_oper.c
--> mys_parse_utilcmd.c
--> mys_parser.c // MySQL语法解析器入口
--> mys_scan.l // lexical scanner for MySQL
--> src/backend/parse/analyze2.c // extra routines to parse analyze
--> src/backend/adapter // MySQL协议适配器 8k左右代码
--> adapter.c
--> errorConvertor.c
--> netTransceiver.c
--> pwdEncryptDecrypt.c
--> systemVar.c
--> userLogonAuth.c
--> uuidShort.c
--> src/backend/commands/mysql
--> mys_prepare.c // 实现MySQL的PREPARE语句
--> mys_sequence.c // 兼容MySQL sequence相关的处理
--> mys_tablecmds.c
--> mys_uservar.c
--> src/backend/commands/trigger2.c // 实现MySQL触发器相关功能
--> src/backend/postmaster/postmaster2.c
--> src/backend/tcop/mysql
--> mys_utility.c
--> src/backend/tcop/postgres2.c
--> src/include/parser/mysql
--> src/include/optimizer/pathnode2.h
--> src/include/optimizer/plan_engine.h
--> src/include/parser/parserapi.h // 解析器API
--> src/include/postmaster/protocol_interface.h
--> src/include/postmaster/postmaster2.h
--> src/include/utils/mysql
--> mys_adtext.h
--> mys_date.h
--> mys_ri_trigger.h
--> mys_ruleutils.h
--> mys_timestamp.h
--> mys_varlena.h
个人思考
个人认为,openHalo数据库中将数据库协议,解析器,执行器进行抽象以支持多种数据库协议多种解析器来实现数据库兼容性是非常好的设计。在进行数据库内核开发时可以借鉴这种设计思想。
参考文档:
多模式兼容




