Details
-
Task
-
Status: Closed (View Workflow)
-
Blocker
-
Resolution: Fixed
-
None
Description
It'd be useful for some plugins to read/store data from/to server tables.
So we can provide the service (as it is defined in the Plugin API) with that functionality.
The functions of that service look exactly as their counterparts from the client library, so users won't be needing extra documentation.
The decided set of functions is:
mysql_init
mysql_real_connect
mysql_options
mysql_close
mysql_real_query
mysql_affected_rows
mysql_errno
mysql_error
mysql_store_result
mysql_free_result
mysql_num_rows
mysql_num_fields
mysql_fetch_row
mysql_fetch_lengths
So that the plugin can do calls like
mysql_real_connect();
mysql_real_query();..
res= mysql_store_result();
...
and it's supposed to work as it does with the client library. The only addition is the
MYSQL_CONNECT_LOCAL option. Connection of that sort attaches to the server itself.
These local queries handled using Server_runnable class model.
Attachments
Issue Links
- blocks
-
-
- Closed
-
- causes
-
MDEV-26890 rpl.rpl_shutdown_wait_slaves fails with timeout on sync_with_master
-
- Closed
-
- includes
-
-
- Closed
-
- is blocked by
-
-
- Closed
-
Activity
I'll comment here, I had some issues with email and your commit email was apparently lost.
The main thought — in this particular case using a service (as in libservice/HOWTO) looks somewhat artificial. A service is a way to create a stable external API out of server internals for plugins to use. But here there is already an API - client-server API, and you build on it (which is great). I'm not sure it adds a lot of value to create a service to proxy another external API, perhaps a plugin can simply use <mysql.h> without requiring a service wrapper?
A couple of smaller comments:
> diff --git a/sql-common/client.c b/sql-common/client.c
|
> index c66cb1a..ace5630 100644
|
> --- a/sql-common/client.c
|
> +++ b/sql-common/client.c
|
> @@ -2844,6 +2844,11 @@ set_connect_attributes(MYSQL *mysql, char *buff, size_t buf_len)
|
> }
|
>
|
>
|
> +#ifdef MYSQL_SERVER
|
> +extern MYSQL * STDCALL do_local_connect(MYSQL *mysql);
|
> +#endif /*MYSQL_SERVER*/
|
|
|
this shouldn't need an #ifdef, one ifdef below is enough.
|
|
|
alternatively, you can remove ifdef below and do here
|
|
|
#ifdef MYSQL_SERVER
|
#define do_local_connect(X) 1
|
#endif
|
|
|
this is, probably, the preferrable way, keeps ifdefs out of the code.
|
|
|
> +
|
> +
|
> MYSQL * STDCALL
|
> CLI_MYSQL_REAL_CONNECT(MYSQL *mysql,const char *host, const char *user,
|
> const char *passwd, const char *db,
|
> diff --git a/sql/service_sql.cc b/sql/service_sql.cc
|
> new file mode 100644
|
> index 0000000..36f65b7
|
> --- /dev/null
|
> +++ b/sql/service_sql.cc
|
> @@ -0,0 +1,43 @@
|
> +/* Copyright 2018 Codership Oy
|
|
|
Is that right?
|
|
|
> +
|
> + This program is free software; you can redistribute it and/or modify
|
> + it under the terms of the GNU General Public License as published by
|
> + the Free Software Foundation; version 2 of the License.
|
for now the plan is to reshpe the patch so it's possible to just include mysql.h to get the interface to the local connection since the server includes the libmysql in it.
Then we should move pieces of the embedded server into the normal server that provide local data transfer, and make loc_... methods with them.
mysql_real_escape_string present in headers but has nothing to link with
if (mysql_real_query(mysql,
|
STRING_WITH_LEN("CREATE TABLE mysql." HISTORY_DB_NAME
|
" ( hash varbinary(512),"
|
" time timestamp default current_time,"
|
" primary key (hash), index tm (time) )")))
|
{
|
crashes the server on attempt to insert with default time
and such table:
if (mysql_real_query(mysql,
|
STRING_WITH_LEN("CREATE TABLE mysql." HISTORY_DB_NAME
|
" ( hash varbinary(512),"
|
" time timestamp default current_time,"
|
" primary key (hash), index tm (time) )")))
|
{
|
inserts 0000-00-00 00:00:00 but according to table definition shoud include current time:
show create table mysql.reuse_password_check_history;
|
Table Create Table
|
reuse_password_check_history CREATE TABLE `reuse_password_check_history` (
|
`hash` varbinary(512) NOT NULL,
|
`time` timestamp NOT NULL DEFAULT current_timestamp() ON UPDATE current_timestamp(),
|
PRIMARY KEY (`hash`),
|
KEY `tm` (`time`)
|
) ENGINE=MyISAM DEFAULT CHARSET=latin1
|
the implementation proposal for today.
https://github.com/mariadb/server/commit/b638b0f515976327c2b11b730747931369101990
This does not built:
FAILED: sql/CMakeFiles/sql.dir/sql_plugin.cc.o
|
/usr/bin/c++ -DDBUG_TRACE -DHAVE_CONFIG_H -DHAVE_EVENT_SCHEDULER -DHAVE_OPENSSL -DHAVE_POOL_OF_THREADS -DMYSQL_SERVER -D_FILE_OFFSET_BITS=64 -Iwsrep-lib/include -Iwsrep-lib/wsrep-API/v26 -Iinclude -Isql -Iextra/pcre2/src/pcre2-build -Iextra/pcre2/src/pcre2/src -Itpool -fPIC -g -DENABLED_DEBUG_SYNC -ggdb3 -DSAFE_MUTEX -DSAFEMALLOC -DTRASH_FREED_MEMORY -Wall -Wextra -Wformat-security -Wno-format-truncation -Wno-init-self -Wno-nonnull-compare -Wno-unused-parameter -Woverloaded-virtual -Wnon-virtual-dtor -Wvla -Wwrite-strings -Werror -std=gnu++11 -MD -MT sql/CMakeFiles/sql.dir/sql_plugin.cc.o -MF sql/CMakeFiles/sql.dir/sql_plugin.cc.o.d -o sql/CMakeFiles/sql.dir/sql_plugin.cc.o -c sql/sql_plugin.cc
|
sql/sql_plugin.cc: In member function ‘virtual bool sys_var_pluginvar::global_update(THD*, set_var*)’:
|
sql/sql_plugin.cc:3622:27: error: ‘PLUGIN_VAR_NO_LOCK’ was not declared in this scope; did you mean ‘PLUGIN_VAR_LONGLONG’?
|
3622 | if (plugin_var->flags & PLUGIN_VAR_NO_LOCK)
|
| ^~~~~~~~~~~~~~~~~~
|
| PLUGIN_VAR_LONGLONG
|
sql/sql_plugin.cc:3627:27: error: ‘PLUGIN_VAR_NO_LOCK’ was not declared in this scope; did you mean ‘PLUGIN_VAR_LONGLONG’?
|
3627 | if (plugin_var->flags & PLUGIN_VAR_NO_LOCK)
|
| ^~~~~~~~~~~~~~~~~~
|
| PLUGIN_VAR_LONGLONG
|
[21/106] Building CXX object sql/CMakeFiles/sql.dir/uniques.cc.o
|
ninja: build stopped: subcommand failed.
|
Problem with timestamp is not solved (branch to check bb-10.7-MDEV-9245-3)
http://lists.askmonty.org/pipermail/commits/2019-April/013658.html