A highly Performant,Safe,Dynamic SQL(Compile time) ORM framework written in Rust, inspired by Mybatis and MybatisPlus.
Why not diesel or not sqlx ?
Framework
Async/.await
Learning curve
Dynamic SQL/py/Wrapper/built-in CRUD
Logical delete plugin
Pagination plugin
rbatis
√
easy
√
√
√
sqlx
√
hard (depends on macros and env. variables)
x
x
x
diesel
x
hard (use FFI, unsafe)
x
x
x
Performance comparison with Golang (in a docker environment)
Framework
Mysql(docker)
SQL statement(10k)
ns/operation(lower is better)
Qps(higher is better)
Memory usage(lower is better)
Rust-rbatis/tokio
1 CPU, 1G memory
select count(1) from table;
965649 ns/op
1035 Qps/s
2.1MB
Go-GoMybatis/http
1 CPU, 1G memory
select count(1) from table;
1184503 ns/op
844 Qps/s
28.4MB
No Runtimes
Dynamic SQL is Zero cost abstraction and implemented using (compile-time,Cow(Reduce unnecessary cloning)) techniques
used json with serde_json for passing parameters and communication
high performance, single threaded benchmark can easily achieve 200,000 QPS - data returned from database directly (
zero lookup time) on a Windows 10 6 core i7 with 16 GB memory machine. Performace will be better using multiple
threads, and it outperforms Go's GoMyBatis.
supports logical deletes, pagination, py-like SQL and basic Mybatis functionalities.
supports future,(in theory, if all io operations are replaced with async_std/tokio, it could achieve higher
concurrency than Go-lang)
supports logging, customizable logging based on log crate
used 100% safe Rust with #![forbid(unsafe_code)] enabled
# add this library,and cargo install
# json(required)
serde = { version = "1", features = ["derive"]}
serde_json = "1"
# Date time (required)
chrono = { version = "0.4", features = ["serde"]}
# logging lib(required)
log = "0.4"
fast_log="1.3"
# BigDecimal lib(optional)
bigdecimal = "0.2"
# rbatis lib(required)
rbatis = { version = "2.0"}
Quick example: QueryWrapper and common usages (see example/crud_test.rs for details)
//#[macro_use] define in 'root crate' or 'mod.rs' or 'main.rs'#[macro_use]externcrate rbatis;use rbatis::crud::CRUD;/// may also write `CRUDTable` as `impl CRUDTable for BizActivity{}`/// #[crud_table]/// #[crud_table(table_name:biz_activity)]/// #[crud_table(table_name:"biz_activity"|table_columns:"id,name,version,delete_flag")]/// #[crud_table(table_name:"biz_activity"|table_columns:"id,name,version,delete_flag"|formats_pg:"id:{}::uuid")]#[crud_table]#[derive(Clone,Debug)]pubstructBizActivity{pubid:Option<String>,pubname:Option<String>,pubpc_link:Option<String>,pubh5_link:Option<String>,pubpc_banner_img:Option<String>,pubh5_banner_img:Option<String>,pubsort:Option<String>,pubstatus:Option<i32>,pubremark:Option<String>,pubcreate_time:Option<NaiveDateTime>,pubversion:Option<i32>,pubdelete_flag:Option<i32>,}// (optional) manually implement instead of using `derive(CRUDTable)`. This allows manually rewriting `table_name()` function and supports code completion in IDE.// use rbatis::crud::CRUDTable;//impl CRUDTable for BizActivity {// type IdType = String; // fn table_name()->String{// "biz_activity".to_string()// }// fn table_columns()->String{// "id,name,delete_flag".to_string()// }//}#[tokio::main]asyncfnmain(){/// enable log crate to show sql logs
fast_log::init_log("requests.log",1000, log::Level::Info,None,true);/// initialize rbatis. May use `lazy_static` crate to define rbatis as a global variable because rbatis is thread safelet rb = Rbatis::new();/// connect to database
rb.link("mysql://root:123456@localhost:3306/test").await.unwrap();/// customize connection pool parameters (optional)// let mut opt =PoolOptions::new();// opt.max_size=100;// rb.link_opt("mysql://root:123456@localhost:3306/test",&opt).await.unwrap();/// newly constructed wrapper sql logiclet wrapper = rb.new_wrapper().eq("id",1)//sql: id = 1.and()//sql: and .ne("id",1)//sql: id <> 1.in_array("id",&[1,2,3])//sql: id in (1,2,3).not_in("id",&[1,2,3])//sql: id not in (1,2,3).like("name",1)//sql: name like 1.or()//sql: or.not_like("name","asdf")//sql: name not like 'asdf'.between("create_time","2020-01-01 00:00:00","2020-12-12 00:00:00")//sql: create_time between '2020-01-01 00:00:00' and '2020-01-01 00:00:00'.group_by(&["id"])//sql: group by id.order_by(true,&["id","name"])//sql: group by id,name;let activity = BizActivity{id:Some("12312".to_string()),name:None,remark:None,create_time:Some(NaiveDateTime::now()),version:Some(1),delete_flag:Some(1),};/// saving
rb.save(&activity,&[]).await;//Exec ==> INSERT INTO biz_activity (create_time,delete_flag,h5_banner_img,h5_link,id,name,pc_banner_img,pc_link,remark,sort,status,version) VALUES ( ? , ? , ? , ? , ? , ? , ? , ? , ? , ? , ? , ? )/// batch saving
rb.save_batch(&vec![activity],&[]).await;//Exec ==> INSERT INTO biz_activity (create_time,delete_flag,h5_banner_img,h5_link,id,name,pc_banner_img,pc_link,remark,sort,status,version) VALUES ( ? , ? , ? , ? , ? , ? , ? , ? , ? , ? , ? , ? ),( ? , ? , ? , ? , ? , ? , ? , ? , ? , ? , ? , ? )/// The query, Option wrapper, is None if the data is not foundlet result:Option<BizActivity> = rb.fetch_by_column("id",&"1".to_string()).await.unwrap();//Query ==> SELECT create_time,delete_flag,h5_banner_img,h5_link,id,name,pc_banner_img,pc_link,remark,sort,status,version FROM biz_activity WHERE delete_flag = 1 AND id = ? /// query alllet result:Vec<BizActivity> = rb.list().await.unwrap();//Query ==> SELECT create_time,delete_flag,h5_banner_img,h5_link,id,name,pc_banner_img,pc_link,remark,sort,status,version FROM biz_activity WHERE delete_flag = 1///query by id veclet result:Vec<BizActivity> = rb.list_by_column("id",&["1".to_string()]).await.unwrap();//Query ==> SELECT create_time,delete_flag,h5_banner_img,h5_link,id,name,pc_banner_img,pc_link,remark,sort,status,version FROM biz_activity WHERE delete_flag = 1 AND id IN (?) ///query by wrapperlet w = rb.new_wrapper().eq("id","1");let r:Result<Option<BizActivity>,Error> = rb.fetch_by_wrapper(&w).await;//Query ==> SELECT create_time,delete_flag,h5_banner_img,h5_link,id,name,pc_banner_img,pc_link,remark,sort,status,version FROM biz_activity WHERE delete_flag = 1 AND id = ? ///delete
rb.remove_by_column::<BizActivity,_>("id",&"1".to_string()).await;//Exec ==> UPDATE biz_activity SET delete_flag = 0 WHERE id = 1///delete batch
rb.remove_batch_by_column::<BizActivity,_>("id",&["1".to_string(),"2".to_string()]).await;//Exec ==> UPDATE biz_activity SET delete_flag = 0 WHERE id IN ( ? , ? ) ///update///if version_lock plugin actived,update method will modify field 'version'= version + 1letmut activity = activity.clone();let w = rb.new_wrapper().eq("id","12312");
rb.update_by_wrapper(&mut activity,&w,&[Skip::Value(&serde_json::Value::Null),Skip::Column("id")]).await;//Exec ==> UPDATE biz_activity SET create_time = ? , delete_flag = ? , status = ? , version = ? WHERE id = ? }///...more usage,see crud.rs
macros (new addition)
Important update (pysql removes runtime, directly compiles to static rust code) This means that the performance of SQL generated using py_sql,html_sql is roughly similar to that of handwritten code.
Because of the compile time, the annotations need to declare the database type to be used
#[py_sql( rb,"select * from biz_activity where delete_flag = 0 if name != '': and name=#{name}")]asyncfnpy_sql_tx(rb:&Rbatis,tx_id:&String,name:&str) -> Vec<BizActivity>{todo!()}
Added html_sql support, a form of organization similar to MyBatis, to facilitate migration of Java systems to Rust(Note that it is also compiled as Rust code at build time and performs close to handwritten code) this is very faster
Because of the compile time, the annotations need to declare the database type to be used
<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.1//EN" "https://github.com/rbatis/rbatis_sql/raw/main/mybatis-3-mapper.dtd"><mapper><selectid="select_by_condition">
select * from biz_activity where
<iftest="name != ''">
name like #{name}
</if></select></mapper>
///select page must have '?:&PageRequest' arg and return 'Page<?>'#[html_sql(rb, "example/example.html")]asyncfnselect_by_condition(rb:&mutRbatisExecutor<'_>,page_req:&PageRequest,name:&str) -> Page<BizActivity>{todo!()}
#[macro_use]externcrate lazy_static;lazy_static!{static ref RB:Rbatis=Rbatis::new();
}/// Macro generates execution logic based on method definition, similar to @select dynamic SQL of Java/Mybatis/// RB is the name referenced locally by Rbatis, for example DAO ::RB, com:: XXX ::RB... Can be/// The second parameter is the standard driver SQL. Note that the corresponding database parameter mysql is? , pg is $1.../// macro auto edit method to 'pub async fn select(name: &str) -> rbatis::core::Result<BizActivity> {}'///#[sql(RB, "select * from biz_activity where id = ?")]pubasyncfnselect(name:&str) -> BizActivity{}//or: pub async fn select(name: &str) -> rbatis::core::Result<BizActivity> {}#[tokio::test]pubasyncfntest_macro(){
fast_log::init_log("requests.log",1000, log::Level::Info,None,true);RB.link("mysql://root:123456@localhost:3306/test").await.unwrap();let a = select("1").await.unwrap();println!("{:?}", a);}
How to use logical deletes plugin (works for fetching or removing functions provided by rbatis,e.g. list**(),remove**(),fetch**())
let rb = Rbatis::new();
rb.link("mysql://root:123456@localhost:3306/test").await.unwrap();letmut tx = rb.acquire_begin().await.unwrap();let v: serde_json::Value = tx
.fetch("select count(1) from biz_activity;",&vec![]).await.unwrap();println!("{}", v.clone());//rb.fetch**** and more method
tx.commit().await.unwrap();//tx.begin().await
Transaction defer
pubasyncfnforget_commit(rb:&Rbatis) -> rbatis::core::Result<()>{// tx will be commit.when func endletmut tx = rb.acquire_begin().await?.defer_async(|mut tx1| asyncmove{if !tx1.is_done(){
tx1.rollback().await;println!("tx rollback success!");}else{println!("don't need rollback!");}});let v = tx
.exec("update biz_activity set name = '6' where id = 1;",&vec![]).await;//tx.commit().await; //if commit, print 'don't need rollback!' ,if not,print 'tx rollback success!'returnOk(());}
How to use rbatis with Rust web frameworks (actix-web is used here as an example, but all web frameworks based on tokio or async_std are supported)
CRUD, with built-in CRUD template (built-in CRUD supports logical deletes)
√
LogSystem (logging component)
√
Tx(task/Nested transactions)
√
Py(using py-like statement in SQL)
√
async/await support
√
PagePlugin(Pagincation)
√
LogicDelPlugin
√
Html(xml) Compile time dynamic SQL)
√
DataBase Table ConvertPage(Web UI,Coming soon)
x
Conlusion: Assuming zero time consumed on IO, single threaded benchmark achieves 200K QPS or QPS, which is a few times
more performant than GC languages like Go or Java.
Support for DateTime and BigDecimal?
Currently supports chrono::NaiveDateTime和bigdecimal::BigDecimal
Supports for async/.await
Currently supports both async_std and tokio
Stmt in postgres uses $1, $2 instead of ? in Mysql, does this require some special treatment? No, because rbatis uses
#{} to describe parametric variabls, you only need to write the correct parameter names and do not need to match it
with the symbols used by the database.
Supports for Oracle database driver?
No, moving away from IOE is recommended.
Which crate should be depended on if only the driver is needed?
rbatis-core, Cargo.toml add rbatis-core = "*"