Tokio based asynchronous MySql client library for The Rust Programming Language.
The library is hosted on crates.io.
[dependencies]
mysql_async = "<desired version>"
Default feature set is wide – it includes all default mysql_common
features
as well as native-tls
-based TLS support.
-
minimal
– enables only necessary features (at the moment the only necessary feature isflate2
backend). Enables:- `flate2/zlib"
Example:
[dependencies] mysql_async = { version = "*", default-features = false, features = ["minimal"]}
*Note: it is possible to use another
flate2
backend by directly choosing it:[dependencies] mysql_async = { version = "*", default-features = false } flate2 = { version = "*", default-features = false, features = ["rust_backend"] }
-
default
– enables the following set of crate's and dependencies' features:native-tls-tls
- `flate2/zlib"
mysql_common/bigdecimal03
mysql_common/rust_decimal
mysql_common/time03
mysql_common/uuid
mysql_common/frunk
-
default-rustls
– same as default but withrustls-tls
instead ofnative-tls-tls
.Example:
[dependencies] mysql_async = { version = "*", default-features = false, features = ["default-rustls"] }
-
native-tls-tls
– enablesnative-tls
-based TLS support (conflicts withrustls-tls
)Example:
[dependencies] mysql_async = { version = "*", default-features = false, features = ["native-tls-tls"] }
-
rustls-tls
– enablesnative-tls
-based TLS support (conflicts withnative-tls-tls
)Example:
[dependencies] mysql_async = { version = "*", default-features = false, features = ["rustls-tls"] }
-
tracing
– enables instrumentation viatracing
package.Primary operations (
query
,prepare
,exec
) are instrumented atINFO
level. Remaining operations, incl.get_conn
, are instrumented atDEBUG
level. Also atDEBUG
, the SQL queries and parameters are added to thequery
,prepare
andexec
spans. Also some internal queries are instrumented atTRACE
level.Example:
[dependencies] mysql_async = { version = "*", features = ["tracing"] }
SSL support comes in two flavors:
-
Based on native-tls – this is the default option, that usually works without pitfalls (see the
native-tls-tls
crate feature). -
Based on rustls – TLS backend written in Rust (see the
rustls-tls
crate feature).Please also note a few things about rustls:
- it will fail if you'll try to connect to the server by its IP address, hostname is required;
- it, most likely, won't work on windows, at least with default server certs, generated by the MySql installer.
There is a set of url-parameters supported by the driver (see documentation on [Opts
]).
use mysql_async::prelude::*;
#[derive(Debug, PartialEq, Eq, Clone)]
struct Payment {
customer_id: i32,
amount: i32,
account_name: Option<String>,
}
#[tokio::main]
async fn main() -> Result<()> {
let payments = vec![
Payment { customer_id: 1, amount: 2, account_name: None },
Payment { customer_id: 3, amount: 4, account_name: Some("foo".into()) },
Payment { customer_id: 5, amount: 6, account_name: None },
Payment { customer_id: 7, amount: 8, account_name: None },
Payment { customer_id: 9, amount: 10, account_name: Some("bar".into()) },
];
let database_url = /* ... */
# get_opts();
let pool = mysql_async::Pool::new(database_url);
let mut conn = pool.get_conn().await?;
// Create a temporary table
r"CREATE TEMPORARY TABLE payment (
customer_id int not null,
amount int not null,
account_name text
)".ignore(&mut conn).await?;
// Save payments
r"INSERT INTO payment (customer_id, amount, account_name)
VALUES (:customer_id, :amount, :account_name)"
.with(payments.iter().map(|payment| params! {
"customer_id" => payment.customer_id,
"amount" => payment.amount,
"account_name" => payment.account_name.as_ref(),
}))
.batch(&mut conn)
.await?;
// Load payments from the database. Type inference will work here.
let loaded_payments = "SELECT customer_id, amount, account_name FROM payment"
.with(())
.map(&mut conn, |(customer_id, amount, account_name)| Payment { customer_id, amount, account_name })
.await?;
// Dropped connection will go to the pool
drop(conn);
// The Pool must be disconnected explicitly because
// it's an asynchronous operation.
pool.disconnect().await?;
assert_eq!(loaded_payments, payments);
// the async fn returns Result, so
Ok(())
}
The [Pool
] structure is an asynchronous connection pool.
Please note:
- [
Pool
] is a smart pointer – each clone will point to the same pool instance. - [
Pool
] isSend + Sync + 'static
– feel free to pass it around. - use [
Pool::disconnect
] to gracefuly close the pool. - [
Pool::new
] is lazy and won't assert server availability.
[Conn::start_transaction
] is a wrapper, that starts with START TRANSACTION
and ends with COMMIT
or ROLLBACK
.
Dropped transaction will be implicitly rolled back if it wasn't explicitly committed or rolled back. Note that this behaviour will be triggered by a pool (on conn drop) or by the next query, i.e. may be delayed.
API won't allow you to run nested transactions because some statements causes
an implicit commit (START TRANSACTION
is one of them), so this behavior
is chosen as less error prone.
This enumeration represents the raw value of a MySql cell. Library offers conversion between
Value
and different rust types via FromValue
trait described below.
This trait is reexported from mysql_common create. Please refer to its crate docs for the list of supported conversions.
Trait offers conversion in two flavours:
-
from_value(Value) -> T
- convenient, but panicking conversion.Note, that for any variant of
Value
there exist a type, that fully covers its domain, i.e. for any variant ofValue
there existT: FromValue
such thatfrom_value
will never panic. This means, that if your database schema is known, than it's possible to write your application using onlyfrom_value
with no fear of runtime panic.Also note, that some convertions may fail even though the type seem sufficient, e.g. in case of invalid dates (see sql mode).
-
from_value_opt(Value) -> Option<T>
- non-panicking, but less convenient conversion.This function is useful to probe conversion in cases, where source database schema is unknown.
MySql text protocol is implemented in the set of Queryable::query*
methods
and in the [prelude::Query
] trait if query is [prelude::AsQuery
].
It's useful when your query doesn't have parameters.
Note: All values of a text protocol result set will be encoded as strings by the server,
so from_value
conversion may lead to additional parsing costs.
MySql binary protocol is implemented in the set of exec*
methods,
defined on the [prelude::Queryable
] trait and in the [prelude::Query
]
trait if query is [QueryWithParams
]. Prepared statements is the only way to
pass rust value to the MySql server. MySql uses ?
symbol as a parameter placeholder.
Note: it's only possible to use parameters where a single MySql value
is expected, i.e. you can't execute something like SELECT ... WHERE id IN ?
with a vector as a parameter. You'll need to build a query that looks like
SELECT ... WHERE id IN (?, ?, ...)
and to pass each vector element as
a parameter.
MySql itself doesn't have named parameters support, so it's implemented on the client side.
One should use :name
as a placeholder syntax for a named parameter. Named parameters uses
the following naming convention:
- parameter name must start with either
_
ora..z
- parameter name may continue with
_
,a..z
and0..9
Note: this rules mean that, say, the statment SELECT :fooBar
will be translated
to SELECT ?Bar
so please be careful.
Named parameters may be repeated within the statement, e.g SELECT :foo, :foo
will require
a single named parameter foo
that will be repeated on the corresponding positions during
statement execution.
One should use the params!
macro to build parameters for execution.
Note: Positional and named parameters can't be mixed within the single statement.
In MySql each prepared statement belongs to a particular connection and can't be executed
on another connection. Trying to do so will lead to an error. The driver won't tie statement
to its connection in any way, but one can look on to the connection id, contained
in the [Statement
] structure.
Warning: You should be aware of Security Considerations for LOAD DATA LOCAL.
There are two flavors of LOCAL INFILE handlers – global and local.
I case of a LOCAL INFILE request from the server the driver will try to find a handler for it:
- It'll try to use local handler installed on the connection, if any;
- It'll try to use global handler, specified via [
OptsBuilder::local_infile_handler
], if any; - It will emit [
LocalInfileError::NoHandler
] if no handlers found.
The purpose of a handler (local or global) is to return [InfileData
].
See [prelude::GlobalHandler
].
Simply speaking the global handler is an async function that takes a file name (as &[u8]
)
and returns Result<InfileData>
.
You can set it up using [OptsBuilder::local_infile_handler
]. Server will use it if there is no
local handler installed for the connection. This handler might be called multiple times.
Examles:
- [
WhiteListFsHandler
] is a global handler. - Every
T: Fn(&[u8]) -> BoxFuture<'static, Result<InfileData, LocalInfileError>>
is a global handler.
Simply speaking the local handler is a future, that returns Result<InfileData>
.
This is a one-time handler – it's consumed after use. You can set it up using
[Conn::set_infile_handler
]. This handler have priority over global handler.
Worth noting:
impl Drop for Conn
will clear local handler, i.e. handler will be removed when connection is returned to aPool
.- [
Conn::reset
] will clear local handler.
Example:
#
let pool = mysql_async::Pool::new(database_url);
let mut conn = pool.get_conn().await?;
"CREATE TEMPORARY TABLE tmp (id INT, val TEXT)".ignore(&mut conn).await?;
// We are going to call `LOAD DATA LOCAL` so let's setup a one-time handler.
conn.set_infile_handler(async move {
// We need to return a stream of `io::Result<Bytes>`
Ok(stream::iter([Bytes::from("1,a\r\n"), Bytes::from("2,b\r\n3,c")]).map(Ok).boxed())
});
let result = r#"LOAD DATA LOCAL INFILE 'whatever'
INTO TABLE `tmp`
FIELDS TERMINATED BY ',' ENCLOSED BY '\"'
LINES TERMINATED BY '\r\n'"#.ignore(&mut conn).await;
match result {
Ok(()) => (),
Err(Error::Server(ref err)) if err.code == 1148 => {
// The used command is not allowed with this MySQL version
return Ok(());
},
Err(Error::Server(ref err)) if err.code == 3948 => {
// Loading local data is disabled;
// this must be enabled on both the client and the server
return Ok(());
}
e @ Err(_) => e.unwrap(),
}
// Now let's verify the result
let result: Vec<(u32, String)> = conn.query("SELECT * FROM tmp ORDER BY id ASC").await?;
assert_eq!(
result,
vec![(1, "a".into()), (2, "b".into()), (3, "c".into())]
);
drop(conn);
pool.disconnect().await?;
Tests uses followin environment variables:
DATABASE_URL
– defaults tomysql://root:password@127.0.0.1:3307/mysql
COMPRESS
– set to1
ortrue
to enable compression for testsSSL
– set to1
ortrue
to enable TLS for tests
You can run a test server using doker. Please note that params related
to max allowed packet, local-infile and binary logging are required
to properly run tests (please refer to azure-pipelines.yml
):
docker run -d --name container \
-v `pwd`:/root \
-p 3307:3306 \
-e MYSQL_ROOT_PASSWORD=password \
mysql:8.0 \
--max-allowed-packet=36700160 \
--local-infile \
--log-bin=mysql-bin \
--log-slave-updates \
--gtid_mode=ON \
--enforce_gtid_consistency=ON \
--server-id=1
Available here
Licensed under either of
- Apache License, Version 2.0, (LICENSE-APACHE or https://www.apache.org/licenses/LICENSE-2.0)
- MIT license (LICENSE-MIT or https://opensource.org/licenses/MIT)
at your option.
Unless you explicitly state otherwise, any contribution intentionally submitted for inclusion in the work by you, as defined in the Apache-2.0 license, shall be dual licensed as above, without any additional terms or conditions.