# Unreleased
# 0.9.2
- Fixed bug in `Rows::filter_some` (it would filter out `Some(false)`).
- Added support for renaming the primary key with `#[primary_key("new_name")]`.
- Improved `#[schema]` docs.
# 0.9.1
- Renamed private feature `mutants` to `__mutants`.
- Updated `k12` from `0.3` to `0.5`.
- Added keywords to `Cargo.toml`.
# 0.9.0
The dependency removal update.
- Added support for renaming tables with foreign key constraints to them.
- Updated rust-version to `1.95`
- Updated `rusqlite` (and removed `modern_sqlite` feature).
- Removed `sea-query`, `sqlite3-parser`, `r2d2` and `r2d2-sqlite` dependencies.
- Check that tables are `STRICT`.
# 0.8.1
- Fix `rust-query` not compiling when using the `jiff-02` feature without the `bundled` feature.
# 0.8.0
- Added support for `jiff::Timestamp` and `jiff::civil::Date` with check constraints.
- Added methods `Expr::to_second`, `Expr::subsec_nanosecond`, `Expr::from_second`,
`Expr::add_nanosecond`, `Expr::to_date_in_tz`, `Expr::year`, `Expr::month`, `Expr::day`,
`Expr::first_of_month`, `Expr::add_day`.
## Breaking
- Deprecated `Expr::unix_epoch`. Create timestamps outside of rust-query instead.
- Fixed missing check constraint for `Option<bool>` columns.
For backwards compatibility, you have to change these columns in your schema to`Option<i64>`.
Then you can define a migration to change the column type to `Option<bool>` with check constraint.
Make sure to use `rust_query::migration::hash_schema` to check that you migrated correctly.
```rust
// this code uses rust-query 0.7.1
#[rust_query::migration::schema(Schema)]
pub mod vN {
pub struct Foo {
pub some_col: Option<bool>,
}
}
#[test]
fn schema_hash() {
use rust_query::migration::hash_schema;
expect_test::expect!["b5feac749e55f5bf"].assert_eq(&hash_schema::<v0::Schema>());
}
```
After migrating, the hash should be the same:
```rust
// this code uses rust-query 0.8.0
#[rust_query::migration::schema(Schema)]
pub mod vN {
pub struct Foo {
pub some_col: Option<i64>, // <-- type was changed
}
}
#[test]
fn schema_hash() {
use rust_query::migration::hash_schema;
expect_test::expect!["b5feac749e55f5bf"].assert_eq(&hash_schema::<v0::Schema>());
}
```
# 0.7.1
- Fixed `Aggregate::exists` giving wrong result when used without any query decorrelation.
- Added more documentation examples (doctests).
# 0.7.0
- Changed table types in schema and `Expr` to be `TableRow<T>` instead of `T`.
- Added support for `use` items in `#[schema]` module.
One can fix their schema by wrapping all references between tables with `TableRow` like this:
```rust
#[schema(MySchema)]
pub mod vN {
use rust_query::TableRow; // imports work!
pub struct Foo {
pub some_bar: TableRow<Bar>, // <-- here
}
pub struct Bar {
pub name: String,
pub foo: Option<TableRow<Foo>> // <-- here
}
}
```
- Made the table insert struct not generic to improve error messages.
Errors can be fixed by manually converting to the required type:
```rust
txn.insert(User {
name: "dsafdsf".to_owned(), // insert fields are not generic anymore, so we need manual conversion.
})
```
- Added `Conflict` type for insert and update with multiple unique constraints.
- Implemented `Error` for `TableRow<T>` and `Conflict`.
- Add required feature `base0` to prevent future breakage.
- Remove `dev` feature from `default`.
- Removed all deprecated methods and types.
- Fixed bug where conflicts during migration could panic or return the wrong row.
# 0.6.11
- Renamed `Expr::as_float` to `Expr::to_f64`.
- Renamed `Expr::truncate` to `Expr::to_i64`.
- Added `Expr::floor`, `Expr::ceil`, `Expr::round_with_precision`,
`Expr::replace`, `Expr::trim`, `Expr::ltrim` and `Expr::rtrim`.
# 0.6.10
- Allow `Transaction::downgrade` in `Migrator::fixup`.
- Deprecated `Config::init_stmt`.
- Improved documentation, added docs for `#[index]`.
# 0.6.9
- Added `Mutable::unique` to mutate unique columns.
- Deprecated `Transaction::update`.
- Added `Migrator::fixup` to mutate the database after a migration and within
the same transaction.
# 0.6.8
- Allow mutating columns that are used in an index.
- Implement `IntoExpr` for `Lazy`.
- Add some more documentation.
# 0.6.7
- Fix to allow implicit reference to primary key in sqlite schema.
- Switch to `sqlite3-parser` for lexing the sqlite schema.
- Fix to allow unquoted column names in the schema.
- Recreate full table on index change.
This is required when a unique index is defined in a table.
- Added cargo-mutants support.
- Fixed bug that would make a unique constraint violation in a migration into a panic.
- Allow changing indices on individual fields without new table.
# 0.6.6
- Check constraints from the database are now verified against the declared schema.
- Added support for booleans in schemas.
This uses a check constraint of the form `"col" IN (0, 1)`.
- Added `Query::order_by` to order rows in queries.
# 0.6.5
- Fix `DatabaseAsync` to allow the waker to change.
# 0.6.4
- Added `Transaction::mutable` and `Transaction::mutable_vec`.
- Deprecated `Transaction::update_ok`.
- Fix, only allow `Transaction::lazy` on table valued expressions.
- Added `Expr::div`, `Expr::modulo`, `Expr::concat`, `Expr::max`, `Expr::min`,
`Expr::truncate`, `Expr::lower`, `Expr::upper`, `Expr::sign`, `Expr::between`,
`Expr::abs`, `Expr::zero_blob`, `Expr::unix_epoch`, `Expr::char_len` and
`Expr::byte_len`.
- Deprecated `UnixEpoch`.
- Relaxed trait bound on `Aggregate::max` and `Aggregate::min` to allow table typed
expressions.
Example usage of the new `Transaction::mutable` API:
```rust
// old
txn.update_ok(
&order.customer,
Customer {
balance: Update::add(total_amount),
delivery_cnt: Update::add(1),
..Default::default()
},
);
// new
let mut customer = txn.mutable(&order.customer);
customer.balance += total_amount;
customer.delivery_cnt += 1;
drop(customer);
```
# 0.6.3
- Added `DatabaseAsync` to run transaction asynchronously on any runtime.
- Added some reuse of connections between transactions.
- Added `Database::new`, to create database without migrations.
# 0.6.2
- Added diagnostics for differences between rust code and database schema.
The new diagnostics use the `annotate-snippets` crate to annotate the rust code.
An example error looks like this:
```
error: Unique constraint mismatch for `#[version(0)]`
╭▸ src/schema/test.rs:146:15
│
LL │ #[unique(baz, field2)]
│ ━━━━━━ database does not have this unique constraint
LL │ pub struct Foo {
╰╴ ━━━ database has `#[unique(baz, field1)]`
```
- Improved schema reading code to be more flexible.
# 0.6.1
- Added automatic addition and removal of column indices without a new schema version.
# 0.6.0
## Breaking changes
- Unique constraints are now unnamed.
Using a unique constraint can be done by using the column names in order e.g.
`Stock.warehouse(w).item(i)` instead of `Stock::unique(w, i)`.
- Support for only filtering on some columns with unique constraint syntax e.g.
you can do `rows.join(Stock.warehouse(w))`, which will join all rows from the Stock
table that match the warehouse.
- To create an empty row you now have to use `txn.insert_ok(v0::Empty {})` instead or
`txn.insert_ok(v0::Empty)`.
- `Optional::then` is renamed to `Optional::then_select`.
- `Optional::then_expr` is renamed to `Optional::then`.
- Migrations now use the `Lazy` type instead of a generic type `T: FromExpr`.
## Added
- `Transaction::lazy` and `Transaction::lazy_iter` are added, these methods
return rows of type `Lazy<'t, Table>`, which lazily queries values when they
are accessed.
- `Transaction::lazy_iter` accepts the same kind of argument as `Rows::join`.
So any table, optionally filtered by an index can be queried. For example:
`txn.lazy_iter(Post.author(my_user))` would iterate over all posts by `my_user`.
- `Query::into_iter` now returns an iterator that can be moved outside the
`Transaction::query`. This makes it possible to return the iterator as a function result.
- `Optional::and_then` was added as a convenient way to combine `Optional::and`
with `Optional::then`.
- Support for extra indices with the `#[index]` attribute. This works exactly
like the `#[unique]` attribute, but it doesn't have a unique constraint.
Defining an index will also add extra methods to filter on the indexed columns.
## Removed
- The generated macros for querying specific columns from tables were removed.
This also means the removal of `MacroRoot`, which was only used by these macros.
`Lazy` should be used instead.
# 0.5.2
- Add option to configure `foreign_keys`.
- Simplify generated query without joins.
- Preserve column definition order when creating unique constraint.
This lets the user choose the ordering, allowing the unique constraint to be used as a covering index.
- Add lock for mutable transactions to fix transaction timeout under load.
The lock is dropped before committing to allow the next mutable transaction to start.
- Optimize `LEFT JOIN` to `JOIN` when the joined row is guaranteed to exist.
This allows sqlite to reorder more joins for faster execution plans.
# 0.5.1
- Changed default `synchronous` to `FULL`.
- Added the option to configure `synchronous` to `NORMAL`.
- Fixed panic propagation from transaction closures.
- Pinned sea-query release candidate version.
# 0.5.0
## Changed table column syntax
- Instead of methods `artist.name()`, you should now use fields `&artist.name`.
- `TableRow` does not have support for accessing columns anymore, instead convert the `TableRow` to an `Expr` using `IntoExpr`.
- Removed `ref_cast` dependency.
## All transactions now run on separate threads
- Removed all transaction lifetimes (`TableRow` no longer has a lifetime).
- Removed `LocalClient` (methods have been moved to `Database`).
- `Database::transaction` and `Database::transaction_mut` now accept a closure to run on a new thread.
- Removed `TransactionMut::commit` (commit now depends on the result returned from the transaction).
- Added `Database::transaction_mut_ok` for when the transaction is always commited.
- Removed `TransactionMut`, it is replaced by `&mut Transaction`.
## Other
- Removed deprecated `Table::join`.
- Removed deprecated `IntoSelectExt` (wit the `map_select` method).
- Removed deprecated `Aggregate::filter_on`.
- Updated to rusqlite 0.37
# 0.4.4
- Add support for doc comments on tables and columns.
- Add `Query::into_iter`, to lazily iterate over query results.
# 0.4.3
- Fix panic when inserting into table without columns.
- Add `Select::map` method.
- Deprecate `IntoSelectExt::map_select`.
- Deprecate `Aggregate::join_on`.
# 0.4.2
- Update the `Rows::join` method to take a constant argument.
This is now the prefered join syntax and all examples have been updated.
- Allow arbitrary correlated subqueries.
This means that `Aggregate` now has an implied bound that allows leaking `Expr` from the
out scope. Correlated subqueries are decorrelated before translating to SQL.
- Fix loose lifetime on `Optional`.
# 0.4.1
- Change conflicts back to using `TableRow` instead of `Expr`.
Changing the conflict type to `Expr` was a mistake, because the `Expr` can be invalidated.
- Fix `#[schema]` macro not showing errors for unique constraints.
# 0.4.0
Blog post: https://blog.lucasholten.com/rust-query-0-4/
## Optional Queries
- Added `optional` combinator.
- Changed `Expr` to be co-variant in its lifetime.
## Basic Datatypes and Operations
- Added support for `Vec<u8>` data type (sqlite `BLOB`).
- Added some more basic operations on expressions.
## Updates, Insert and Query
- Added safe updates of a subset of columns for each table.
- Update statements now use the `Update` type for each column.
- Insert and update conflict is now an `Expr` (`find_or_insert` returns an `Expr` now too).
- `Rows::into_vec` is no longer sorted automatically.
## Schema and Mirations
- Changed `#[schema]` syntax to be a module of structs.
- Added `#[from]` attribute to allow renaming tables and splitting tables in the schema.
- The generated migration structs have moved from e.g. `v1::update::UserMigration` to `v0::migrate::User`.
- Migrations now require explicit handling of potential unique constraint violations.
- Migrations now require explicit handling of foreign key violations.
## Type Driven Select
- Added a macro for each table to create ad-hoc column selection types like `User!(name, age)`.
- Added the `FromExpr` trait to allow custom column selection and conversion.
## Feature Flags and Dependencies
- `TransactionWeak::rusqlite_transaction` is renamed and no longer behind a feature flag.
- `hash_schema` method was moved behind `dev` feature which is enabled by default.
- Updated dependencies.
## Renaming
- Renamed `Dummy` to `Select`.
- Renamed `Column` to `Expr`.
- Renamed `try_insert` to `insert` and `insert` to `insert_ok`.
- Renamed `try_delete` to `delete` and `delete` to `delete_ok`.
- Renamed `try_update` to `update` and `update` to `update_ok`.
# 0.3.1
- Added error message when defining an `id` column.
- Added support for sqlite `LIKE` and `GLOB` operators (Contributed by @teamplayer3).
- Added support for `DELETE` using `TransactionWeak` and `#[no_reference]`.
- Added `TransactionWeak::unchecked_transaction` behind feature flag.
- Added `impl ToSql for TableRow` behind `unchecked_transaction` feature flag.
- Removed `impl RefCast for Transaction`, it was not intended to be public.
- Removed `impl FromSql for TableRow`, it was not intended to be public.
# 0.3.0
- Added support for updating rows.
- Added `Table::dummy` method, which makes it easier to do partial updates.
- Reused table types in the generated API for both naming `TableRow<User>` and dummies `User {name: "steve"}`.
- Forbid `Option` in unique constraints.
- Renamed `ThreadToken` to `LocalClient`.
- Renamed and moved `read` and `write_lock` to `transaction` and `transaction_mut`.
- Check `schema_version` at the start of every transaction.
- Simplify migration and borrow `LocalClient` only once.
- Renamed `Prepare` to `Config` and simplified its API.
# 0.2.2
- Bound the lifetime of `TableRow: IntoColumn` to the lifetime of the transaction.
Without the bound it was possible to sneak `TableRow`s into following transacions. <details>
`query_one` now checks that its input lives for as long as the transaction.
To make sure that `query_one` still checks that the dummy is "global", the transaction now has an invariant lifetime.
</details>
# 0.2.1
- Relax `Transaction` creation to not borrow the `Database`.
- Add missing lifetime bound on `try_insert`s return value.
Technically this is a breaking change, but it fixes a bug so it is still a patch release.
- Fix the version of the macro crate exactly (=0.2.0) to allow future internal API changes with only a patch release.
# 0.2.0
- Rewrote almost the whole library to specify the schema using enum syntax with a proc macro.
- Added a single Column type to handle a lot of query building.
- Dummy trait to retrieve multiple values at once and allow post processing.
- Added support for transactions and multiple schemas.
# 0.1.x
- This version was SQL schema first. It would generate the API based on the schema read from the database.