All Versions
43
Latest Version
Avg Release Cycle
33 days
Latest Release
1284 days ago

Changelog History
Page 3

  • v0.7.1

    October 04, 2019
  • v0.7.0 Changes

    May 26, 2019

    This is a compatibility-breaking release, although most source code should compile unmodified.

    The big changes are improvements in resource safety:

    • Replaced MonadError-based guarantee with bracket, which prevents potential resource leakage when using cancelable IO; and improves transactor resource safety by using Resource. The migration guide contains slightly more information. Thanks Sam Guymer for these updates!
    • ⚡️ Fragment concatenation and derived combinators like Fragments.in are now stacksafe, so you can now have gigantic IN clauses. This required API changes in Param, Fragment, Update[0] and Query[0] but these should not affect most users.

    🆕 New features:

    • At long last there is a doobie-quill module that allows constructing ConnectionIO values via Quill quotes. See the new book chapter for examples. Many thanks to Alex Ioffe and Chris Davenport for their assistance!

    And an assortment of other improvements:

    • Postgis has been upgraded to v2.3.0. Thanks Erlend Hamnaberg!
    • Added support for Scala 2.13.0-M5, many thanks to Sam Guymer for setting this up!
    • TypeTag constraint on Analyzable has been relaxed to WeakTypeTag. Thanks nigredo-tori!
    • New syntax allows .transact on OptionT and EitherT of ConnectionIO, rolling back in the failure cases. Thanks Julien Truffaut!
    • Postgres now supports arrays of Scala BigDecimal. Thanks Nicolas Rémond!
    • Most calls to throw have been replaced with raiseError, which allows doobie to work with effect types that don't catch. Thanks Sam Guymer!
    • Some things in the build were preventing doobie from cooperating with the community build. This has been fixed, thanks Seth Tisue!
    • Text[Char] instance was added for Postgres COPY FROM STDIN. Thanks Dermot Haughey!
    • You can now create a HikariTransactor from a HikariConfig. Thanks Yuriy Badalyantc!
    • Thanks to Harry Laoulakos and Daan Hoogenboezem for documentation updates!
    • Fragment combinators now parenthesize arguments so associativity in Scala is reflected in generated SQL. Thanks Katrix!

  • v0.7.0-M5

    May 07, 2019
  • v0.7.0-M4

    April 20, 2019
  • v0.7.0-M3

    February 19, 2019
  • v0.7.0-M2

    January 17, 2019
  • v0.7.0-M1

    January 04, 2019
  • v0.6.0 Changes

    Many thanks to Arber Shabhasa, Bjørn Madsen, Chris Davenport, Cody Allen, Dmitry Polienko, Kai(luo) Wang, Kevin Walter, Mark Canlas, and Quang Le Hong for their contributions to this release.

    💥 :exclamation: This is a major update with breaking changes. Please read the following notes carefully.

    Transactors and Threading

    Prior to 0.6.x doobie had nothing to say about threading; it was up to users to shift interpreted IO programs onto dedicated pools if desired. This has changed. We now identify three distinct execution contexts that are relevant to database applications.

    1. All non-blocking work is performed on the execution context identified by ContextShift[F]. If you are using IOApp (which you should) this instance is provided for you. All interpreters (and thus all transactors) need this instance on construction.
    2. Requesting a JDBC connection is a blocking operation, so to avoid deadlock these requests cannot be placed in competition with running programs (which need to make progress in order to finish up and return their connections to the pool). Therefore we need a distinct, bounded, blocking execution context for the single purpose of awaiting database connections. All transactors that use a connection pool will require this execution context to be specified. ⏱ 1. All JDBC primitive operations are [potentially] blocking, so we need a distinct, typically unbounded (since the connection context above provides a logical bound) execution context for scheduling these operations. All transactors that use a connection pool will require this execution context to be specified.

    ✅ For convenience we provide an ExecutionContexts module that provides Resources yielding the kinds of ExecutionContexts that will tend to be useful for the scenarios above. Note that DriverManagerTransactor provides an unbounded number of connections and is unsuitable for production use anyway, so we do not require the blocking pools here (it uses an unbounded pool internally). Fine for testing but don't use it in real life.

    See the book chapter on Managing Connections for more information and examples.

    Streams and Resources

    In 0.5.x we provided some single-element Streams that would emit values and guarantee resource cleanup. This has been generalized as Resource in cats-effect 1.x, and all such constructors (H2Transactor.newH2Transactor for example) now use Resource rather than Stream. You can use Stream.resource(rsrc) to regain the old functionality if desired.

    Splitting of read/write functionality.

    Prior to the 0.6.x series we provided two typeclasses for bidirectional type mapping:

    • Meta defined nullable mappings between column/parameter values and scala types.
    • Composite defined null-safe mappings between column/parameter vectors and scala types.

    Starting with version 0.6.0 type mappings are unidirectional:

    • Meta has been split into Get and Put typeclasses, for reads and writes of column/parameter values, respectively.
    • Composite has been split into Read and Write typeclasses, for reads and writes of column/parameter vectors, respecitively.

    👍 Note that Meta does still exist, but only as a mechanism for introducing Get/Put pairs. An implicit Meta[A] induces both an implicit Get[A] and an implicit Put[A], and the old mechanism of Meta[A].imap(...)(...) is still supported for this purpose. The xmap method has been replaced with parametric imap and TypeTag-constrained timap. Prefer timap when possible because it yields better diagnostic information when typechecking queries.

    To summarize:

    0.5.x 0.6.x Notes
    [A: Meta] [A: Get : Put] Or just one, depending on usage.
    [A: Composite] [A: Read : Write] Or just one, depending on usage.
    Meta[A].xmap(..) Meta[A].timap(...) Or imap when a TypeTag is unavailable.
    Composite[A].xmap(...) Read[A].map(...) Write[A].contramap(...) This takes two steps now.

    Please refer to book chapter on Custom Mappings and the examples project for more details.

    Other Changes
    • postgres-circe module created exposing Get and Put instances for io.circe.Json
    • ⬆️ upgraded to cats-effect 1.0 and fs2 1.0

  • v0.5.4

    December 07, 2018
  • v0.5.3 Changes

    ⚡️ Minor updates, see below.

    • Updated to refined 0.9, fs2 0.10.4, Hikari 3.1.0, ScalaCheck 1.14.0, and Specs2 4.2.0.
    • ➕ Added .execWith method to Fragment, allowing for custom handling of the associated statement.
    • ➕ Added pure to generated algebras, so you can now say FC.pure(1) instead of 1.pure[ConnectionIO] if you like. Thanks wedens!
    • ➕ Added an example of generic DAOs for simple schemas. See Orm.scala in the example project.
    • ⚡️ Minor doc updates.

    🚀 This is the last planned release in the 0.5.x series. Work will start soon on 0.6!