GitLab

This is not a permament solution and we'll add a test harness soon
that executes tests against an actual database, but for now this will do.

Arve Knudsen authored Aug 22, 2019 and Katharina Fey committed Jan 06, 2020

Signed-off-by: Arve Knudsen <arve.knudsen@gmail.com>

The previous builder API was very suboptimal because it required
database-specific string escaping on the side of the user. Instead
`barrel` now takes a table to reference as a string, then either a
single column name or a vector of columns.
This is currently still "stringly-typed" but good enough for now!

It also adds a `foreign_schema` function which can be passed a schema
that is _different_ from the current table schema.

This PR changes the way that the `types::primary()` builder works and
adds a `primary(bool)` function to the Type pattern. The `PRIMARY`
base type still exists, but is handled seperately, for Pg reasons
(should be removed next version).  All other types that use the
`primary(...)` function will be able to become `PRIMARY KEY` in the
generated SQL.

Additionally this PR fixes the way that `WrappedDefault` values are
encoded. i.e. it is now possible to properly specify a default value
that is then mapped into the apropriate type.

The previous builder API was very suboptimal because it required
database-specific string escaping on the side of the user. Instead
`barrel` now takes a table to reference as a string, then either a
single column name or a vector of columns.
This is currently still "stringly-typed" but good enough for now!

It also adds a `foreign_schema` function which can be passed a schema
that is _different_ from the current table schema.

This PR changes the way that the `types::primary()` builder works and
adds a `primary(bool)` function to the Type pattern. The `PRIMARY`
base type still exists, but is handled seperately, for Pg reasons
(should be removed next version).  All other types that use the
`primary(...)` function will be able to become `PRIMARY KEY` in the
generated SQL.

Additionally this PR fixes the way that `WrappedDefault` values are
encoded. i.e. it is now possible to properly specify a default value
that is then mapped into the apropriate type.

mikong authored Aug 19, 2019 and Katharina Fey committed Sep 09, 2019

* Remove incorrect double quotation marks from the expectation values
* Remove unnecessary PRECISION from the expectation value of `double()`
* Remove the `uuid()` test as it is `unimplemented!()`

mikong authored Aug 19, 2019 and Katharina Fey committed Sep 09, 2019

* Update comments to refer to mysql instead of pgsql or sqlite3
* Fix unresolved import errors by updating `use` statements
* Fix `types::UUID` error

Eric Dattore authored Jun 07, 2019 and Katharina Fey committed Jun 13, 2019

The tmp package wouldn't build because the feature specified by the
backend wasn't properly wrapped in quotes. Cargo would complain about
parsing the `Cargo.toml` file.