- 🛠 Fix a bug where rows may not be closed when bind failed in some way
- 🛠 Fix aliasing of primary key column names in exist template
- 🛠 Fix a bug in the psql driver where
double precisionwas converted into a decimal type when it indeed should be float64.
- 🛠 Fix a bug in the psql driver where
realarrays were converted into decimal arrays, now they are converted into float64 arrays. For
realthis may not be ideal but it's a better fix for now since we don't have a float32 array.
- ➕ Add alternative syntax that's case-sensitive for defining aliases. This is particularly useful if toml key syntax is not good enough or viper keeps lowercasing all your keys (and you have uppercase names in say mssql).
- ➕ Added
aliasCols, a template helper that can be used in conjunction with stringMap to transform a slice of column names into their aliased names.
- 🛠 Fix several places that referenced primary keys in the templates that were not alias aware.
- 🛠 Fix strmangle's CamelCase such that it forces the first char to be lowercase.
- 🛠 Fix a bug where relationship to_many eager had an old argument in the query call.
- 🛠 Fix a bug where relationship to_many eager call for afterSelectHooks was not using aliases.
- ✂ Remove dependency on spew package in tests for boilingcore
- 🛠 Fix a badly templated value in the mssql upsert template when using context
- 🛠 Fix a bug where the database schema name was not properly being given to the templates. This broke explicit schemas where implicit schemas like in mysql, psql public etc. would still work.
- 🛠 Fix generation failure on windows due to path manipulation issues (#314)
- 🛠 Fix an issue in mysql driver where the null json field was getting the wrong type (#311)
- 🛠 Fix compilation failures when using a nullable created_at column (#313)
- ➕ Add CockroachDB out of band driver link to readme (thanks @glerchundi)
- 🛠 Fix an issue in psql driver where if you had a primary key with the same name in two different tables, it could possibly get the wrong one (#308)
- 👍 Allow errcheck to run successfully (exclude fmt.Fprint(|ln|f) function) on models
- 🛠 Fix an issue where debug output wouldn't output on crash (exactly when you want it)
- 🛠 Fix issue with boolean values 'true' and 'false' not being accepted (thanks @glerchundi)
- ➕ Add ability to rename tables, columns, and relationships.
- ➕ Add support for rows affected to Update/Delete calls.
- ➕ Add support for geotypes for psql (thanks @saulortega)
- ➕ Add a flag to set the config file (thanks @l4u)
- ➕ Add support for citext to psql (thanks @boxofrad)
- ➕ Add virtual columns ignoring for mysql (thanks @Nykakin)
- ➕ Add ability for drivers to provide their own templates/replacement templates
- ➕ Add ability for drivers to specify imports
- ➕ Add a boil.sh that includes many commands that help build and test sqlboiler
- ➕ Add many more driver dialect flags to be able to remove all the DriverName comparisons inside sqlboiler templates. This allows us to more cleanly support more drivers.
- ➕ Add ability to override imports via config file
- ➕ Add ability to replace types using the config file
- ➕ Add ability to re-use queries with SetArgs
- ➕ Add ability to specify blacklisted or whitelisted columns in a table by using the syntax tablename.columnname in the driver's whitelist/blacklist config setting.
- ➕ Add way to create a relationship struct:
- 🛠 MySQL numeric, fixed, dec types now generate using the new types.(Null)Decimal
- 0️⃣ Use bindata as the default method of accessing templates, this prevents many bug reports we've had in the past. During development or to otherwise opt out the --templates flag will not load from bindata (drivers are the exception).
- ➕ Add the ability to generate non-go files. The new --templates flag allows you to override the default bindata templates/ and templates_test/ directory that ship with sqlboiler.
- 🏗 Export the queries.BuildQuery method for public use. This allows building queries without executing.
Drivers are now separate binaries. A lot of the reason for this is because having to keep all the drivers together inside sqlboiler was going to later cause more churn that it was worth. sqlite showed us this immediately with it's cgo dependencies as an example. This was the source of a huge number of changes, many of them breaking but other than the new workflow of using 👀 a new binary instead of just a string, users shouldn't actually see too much difference.
- ⚡️ Insert, Update and Upsert now take a
boil.Columnsinstead of a
whitelist string. This allows more flexibility in that we can use different column list kinds to control what gets inserted and updated. See
boil.Greylist. This is a breaking change to the syntax of Insert and Update as this argument is no longer optional. To migrate an app and keep the same behavior, add
boil.Infer()where there's a missing argument, and add
boil.Whitelist("columns", "here")where columns were previously specified.
- Eager loading can now accept QueryMods. This is a breaking change. To migrate
an application simply take calls that were of the form:
Load("a.b", "a.c")and break them into two separate
Load()query mods. The second argument is now a variadic slice of query mods, hence the breaking change. See the docs for
Load()for more details.
- Eager loading now attaches both sides of the relationship in the
Rstructs. This is consistent with the way the set relationship helpers work.
- Queries no longer keep a handle to the database and therefore the db parameter is no longer passed in where you create a query, instead it's passed in where you execute the query.
- context.Context is now piped through everything, this is a breaking change
syntactically. This can be controlled by the
- 📇 Rename postgresql driver to psql
- MySQL driver no longer accepts
tinyint_as_boolvia command line. Instead the flag has been inverted (
tinyint_as_int) to stop the automatic bool conversion and it's now passed in as a driver configuration setting (via the env or config).
- 🔧 Schema is no longer an sqlboiler level configuration setting. Instead it is passed back to sqlboiler when the driver is asked to produce the information for the database we're generating for. This is a breaking change.
- ⚡️ Rows affected is now returned by update and deletes by default. This is
syntactically a breaking change. This can be controlled by the
- Panic and Global variants are now hidden behind flags. This reduces the size of the generated code by 15%. This is a breaking change if you were using them.
- Randomize now counts on an interface to do proper randomization for custom types.
- 📦 Imports is now it's own package and is cleaned up significantly
- 🔄 Changed all
$.as is proper
- Drivers are now responsible for their own config validation
- 0️⃣ Drivers are now responsible for their own default config values
- Upsert is no longer a core part of sqlboiler, and is currently being provided by all three supported drivers, but other drivers may choose not to implement it.
- 🔧 Drivers can now take any configuration via environment or config files. This allows drivers to be much more configurable and there's no constraints imposed by the main/testing of sqlboiler itself.
- 🏁 Replace templates (hidden flag) now use the
:for windows compatibility.
- 🛠 Postgres numeric, fixed, decimal types now generate using the new types.(Null)Decimal instead of float64, this is a breaking change, but a very correct and important one.
- ✅ The concept of TestMain is now gone from both templates and imports. It's been superceded by the new driver abilities to supply templates and imports. The drivers add their mains to the TestSingleton templates.
- ✂ Remove the reliance on the null package in the templates. Instead favor using the sql.Scanner and driver.Valuer interface that most types will implement anyway (and the null package does too).
- 🛠 Fixed a bug in Bind() where all the given
*sql.Rowswould be consumed even in the event where we were binding to a single object. This allows for slower scanning of a
- 🛠 Fixed a problem in eager loading where the same object would be queried for multiple times. This was simply inefficient and so we now de-duplicate the ids before creating the query.
- 🛠 Fix a bug with insert where if the columns to insert were nil, values would never be loaded into the query (RETURNING clause missing).