ecto_interface

# Changelog

## 6.1.1

- [minor] Losen the restrictions on postgres versions

## 6.1.0

- [feature] A new set of reader functions that can be used as part of an Ecto.Multi chain
- [bug] random\_\* functions weren't fully using the keywords they were given and also formed their queries awkwardly

## 6.0.2

- [patch] `Module.safe_concat` doesn't account for load order issues that you might encounter, so I switched to the unsafe variant

## 6.0.1

- [patch] Massively reduce the compile-time dependency costs of using ecto interface

## 6.0.0

- [breaking] Currently whenever a subscription is called N-times there are, exectedly, N-times as many event handler calls. However given the goals of the library (to provide simple interfaces for commonly defined functionality) it's more expected to unsubscribe from all listeners first and then subscribe.

## 5.0.3

- [bug] Plural subscription without tenancy weren't actually subscribing
- [bug] Documentation around handle_info/2 definitions weren't correct

## 5.0.2

- [bug] broadcast_resources_insert/2 had a bad arguement causing confusion in the pubsub call

## 5.0.1

- [bug] We were accidentally listening to a different key than publishing, that has now been fixed

## 5.0.0

- [breaking] To reduce compile time depenencies we've made it so you no longer have to reference the entire module name in the `source` property, now it's just the name of the module itself.

## 4.3.0

- [feature] Allow for broadcast_resources_event(event, key, options) which enables special event broadcasts other than inserted or changed.

## 4.2.0

- [feature] Allow for pagination to deal with "last page" functionality.
- [bug] Allow random functions to receive Ecto.Repo options.
- [bug] Fixed some documentation misconceptions.

## 4.1.1

- [bug] The subscription channel for an individual item change with options and the broadcast of the same parameters weren't the same. I've now corrected them.

## 4.1.0

- [feature] Now you can subscribe and broadcast to specific tenant keys and/or primary keys (Detailed below)

```elixir
# For the below examples `charge.id == 1` and `charge.merchant_slug == :johns_tackle_shop`.

# This will subscribe to a topic with the id given, ie: `App.Transactions/charges:1`
App.Transactions.subscribe_to_charge(charge.id)

# And to broadcast:
App.Transactions.broadcast_charges_insert(charge.id)
App.Transactions.broadcast_charges_change(charge.id)

# You can also subscribe to charges with a specific set of options
App.Transactions.subscribe_to_charges(tenant: charge.merchant_slug)
App.Transactions.subscribe_to_charges(tenant: charge.merchant_slug, prefix: "sandbox")

# This will publish a message of `{:changed, {:charges, 1}}` to `App.Transactions/charges:1` and `App.Transactions/charges/merchant_slug:johns_tackle_shop`
App.Transactions.broadcast_charges_change(charge.id, tenant: charge.merchant_slug, prefix: "sandbox")
```

## 4.0.1

- [fix] If a module didn't define a changeset/2 then you would get a compilation warning, now you get a runtime warning.

## 4.0.0

- [breaking] Removing `new_*` and `change_*` as they stopped fitting once I added the `create_*_by` type functionality. You should just use your changesets.
- [breaking] Removing insert/update changeset function options and instead either default to `changeset/2` or whatever you pass to `*_*_by`.
- [breaking] [feature] Refactored the options for the `use(EctoInterface)` calls to use keywords because we want to be able to pass in repo on a per-definintion basis. This is especially useful for multi-database applications.
- [feature] Instead of calling `use(EctoInterface.Pubsub)` seperately you can now pass `pubsub: Core.PubSub`, `tagged: :tags`, and `slug: :username` respectively to setup those three extensions.

## 3.1.0

- [feature] Allow a per-configuration per-module definition for repo in addition to the default repo.

## 3.0.0

- [breaking] [feature] Now when you pass preloads to `create` or `update` it will then preload after the insert/update as well.

## 2.3.1

- [fix] Both `change_#{singular}` and `new_#{singular}` were incorrectly still enforcing a `map()` type for the value.

## 2.3.0

- [feature] Prior understandings of Ecto's "changeset" functions were that they took a record and a mapping of attributes. My understanding has grown since then, for example compounding changesets makes sense and also no second (or multiple) arguments. I've at least removed the restriction of it being a map for the second arugment. You can now do: `Core.Users.create_account(:pending)` and it should be compatible.

## 2.2.1

- [fix] Relax the requirements of the optional postgrex version

## 2.2.0

- [feature] Improved type annotation for the create functions
- [fix] Now you can pass repository options to `paginate_` functions whereas before they were swallowed

## 2.1.0

- [feature] Now includes pagination functionality via `paginated_*/*`, much like `stream_*`. This is incorporated into the library by copying the Paginator code. I wanted to just use the library as normal but it's kinda outdated.
- [breaking fix] `new_*/1` and `new_*/2` had conflicts, now there is only `new_customer(record, attributes, changeset_function)` and `new_customer(attributes, changeset_function)`
- [breaking fix] `stream_*_by/2` wasn't actually calling `stream/0`.
- [fix] Multiple specs are now correctly showing the options variant

## 2.0.0

- [feature] All functions that can now support the `options` optional argument, which gets passed to `Repo.all/*`, `Repo.one/*`, `Repo.get/*`, `Repo.stream/*`, `Repo.insert/*`, `Repo.update/*`, and `Repo.delete/*`. This is useful for: Timeouts, prefixes, and more.
- [breaking] All `Read`, `Tagged` and `Slug` functions that previously took `subquery` are now named `_by/`, for example `Core.Users.list_accounts(&where(&1, name: "Kurtis Rainbolt-Greene"))` now becomes `Core.Users.list_accounts_by(&where(&1, name: "Kurtis Rainbolt-Greene"))`.
- [breaking] All `Write` functions that previously took custom changeset functions are now named `_by/`, for example `Core.Users.create_account(&Core.Users.Account.update_name_changeset/2, %{name: "Kurtis Rainbolt-Greene"})` now becomes `Core.Users.create_account_by(%{name: "Kurtis Rainbolt-Greene"}, &Core.Users.Account.update_name_changeset/2)`.
- [breaking] I've removed the `get_*_by(Keyword.t())` function because it doesn't make sense given the `subquery` enabled read function.
- [breaking] I've moved `EctoInterface.Read.PubSub` to `EctoInterface.PubSub` since it has both read/write functions.

## 1.3.0

- [feature] New `EctoInterface.Read.PubSub` module for supercharging your contexts with pubsub capabilities. Requires `default_pubsub` as a configuration.
- [feature] New `stream_*/0` and `stream_*/1` functions that return an `Stream` from `Ecto`. Must be executed inside a transaction like all ecto streaming.
- [feature] Added the ability to define which changeset is used in `new_*/*`, `change_*/*`, `update_*/*`, and `create_*/*`.
- Increased test robustness.

## 1.2.0

- [feature] `get_*/1` take a `id` (a primary key) _or_ a `subquery`, but that forces the query version to do the `limit`, `where`, and `one`, now we have two: `get_*/1` that takes an `id` _or_ a query and `get_*/2` which takes an `id` and `subquery` where we do the expected work.

## 1.1.1

- [bugfix] Accidentally was passing the changeset to the preload function, added some tests as well

## 1.1.0

- [feature] Now when create/updating a resource you can pass in a `preload: []` keyword like for `Repo.preload/1`

## 1.0.0

- First release