* test subtotals on draft order operations
* fetch cart with subtotals when doing draft order operations
* write test for updating discounts seeing changes in totals
* formatting
* udpate test
* force taxes
* missing select
* add import
* rm force_taxes
**What**
- enable to filter sales channels of a Publishable API Key with a free text search param
Co-authored-by: Oliver Windall Juhl <59018053+olivermrbl@users.noreply.github.com>
### What
Explicitly declare `@medusajs/medusa` as a devDependency for `@medusajs/medusa-js` client package.
### Why
With our latest version upgrade of `turbo`, peerDependencies aren't taken into account in the build graph. The recommendation is to also explicitly declare them in devDependencies.
### Test
* Run `yarn build --graph`
* Expect to see medusa-js having a dependency on medusa
```
"[root] @medusajs/medusa-js#build" -> "[root] @medusajs/medusa#build"
```
### What
Flip Turborepo cache glob pattern from an allow-list to deny-list pattern.
### Why
Packages within medusa's monorepo will output their build to a `dist` directory.
This convention does not apply to plugins since the core plugin loader expects the content of plugin packages `src` directory to be outputted at the root of the package.
i.e. `packages/foobar-plugin/src/utils/index.ts` -> `packages/foobar-plugin/utils/index.js`
Manually maintaining an allow-list of known plugin output directories is not scalable. When a directory exists in a plugin package but is not know to the allow-list, the directory will not be re-built on subsequent builds. Troubleshooting the issue requires intimate knowledge of Turborepo caching strategies.
### How
By using a deny-list glob pattern, plugin packages can now declare any not-known directory within their `src` folder without facing any potential omission issues during the build process.
We declare `src` and its content as the only directory for turbo cache to ignore.
### Additional scope
* Use `turbo.json` file content in cache hashing algorithm in order to break CI cache when changes are made to the Turborepo config.
* Upgraded turbo minor verion.
* Added missing dependency to medusa package.
### Test
* Delete previously built output in packages. Run `yarn build --force` (replace any existing cache)
* Expect all src content to be outputted
* Run `yarn build` right after
* Expect a fast build time since cache will be fresh
* Add a new directory with an index.ts file in a plugin package src folder. Run `yarn build`
* Expect a fast build time, except for the modified plugin package.
* Expect the newly added directory to be outputted.
* Delete the newly outputted directory. Run `yarn build`
* Expect the outputted directory to reappear.
### References
* 5093b82f3a/packages/medusa/src/loaders/plugins.ts
* https://turbo.build/repo/docs/reference/configuration#outputs
* https://turbo.build/repo/docs/reference/configuration#globaldependencies
Resolves CORE-891
**What**
When a discount is deleted from a cart, the line item adjustments must be refreshed. Now, the cart service `removeDiscount` includes updating the adjustments.
**Tests**
**Add a new integration test which:**
- create a cart with a discount
- add a line item
- validate the adjustments
- remove the discount
- check that the adjustments are not present anymore
**Update unit tests**
The actual tests cases now check that the adjustments repository is called when needed
FIXES CORE-890
A discount should always be associated to atleast 1 region - we validate it against the region in the cart service.
Adding a request parameter validation to pass in atleast one regionID in an array of regions will ensure that regionID will be validated as a part of the discount service and fail if its not a valid region ID.
**What**
I have created a new method on the cart service which is `addLineItems`, allowing a user to add one or multiple items in an optimized way. Also updated the `generate` method from the line item service which now also accept a object data or a collection of data which. Various places have been optimized and cache support has been added to the price selection strategy.
The overall optimization allows to reach another 9000% improvement in the response time as a median (Creating a cart with 6 items):
| | Min (ms) | Median (ms) | Max (ms) | Median Improvement (%)
|---|:-:|---|---|---|
| Before optimisation | 1200 | 9999 | 12698 | N/A
| After optimisation | 63 | 252 | 500 | 39x
| After re optimisation | 56 | 82 | 399 | 121x
| After including addressed feedback | 65 | 202 | 495 | 49x
FIXES CORE-722
**What**
- add/ remove sales channels to/from PublishableApiKey in batch
- associate created cart with SC defined by PK
- filter products if PK with SC association is present
- retrieve a list of sales channels for a PK
- implement 3 new middleware
- `extendRequestParams`
- _extend req object with PK scopes (a list of sales channels) if a publishable key is present in the header of the request_
- `validateProductSalesChannelAssociation`
- _validate if the passed product id belongs to a SC from the PK's scope_
- `validateSalesChannelParam`
- _validate that passed SC ids in the req body/query are within the scope of the PK_
**How**
- The general idea was to reuse existing logic in the controller layer which expects `sales_channel_id` array to be passed. The middleware sets associated SC ids in the request context if a PK is present. These ids are then merged in the controller and passed to the service layer.
**TODO**
- filter response from the search endpoint (CORE-824)
**Testing**
- _integration tests_
- add sales channels to the publishable API key scope
- remove sales channels from the publishable API key scope
- returns products from a specific channel associated with a publishable key
- returns products from multiples sales channels associated with a publishable key
- returns all products if PK is not passed
- returns all products if passed PK doesn't have associated channels
- should assign sales channel to order on cart completion if PK is present in the header
- list sales channels from the publishable api key
- throws because sales channel in query/body is not in the scope of passed PK
---
**Discussion**
- what about the other endpoints (e.g. GET /store/product/:id - do we return 404 if the product is not in the SC associated with passed PK)
- what about products search route
- what about `/admin/products` & `/admin/orders` routes (do we add the middleware there as well)
---
RESOLVES CORE-792
RESOLVES CORE-793
RESOLVES CORE-816
### What
OAS: Explicitly declare type:object on schemas with properties.
### Why
While not officially required, schemas with properties should have their type explicitly declared as "object". Omitting the type translates to type: any with properties XYZ. The ambiguity can lead to issues when using code generators.
### How
Multiple regex searches to identify schemas missing type: object. Manually edit each schema. Spot check generated OAS.
### Testing
- Ran OAS validator.
- Ran docs build script
### Proof

RESOLVES CORE-847
**What**
Allow DraftOrders to be created with an IdempotencyKey.
Note this doesn't implement idempotency for the DraftOrder create endpoint but allows the service layer to ingest the key and store it in the database. This is a preliminary step to being able to support an idempotent API request.
* use correct authentication middleware
* remove guard from get-session since it's guarded by middleware doing the same check
* Add integration tests
* Create lazy-swans-agree.md
Co-authored-by: olivermrbl <oliver@mrbltech.com>
Co-authored-by: Oliver Windall Juhl <59018053+olivermrbl@users.noreply.github.com>