refs https://github.com/TryGhost/Toolbox/issues/241
- The `engines.ghost-api` property has been deprecated and the support for it will be dropped in Ghost v5 due to versionless nature of the Content API.
- When uploading a new theme or activating existing one that uses ghost-api in it's config a warning will be shown to the user
refs: https://github.com/TryGhost/Toolbox/issues/245
- we don't need this serializer because the default serializer will call the tags mapper
- for now I've changed, rather than removed the tag serializer test as this shows default works the same!
refs: https://github.com/TryGhost/Toolbox/issues/245
- There's no need for a mapper or serializer as labels uses the default behaviour
- Added a full suite of tests, consolidating from regression and using the new framework to prove nothing is broken
- settings cache was appearing as untested because we use rewire!
- rewire was totally unnecessary in this case, so I removed it
- updated to 100% test coverage whilst there, including removing one unreachable branch
- commented some undesirable behaviour I found whilst trying to reach all branches
- I don't want to change the behaviour to return false correctly without having a reason beyond improving coverage
refs: https://github.com/TryGhost/Toolbox/issues/245
- Added a serializer called default to the canary API
- Ideally, this would be part of the shared framework, but this would change v2/v3 and we're about to get rid of them
- Therefore, we change just canary for now, and we can refactor again later.
- Added wiring to handler that uses the default serializer, if there is a default, and isn't an explicit serializer for the endpoint
- Removed the invites serializer, so that one endpoint now uses the default
Note: previous commits have added explicit serializers to every endpoint, this is the first step towards paring
that back so that we have less serializers overall, not more!
refs: https://github.com/TryGhost/Toolbox/issues/245
- Upload, updateMembersEmail, validateMembersEmailUpdate & disconnectStripeConnectIntegration were all missing serializers
- Upload is an as-is response, same as download
- updateMembersEmail, validateMembersEmailUpdate & disconnectStripeConnectIntegration are all passthroughs with no response
- With the upcoming refactor we want to have all the serializers defined explicitly
- This will allow us to change the default behaviour
- Updated the file based tests to prove the body doesn't change
- Tests were added to cover updateMembersEmail, validateMembersEmailUpdate & disconnectStripeConnectIntegration in 68c1bc0285
refs: https://github.com/TryGhost/Toolbox/issues/245
- sendTestEmail was missing a serializer because it's deliberately a passthrough
- With the upcoming refactor we want to have all the serializers defined explicitly
- This will allow us to change the default behaviour
- Updated the tests to prove the body doesn't change
refs: https://github.com/TryGhost/Toolbox/issues/245
- The destroy endpoint was missing a serializer
- Instead of adding one, I've refactored to use an all method that's a passthrough
- Updated the tests to use the same pattern as others to make it clearer this is tested
refs: https://github.com/TryGhost/Toolbox/issues/245
- The destroy endpoint was missing a serializer
- As this serializer uses the same createSerializer pattern as members, I've copied the passthrough from members into here
- This ensures the behaviour will stay the same when the default behaviour changes
- Updated the tests to prove the body doesn't change
refs: https://github.com/TryGhost/Toolbox/issues/245
- The destroy endpoint was missing a serializer
- Instead of adding one, we've refactored to use the standard structure for this serializer
- Updated the tests to prove the body doesn't change
refs: https://github.com/TryGhost/Toolbox/issues/245
- There was only a serializer in place for redirects.download.
- Upload was falling through, which means nothing happens by default atm
- We want to change this default, so I'm making sure all our routes have serializers declared and tests
- Updated the tests and checked the behaviour was the same before and after:
- We can't use our new framework here yet because it doesn't support uploads or downloads
- Instead, just add simple matching for the body of the responses
refs https://github.com/TryGhost/Toolbox/issues/241
- The `engines.ghost-api` property has been deprecated and the support for it will be dropped in Ghost v5 due to versionless nature of the Content API.
- When uploading a new theme or activating existing one that uses ghost-api in it's config a warning will be shown to the user
- split out the two tests that use files, as the new framework doesn't support this yet
- convert the 3 existing tests for the settings endpoint to use the new framework
- introduced a new pattern of using a function to generate a matching array, so that we can do extra stuff
- in this case, just one of the items needed a matcher for the value, which results in slightly weird code
- wrapping in a function gives us somewhere to do this and leave a comment
- some tests were inconsistently using copy-pasted matchers
- instead, we should define them once and use them everywere
- this keeps the tests easy to read and maintain
refs https://github.com/TryGhost/Team/issues/1141
- switched to the same member fetch method as used in `GET /member/:id/` so there's consistent data available when rendering the API responses
- Attempted to make it clearer what the db utils do and what each one is for
- Clearly marked the old clearData & initData as deprecated, we don't want to mix db and url service resets
- The fast truncateAll method can fail
- If it does, use knex-migrate to do a safe reset of the DB instead
- This is being done to try to make test runs more reliable in the face of database errors
- if the database does not exist, ignore the error
- tables that don't exist don't need truncating
- instead wait until the next thing calls init, that will cause a fresh db to be created in the correct state
- inside of reset, if the truncate fails then do a full reset instead
- some code in tests were still using various unreliable methods to determine the database in use
- DatabaseInfo is the correct method to check if the db is mysql or sqlite
- Exposed DatabaseInfo via our test db-utils to make this easier
refs c4470ff732
- labs flag was removed under the false assumption it was a client-side only flag but the `last_seen_at` property in API responses was also gated meaning the member details screen showed "Not seen yet" and the members list did not show the last seen date of all members when filtering
refs https://github.com/TryGhost/Team/issues/1367
Because we are passing through a different member object as the context
in the get helper, the content gating was not working correctly, as the
member was missing a status property, this adds the property which fixes
content gating.
- Added extra tests for get helper {{access}} property
- Added extra test for {{access}} property in next_post helper
- In the future we might want to update the tests so they test the whole request -> HBS context flow. Currently the has context is still stubbed manually.
- We've been seeing weird errors with tables not existing when running tests locally
- This appears to happen after an error causes the tests to abort
- This change includes two fixes:
1. we triggers a full DB reset just before the entire test suite runs
- this is done by wrapping the override file for tests that use a db, and supplying a mochaGlobalSetup hook
2. catch errors when attempting to do a fast truncation-based reset & init, and do a full reset & init instead
- These two changes should ensure the DB is always in the state we expect when running a new test suite
closes https://github.com/TryGhost/Team/issues/1426
When fetching tiers using the content API, we incorrectly returned all tiers including archived ones unless the active:true filter is passed. Correct behaviour is to always hide archived tiers, so this filter should not be required.
- forces `active:true` filter for tiers content api browse
- updates test to check for archived test removal in tiers content api
- the setup() function returns a function that's expected to be called as a mocha hook
- wrapping this in a function means it doesn't get called properly
- therefore the db setup was never being called for this test
fixes: https://github.com/TryGhost/Team/issues/1424
- This moves the require of the members service after Ghost has been booted normally.
- This gives the tests time to properly setup temporary folders before the paths get used.
- In turn this prevents issues in the legacy-members tests, which were affected by the Member Service being called too early in this test.
refs: https://github.com/TryGhost/Toolbox/issues/245
- Ghost's API framework has a mixed up concept of what a serializer is. Mappers are true serializers! What we call serializers are little more than a small formatting step.
- This PR splits mappers into individual files and uses the endpoint's docname as the mapper name. This will help us to automate the calling of a mapper for an endpoint later.
- This is one tiny step in reworking the framework to need less code to make it work, and to have clearer concepts for how to do things.
As multiple tiers is now GA, we want to allow devs to be able to work with Tiers and offers via content/Admin API. This change -
- updates fixtures to add permissions to admin integration role for new sites
- adds migration to update existing sites to have correct permissions for role
- whitelists add/edit/read/browse on tiers and offers API for integrations
refs https://github.com/TryGhost/Toolbox/issues/169
- There was not much clarity around how the tokens created for the versioned API audience would behave when non-versioned API is introduced. The tests added here illustrate the tokens being forward compabible (created for versioned -> verified at non-versioned API) and not backwards combatible (created for non-versioned -> verified at versioned)
refs https://github.com/TryGhost/Toolbox/issues/169
- Before releasing Ghost v5 we would like to move all canary-related URLs to a non-versioned format, which will become a default in v5.
- 'canary' is by definition unstable, so breaking any unprepared client explicitly using the canary is expected
- Removed the aliased /content/ and /admin/ apps from app.js because with updated configuration they become duplicates of 'canary' endpoints
refs https://github.com/TryGhost/Team/issues/1267
Multiple tiers is coming out of beta -
- allows site owners to create multiple tiers
- allows setting individual welcome page for each tier
- allows setting visibility for individual tiers for portal and themes
refs https://github.com/TryGhost/Team/issues/1363
- When uploading a zip of images in Settings > Labs > [Import], it will skip images that have an uppercase extension, citing an 'unsupported file type' error.
- Cause: Glob ignored those files when matching extensions in ImportManager
- Fix: Added nocase option where needed
- Extended tests to also test the processZip method of ImportManager with getFilesFromZip
- Added isValidZip for zip with uppercase image
- Cleaned up JSDoc in ImportManager, and replaced some older JS syntax
Fixed zipContainsMultipleDataFormats error never thrown:
When a zip combines two data formats, no error was thrown.
- The promise error was only returned in an _.each loop, but never thrown
- Previously when combining multiple data types in a zip file, no error got thrown
- Added a test for this error
- Also added a test for noContentToImport error
Other errors and fixes:
- Added missing length in getBaseDirectory check
- getContentTypes fixed (returned duplicate values). Type error came up after adding all JSDocs
- updated tests to match real types from JSDoc and pass type validations
- Rewrote some methods in the async await syntax
- Added tests for ImportManager clean up
- Our old fixtures were designed as a guide to getting started to Ghost, but they got in the way
- The old fixtures now live as part of ghost.org/resources - a living guide to starting with Ghost
- These new fixtures mean the site is ready to go as soon as it's setup
Co-authored-by: Hannah Wolfe <github.erisds@gmail.com>
- we send the roles data array in when we're changing the role of the
user
- if we send an empty array, we don't want to edit the user's role
- the code _thought_ that's what it was doing, but we only check the
falsiness of the array, which is truthy for `[]`
- it also needs to check the length of the array
- this commit includes a test which would fail with a 500 error without
the fix
refs https://github.com/TryGhost/Team/issues/1284
When you create a new post with a tag slug that contains spaces, those spaces will get replaced by dashes. But instead of reusing an existing tag, a new tag is always created.
- New tag slugs are cleaned up before matching with existing tags in the Post model onSaving method
- Cleaned up multiple loops in onSaving of Post model
- Cleaned up syntax when cleaning up tag slug
- Added tests for slugs with spaces
- Added test for too long tag slug causing duplication
Some tests need to run without Stripe connected/enabled, which is the
default at the moment. This method will disconnect stripe for the
lifecycle of the API agent.
refs https://github.com/TryGhost/Team/issues/1188
The products API did return a 500 error when you tried to store product benefits with an empty name. This should throw a 422 Validation Error instead. This change includes some tests for this error. The error itself has been solved by updating the bookshelf-relations dependency earlier.
- Added test when creating a new product with an empty benefit name
- Added test when updating an existing product with new benefits, with an empty name
- Added a test that creates a tier with benefits
no issue
- The support for the misformated roon imports was temporary anyway, based on the comments in code. It's also unecessary to keep around any code related to Roon as it's been _ages_ since anybody needed this kind of migration
