refs https://github.com/TryGhost/Team/issues/1545
**Changes (`members-api`)**
- Compare via https://github.com/TryGhost/Members/compare/%40tryghost/members-api%406.0.0...%40tryghost/members-api%406.1.0
- Added mapping from member subscribed to newsletters on edit/create
- When editing or creating a member with the subscribed property, it is mapped to the corresponding newletters value
- Defaults to all active newsletters with visibility = members and subscribe_on_signup = true
**Tests**
- Adds test that adds a member with subscribed = true
- Adds test that adds a member with subscribed = false
- Adds test that edits a member with subscribed = true
- Adds test that edits a member with subscribed = false
refs https://github.com/TryGhost/Team/issues/1561
With multiple newsletters, unsubscribe links will also need to have a unique reference to the newsletter that the email is for, so that we can unsubscribe members from the particular newsletter automatically when they click on the link.
As our existing pattern for members is to use UUID as the external unique reference, this change adds UUID to newsletter schema and populates the existing newsletters with a UUID value.
- adds new `uuid` column to newsletter schema
- updates newsletter model to add default uuid
- updates default newsletter migration to add `uuid`
- drops nullable on `uuid` column later in migrations once we have populated existing newsletters
- these packages contain only code changes or dependency updates but
we're force to publish new versions due to Lerna's limitations
- this commit bulk updates the packages to save having many commits
refs 58ace0af76
- please see the referenced commit above for full context but this
commit bumps `@tryghost/nodemailer`, which contains a fix for reading
credentials when using the SES mail transport
- credits to @touzoku
closes: https://github.com/TryGhost/Team/issues/1553
- we want to be explicit in what values are supported
- we want the values that are supported to be supported to also be explicit and clear
- without this API users can set the value to anything and the active/not active logic will work, until such time as we introduce further statuses
- this means introducing a new status could be a potential breaking change and lands us in horrible hot water
refs: https://github.com/TryGhost/Team/issues/1557
- There is currently no default order for the members API
- This is done at the API level purely for the endpoint, not in the model using orderDefaultOptions
- This is because orderDefaultOptions affects findPage and findPage is wrapped by membersAPI.members.list
- That in turn is used in multiple places, e.g. getting member counts for emails and getting members for exports
- There is currently no created_at DESC index on the table, so we don't to impact performance too much
- This ensures it's only affected when paginating
refs https://github.com/TryGhost/Team/issues/1478
- Moved all admin API members tests to enable the multiple newsletters flag
- Checks if the susbcribe events are added correctly when adding or removing newsletters
- Checks if susbcribe events are added for default newsletters
refs https://github.com/TryGhost/Team/issues/1550
- Switched to using the newsletter design settings over the global settings
- Made the `newsletter_id` property available in the Admin API Post resource
- Added the `showHeaderName` variable that can be used in the post html template
refs https://github.com/TryGhost/Team/issues/1551
- Updated existing migration to insert a nullable created_at column
- Added a migration to update all the created_at values to now
- Added a migration to drop nullable
- Also includes new helper methods to set and drop nullable
refs https://github.com/TryGhost/Team/issues/1545
- Remapped `member.subscribed` value based on newsletter subscriptions in API output
- Enabled filtering by subscribed status for multiple newsletters
Co-authored-by: Matt Hanley <git@matthanley.co.uk>
refs https://github.com/TryGhost/Team/issues/1552
- The API doesn't enforce a unique sort order but we infer the "default newsletter" based on ordering so we need to ensure a consistent and deterministic ordering behaviour
- Many of these tests were using API calls to get IDs or check side effects
- This makes snap files much harder to read, where keeping tests more minimal gives us the same test coverage
- Also updated the tests to include members, so we have some real live counts
refs https://github.com/TryGhost/Team/issues/1524
- We need to fetch the post newsletter to grab the slug as it's needed for the member NQL filter.
- We can then use the newsletter slug and append it in the existing member NQL filter.
- Removed `subscribed:true` when an email is sent to a newsletter and replaced it with the newsletter id
- Added `status:-free` when an email is sent to a newsletter with `visibility` set to `paid`
- Added tests what happens when you publish without newsletter_id
- Added tests what happens when you publish with newsletter_id
Co-authored-by: Simon Backx <simon@ghost.org>
- Updated the member fixtures to have 6 subscribed and 2 unsubscribed members
- Added an inactive newsletter
- Updated newsletter-member relations to reflect the 6 subscribed and 2 unsubscribed, but with different cases:
- 3 subscribed to default
- 1 subscribed to secondary only
- 1 subscribed to default + secondary
- 1 subscribed to secondary + inactive
- 1 subscribed to inactive only (i.e. not subscribed)
- 1 not subscribed at all
- With these changes, I needed to update the members snap as 2 members appear as subscribed:false in many tests
- I also needed to update some posts regression tests, as 2 less members get emailed in 2 tests
refs https://github.com/TryGhost/Team/issues/1524
- This enables admins in the ghost admin to have an overview of the total posts/members associated with a newsletter.
- Follows the `?include=count.x` convention used by other resources
Closes https://github.com/TryGhost/Team/issues/1508
Refs https://github.com/TryGhost/Ghost/pull/14468
- Maps existing subscribers to the default newsletter
A note on performance:
We loop over the rows in a potentially large table (members) but I've minimised the impact by limiting the columns we fetch. The alternative is a raw SQL query like the one below: the SQL version takes ~0.9s vs ~1.1s for the migration (my laptop, ~30k members). The disadvantage of the raw SQL implementation is the approximation of the ObjectID (instead of a legit bson ID) which isn't sequential and may impact index size/performance.
```sql
insert into members_newsletters (id, member_id, newsletter_id)
select
substr(replace(uuid(),'-',''),1,24),
id,
'62595dcbfad4e031d85d166f'
from members
where members.subscribed=true;
```
refs https://github.com/TryGhost/Team/issues/1495
- removes subscription from all newsletters for a member on click of unsubscribe link in email
- allows the new multiple newsletter system to work with existing unsubscribe flow
refs https://github.com/TryGhost/Team/issues/1478
The `newsletter_id` is nullable for now to remain compatible until we have a proper data migration + updated code to set it on inserts
refs https://github.com/TryGhost/Team/issues/1473
- Added the default newsletter
- We use the title to populate the newsletter same, slug and sender name
- We use the description to populate the newsletter description
- We use the global design settings to populate the corresponding newsletter design settings
Co-authored-by: Matt Hanley <git@matthanley.co.uk>
refs https://github.com/TryGhost/Team/issues/1532
- Added before the migration in https://github.com/TryGhost/Ghost/pull/14468 to populate the default newsletter
- The fixture for the default newsletter has a different value than the model and schema default
- This is because by default the newsletter name is the same as the site title, and the site title is already shown
- I had to fork this dependency to add support for Knex 2.0 but yarn is
weird and won't bump it if we use `master` because "nothing changed"
- Using a hash should force yarn to pull the changes if the hash gets
updated
refs fccee0614f
- `@tryghost/nql-lang` had `date-fns` declared as a devDependency but it
was used within the library itself
- the referenced commit moved it to `dependencies` and this commit bumps
the packages in Ghost
- we had to switch to `@vscode/sqlite3` a while back because `sqlite3`
was unmaintained
- this fork didn't come with prebuilt binaries, so everyone had to
compile them on their machine
- this brought a lot of issues with installing Ghost
- since then, the Ghost team have picked up maintenance of `sqlite3` and
Knex has switched back, so we can switch back here too
