ref ENG-1490
- Regression from the secondary lexical instance.
- Replaced the visibility and positioning styles with width, height, and
overflow to prevent layout space issues.
- Ensured the element takes up no space and is not visible, while
allowing proper initialisation.
fixes https://linear.app/tryghost/issue/ENG-1484
- in Ghost release
[v5.89.0](https://github.com/TryGhost/Ghost/releases/tag/v5.89.0), we
have added a safeguard around bulk member deletion, due to a limitation
in NQL for member filters (commit: 2484a77)
- with this change, we limit the safeguard to only the cases we know are
problematic, and remove it for other useful and safe queries
- more precisely, the safeguard is in place only when:
- Multiple newsletters exist, and the filter contains 2 or more
newsletter filters
- If any of the following stripe filters are used even once:
- Billing period
- Stripe subscription status
- Paid start date
- Next billing date
- Subscription started on post/page
- Offers
no issue
- added passthrough of `feature.contentVisibility` when rendering Lexical to HTML to allow for labs-feature specific rendering of HTML cards
- updated golden post email render test to account for labs flags being on/off
ref PLG-109
- Added a boolean config property `stripeEnabled` to the configuration.
- This property is now passed down to the Koenig Editor to manage
features based on Stripe connectivity.
closes https://linear.app/tryghost/issue/PLG-15
- removed `internalLinking` GA labs flag
- renamed search providers to `flex` and `basic`
- keeps old search provider around as it can handle non-English languages unlike the faster flex provider
- updated `search` service to switch from `flex` to `basic` when the site's locale is not english
- bumped Koenig packages to switch from a feature flag for toggling internal linking features to the presence of the `searchLinks` function in card config
- updated tests to correctly switch between flex and basic providers in respective suites
fixes
https://linear.app/tryghost/issue/DES-435/excerpt-in-post-settings-has-an-inconsistent-error-state
The excerpt form field seemed to not be properly handling errors.
However, it was a case of the error styling being overruled by the
regular styling, causing the red border to only show upon `:focus` when
there is an error in the excerpt.
I've rewritten the logic to be slightly less obfuscated and added some
CSS to circumvent the issue.
ref ONC-225
- Wires up the `editor_default_email_recipients` key to the settings
public / content api endpoint.
- This key is then wired up to Portal to determine whether it's hiding or
showing the Member subscribe toggle
refs https://ghost-foundation.sentry.io/issues/4907452370/
- we want to ignore these errors but the caret is stopping us from doing
so because the errors usually start with AbortError
- we can remove the caret to do so and clean up Sentry
refs ENG-661
Fixes a long-standing issue where an outdated Lexical schema in the
database triggered the unsaved changes confirmation dialog incorrectly.
Implemented a secondary hidden Lexical instance that loads the state
from the database, renders it, and uses this updated state to compare
with the live editor's scratch.
This ensures the unsaved changes prompt appears only when there are real
changes from the user.
ref
https://linear.app/tryghost/issue/ENG-1440/backfill-offer-redemption-data-with-a-migration
There was a bug that caused offer redemptions to not be recorded in the
database for some subscriptions that were created with an offer.
However, we still have the `offer_id` attached to the subscriptions, so
we are able to backfill the missing redemptions. The bug was fixed in
bf895e6e99
This commit only contains a migration, which queries for subscriptions
that have an `offer_id` but do not have any offer redemptions recorded,
and adds any missing redemptions to the `offer_redemptions` table.
ref https://linear.app/tryghost/issue/ONC-197
- YouTube has started responding to video page requests with localised content when requested from certain IPs, with that localised content not containing the required `<link rel="alternate" ...>` tag pointing to the oembed endpoint
- we were fetching video pages rather than the oembed endpoint for YouTube Live URLs because they are not recognised by the oembed extraction library we use
- by modifying the URL from a live URL to a watch URL before we perform oembed lookup/extraction we are able to bypass the (localised) page fetch and instead grab the oembed content directly
ref https://linear.app/tryghost/issue/ONC-199
The `updateSubscriptionItemPrice()` method in our Stripe library used by the importer when moving a subscription over to a Ghost product/price was setting `proration_behavior: 'always_invoice'`. This resulted in invoices being created when changing the subscription (even though no prices were changing as far as the customer is concerned) and in some cases where a customer previously had a one-off discount the customer was incorrectly charged the proration difference because the discount was no longer applied to the new invoice.
- updated `updateSubscriptionItemPrice()` to accept an `options` param allowing the `proration_behavior` property passed to the Stripe API to be overridden on a per-call basis
- updated the `forceStripeSubscriptionToProduct()` method used by the importer to pass an options object with `prorationBehavior: 'none'` when updating the subscription item price so that no invoice and no unexpected charges occur when importing
ref https://linear.app/tryghost/issue/ONC-199
The `updateSubscriptionItemPrice()` method in our Stripe library used by the importer when moving a subscription over to a Ghost product/price was setting `proration_behavior: 'always_invoice'`. This resulted in invoices being created when changing the subscription (even though no prices were changing as far as the customer is concerned) and in some cases where a customer previously had a one-off discount the customer was incorrectly charged the proration difference because the discount was no longer applied to the new invoice.
- updated `updateSubscriptionItemPrice()` to accept an `options` param allowing the `proration_behavior` property passed to the Stripe API to be overridden on a per-call basis
- updated the `forceStripeSubscriptionToProduct()` method used by the importer to pass an options object with `prorationBehavior: 'none'` when updating the subscription item price so that no invoice and no unexpected charges occur when importing
Ref https://linear.app/tryghost/issue/SLO-193/optimise-count-query-skip-distinct-from-count-query-for-members-events
The member events endpoint have many queries like:-
select count(distinct members_subscribe_events.id) as aggregate
from `members_subscribe_events`
where `members_subscribe_events`.`created_at` < '2024-07-30 11:30:39'
In these queries, distinct is not required as id is a primary key. Skipping distinct would improve the performance.
This PR will changed the query to:-
select count(*) as aggregate
from `members_subscribe_events`
where `members_subscribe_events`.`created_at` < '2024-07-30 11:30:39'
ref 8ea1dfb
ref https://linear.app/tryghost/issue/ONC-111
* undid the reversion for the performance improvements
* built upon new tests for the posts list functionality in admin,
including right click actions
* added tests for pages view in Admin
This was reverted because it broke the Pages list view in Admin, which
is a thin extension of the Posts functionality in admin (route &
controller). That has been fixed and tests added.
This was originally reverted because the changes to improve loading
response times broke right click (bulk) actions in the posts list. This
was not caught because it turned out we had near-zero test coverage of
that part of the codebase. Test coverage has been expanded for the posts
list, and while not comprehensive, is a much better place for us to be
in.
fixes https://linear.app/tryghost/issue/ONC-206
ref https://app.incident.io/ghost/incidents/90
- when multiple member filters are used in combination, NQL sometimes
hit a limitation that results in the wrong members being returned
- while we work on the NQL limitation, we are temporarily disabling bulk
member deletion when more than one member filter has been applied