https://github.com/TryGhost/Team/issues/599
- When custom integration limit is enabled all requests from existing integrations should not be accepted. With the exception of internal integrations like backup and scheduler
refs https://github.com/TryGhost/Team/issues/599
- This is a precursor change to tests which verify the hostSettings limits are working correctly
- Bumped limits-service version which allows for multiple calls of loadLimits on the same service instance
no issue
- There is a valid subset of statuses that can be set for the users but there's no "isIn" validation for possible values
- Additionally some of the statuses like warn-1, warn-2, etc. don't have a clear usecase (or at least nothing was found in the codebase for them to be used). They might be up for removal if this assumption is correct
refs https://github.com/TryGhost/Team/issues/586
- Member model now has `products` relation, sorted using `sort_order`, following convention from `labels`
- Product model has handling to set `slug` from name, following convention of Label model
- Updated filter plugin to handle filtering Member models by their `product` relations e.g. `product:[slug, slug]`
refs https://github.com/TryGhost/Team/issues/586
- Add the `stripe_products` table, so that we can map Stripe Products to Products in Ghost
- Add the `stripe_prices` table, so that we can associate Stripe Prices to Products table
refs https://github.com/TryGhost/Team/issues/586
- Add the products table, so that we can store Products in Ghost
- Add the members_products table, so that we can associate Members w/ Products
- Use sort_order on the members_products table to follow the same convention in members_labels
- Populate the products table with a single product, using the name from the stripe_product_name setting
- Populate the members_products table with relations based on the status column of the members table
Populating the tables allows us to transition from the current system, which does not care about products, into the
new system, where Products are used to group members. The intention is that all existing paid members have the
same product
refs https://github.com/TryGhost/Team/issues/585
- adds `DELETE /members/` route to the Admin API
- supports `?filter`, and `?search` query params to limit the members that are deleted
- `?all=true` is required if no other filter or query is provided
- uses `models.Member.bulkDestroy` which _will not_ cancel any Stripe subscriptions if members have them but _will_ clean up the Stripe relationship data in Ghost's database
refs https://github.com/TryGhost/Team/issues/510
- This version bump includes follwing interface improvements of the limit-service package: passing in errors as a parameter to "loadLimits" and allowing for custom "currentCountQuery" method implementations per limit
closes https://github.com/TryGhost/Team/issues/595
Due to a bug in `mrr_delta` calculation, we ended up reducing the MRR delta by twice the original amount when a subscription goes from active to canceled and storing it in `members_paid_subscription_events` table, which is used to show the MRR chart on Dashboard. The way we identify the incorrect events in the table which got the double negative value is by checking if they match certain criteria - Both `from_plan` and `to_plan` have same value as a subscription changes status while being on same plan.
This migration halves the `mrr_delta` for incorrect events to restore the correct MRR change for the site.
no refs
- Removes old `/members/stats` endpoint in favor of new `/members/stats/count` in canary/v4 which captures members counts using new events table
- Removes tests for old `/members/stats` endpoint
- Added test for new `/members/stats/count` endpoint
refs https://github.com/TryGhost/Team/issues/555
- Export files included a lot of data which was not used in the importer, for example: members, labels, migrations and many more. This lead to a lot of clutter in the import files and made it hard to reason about their purpose.
- The main purpose of exports - is to export importable resources. These are posts, tags, and users. The rest of data like members or migrations either have their own importer (like CSV importer for members) or does not and should not have any ways to be imported.
- These changes are in now way complete. It's a first step towards resource-based exports which could be properly versioned in the future on API level and not be a mirror of the DB structure.
- This is sort of a breaking change. But we are doing it because: (1) its an internal API that should not be used by external clients, (2) there was no public contract to have this API stable at any point, (3) we really need to get back the control over export files structure and size
- In case an external client was dependent on some structure of the exported json file they can still pass in ALL of previously exported data by passing table names in `include` query parameter.
refs https://github.com/TryGhost/Team/issues/555
- Previous blocklist approach was resulting in adding every single new table into an export automatically. Which creates possibility to leak sensitive data if not used porperly. Allowlist approach gives better control over what is exported, makes this information explicit, and version-control friendlier
refs https://github.com/TryGhost/Team/issues/555
- The getVersionAndTables was doing too much and was only used once creating clutter in doExport method. Refactored code doing 2 direct calls instead of destructuring "dances".
refs https://github.com/TryGhost/Team/issues/397
The default newsletter/support email address for a site is currently setup as noreply@DOMAIN , which means for a custom domain setup with www the email address becomes noreply@www.somesite.com which is not the expected behavior normally. This removes the `www` subdomain if present for those email addresses, but doesn't change any other subdomain
refs https://github.com/TryGhost/Team/issues/536
From 4.0, we ensure and require that accent colour is always set. This change removes hardcoded accent color fallbacks to avoid confusion as well as cause accidental fallback that is undesired causing themes to look different
closes https://github.com/TryGhost/Team/issues/571
- `post.plaintext` values were being transformed with the markdown URL transformer but that wasn't picking up the link format used in our plaintext fields resulting in absolute URLs being stored in the database rather than `__GHOST_URL__` URLs
- meant that if the `url` config is changed then plaintext and other calculated fields that used it would have URLs that referred to the old domain rather than the new one
- re-saving the posts would have updated it but that's not feasible to do manually for large sites
- bumped `@tryghost/url-utils` to a version that has plaintext transform utils and updated the post model's transform map
refs https://github.com/TryGhost/Team/issues/510
- contributors don't count towards the staff limit, therefore they should be allowed to be unsuspended
- currently, we don't check the role when unsuspending, which is incorrect
- this bug is pure oversight!
closes https://github.com/TryGhost/Ghost/issues/12791
closes https://github.com/TryGhost/Team/issues/566https://github.com/TryGhost/Ghost/pull/12787 introduced a significant performance regression due to a misunderstanding of when Bookshelf calls `.format()` ([related upstream issue](https://github.com/bookshelf/bookshelf/issues/668)). We expected `.format()` to only be called on save but it's also called when Bookshelf performs fetching and eager loading which happens frequently. `.format()` can be a heavy method as it needs to parse and serialize html and markdown so it should be performed as infrequently as possible.
- override `sync()` in the base model so we can call our own `.formatOnWrite()` method to transform attributes on `update` and `insert` operations
- this was the only feasible location in Bookshelf I could find that is low enough level to not require modifying model instance attributes
- gives models the option to perform heavy transform operations only when writing to the database compared to the usual `.format()` method that is also called on fetch in many situations
closes https://github.com/TryGhost/Team/issues/570
When site owner/admin updates their newsletter/support email address from settings, they receive an email with confirmation link which on success takes them to Ghost Admin on email settings screen with a toast about success. Since the path for email settings in Ghost Admin changed in v4, the fix updates the redirect link to new Admin settings URL.
refs 4dc413d6a1
- Fixed failing test cases which were designed to check non-major upgrade messages. There's no clear use cases for those, but still worth keeping such case in mind
closes https://github.com/TryGhost/Team/issues/564
refs https://github.com/TryGhost/Ghost/issues/10236
- The notification to upgrade to new 4.0 Ghost version was still visible to users after upgrading the instance to 4.0. This was caused by notification filtering not taking into account 3.x or 4.x versions.
- The fix filters out notifications that detect a major version notification using `x.0 is now available` pattern and compares current version to that major. This should future proof the issue from happening in Ghost 5.0 (but a proper holistic fix is preferable!)
closes https://github.com/TryGhost/Team/issues/552
Refactors URL transforms so they take place at the model layer rather than the API serializer layer. Continuation of the pattern created for the settings model in https://github.com/TryGhost/Ghost/pull/12738
- Added checks to all front-end tests to ensure output does not contain the magic replacement string
- includes failing acceptance test for `__GHOST_URL__` appearing in sitemaps
- Removed all transform-ready URL transforms from API serializers
- input serializers transform image urls relative->absolute to keep absolute-urls as the consistent "outside of the database" format
- output serializers should not need to perform any URL transforms as that will be done at the model layer
- Added url transforms to models layer
- removes knowledge from the API serializers which shouldn't need to know how data is stored internally in the database
- makes absolute urls the consistent "outside of the database" URL format
- adds transform step to the sitemap generator because the data used for that is fetched directly via knex which will not run through the bookshelf `parse()` methods
closes https://github.com/TryGhost/Team/issues/557
After updating to v4 on SQLite from v3, we were unable to delete members due to a foreign key mismatch error. This is because the migrations which recreates the tables for `members_stripe_customers_subscriptions` and `members_stripe_customers` doesn't add back the unique constraint which is needed by FKs for reference.
The migration creates the missing UNIQUE constraints on the tables.
refs 6b07d4b2a0
- The model is needed here, because it contains full set of fields. In some cases, like email-preview, the "plaintext" field is not present in "attrs" which causes the logic to fail.
- This should be sorted along with https://github.com/TryGhost/Ghost/issues/10396
refs https://github.com/TryGhost/Team/issues/467
refs a6f5eb71be
- When a generated excerpt is calculated for posts/page resources it uses raw model! to get the data. Model contains untranformed __GHOST_URL__ markup which has to be additionally processed before extracint an excerpt or use the transformed `plaintext` from available attributes (chose the latter to decrease complexity)
- Removed model dependency as `attrs` at this point of serialization should always contain the `plaintext` field. It's ugly and has an unsolved bug report here - https://github.com/TryGhost/Ghost/issues/10396. The reliance should be solved at some point, but definitely not a part of this issue
refs: https://github.com/TryGhost/Team/issues/510
refs: e30b9735fa
- In the case that there is a staff user limit, and the limit is maxed out, it is no longer possible to invite new staff users
- However, Contributors are not considered staff users and therefore it should always be possible to invite new Contributors
no issue
Comped members were not able to view paid-member content because content gating was only looking for `member.status === 'paid'` which doesn't take into consideration members on a "complimentary" plan.
- added front-end acceptance tests for member access to posts
- updated content-gating check to take comped members into consideration
- With 4.0 we have a brand new version of Casper, new fixtures and new default settings
- Fixture posts cover the key features and give users an introduction to how to use their site
- This all comes from the marketing and design teams to refresh the look and feel of Ghost and give users the best possible onboarding experience
Note: this fixture overhaul includes
- new content for new 4.0 features
- regenerated post content using our updated mobiledoc structure
- a switch from British to US English
refs 2bba9989db
- Note: this will require new fixtures so that the navigation links actually work
- These updates are all in aid of getting the best possible default setup and onboarding experinence for new Ghost users
- With 4.0 we have a brand new version of Casper, new fixtures and new default settings
- This all comes from the marketing and design teams to refresh the look and feel of Ghost
Note on accent color:
This commit changes the default accent colour again.
The intention is that new sites should get #FF1A75 (pink) as their default.
Any existing sites that do not have an accent colour set yet, should get #15171A (black) on upgrading to 4.0.
These are different as they are different experinces. Fresh sites will be guided to pick a color, so
a bright color is more visible and helps to see what can be done, whilst existing sites get a muted
black, that should be a sensible fall back color.
refs TryGhost/Team#535
We want to ensure that a site will always have a default value of
`'#15171A'` for the accent_color setting.
Since the boot process changed we have three cases to account for:
1. Setting does not exist
2. Setting exists with no value
3. Setting exists with a value
It is only in the case of 2. that we want the migration to update the
database with a default value.
In the case of 3. the site owner has already set a value, which we do
not want to override.
In the case of 1. the setting will be created (and populated with
default value) from the default-settings.json file, by the
populateDefaults method called from the settings service
We also update the accent_color setting to include a non-empty
validation, to ensure that the setting will always have a value, as
sites before 4.x may have an empty accent_color, we must update the
importer to set the default value if one is not present. Otherwise we
would run into validation errors and even if we didn't would have an
invalid database state.
no-issue
We were originally checking the state of the database, e.g. if a foreign
key constraint existed, so that we could conditionally act upon it. This
was to ensure that our migrations are idempotent.
Some database configurations, for example if you have many databases on
a single MySQL instance, would cause these information_schema queries to
take an exceptionally long time.
In order to speed up migrations, we instead attempt the action we want
to apply to the database, and then catch relevant errors to ensure the
migration is idempotent.
SQLite does not error when adding duplicate foreign or primary key
constraints, meaning that we must keep in pre-checks for these
operations, when running on SQLite
Co-authored-by: Daniel Lockyer <hi@daniellockyer.com>
no issue
Upgrading from v1 or v2 can result in successful upgrades but with mobiledoc errors showing in the logs:
```
NAME: InternalServerError
MESSAGE: Mobiledoc card 'card-markdown' not found.
```
The errors do not signify a problem as long as the 4.0 migrations run because those rename the deprecated card before re-rendering.
- `@tryghost/kg-default-cards` dropped support for `card-markdown` cards. 4.0 migrations handled this by renaming all `card-markdown` cards to `markdown` before re-generating any content
- 2.0 and 3.0 also had migrations that re-generated content but they are run before the 4.0 card rename migration meaning that the mobiledoc renderer sees cards that it doesn't know about. The behaviour for unknown cards is to log an error and skip rendering of that card
- by NOOPing the 2.0 and 3.0 migrations we eliminate the incompatibility errors and reduce the amount of processing the upgrade needs to perform
refs https://github.com/TryGhost/Ghost/pull/12736
refs https://github.com/TryGhost/Team/issues/467
knex's `parse()` method is only called on data when directly fetched from the db. This was causing problems when model instances are passed around via events for example because `.get('key')` will return data that was directly set on the model without having gone through the `parse()` transformations. The result of this inconsistency was settings appearing correct when Ghost started up but then being broken as soon as a setting was changed.
- moved absolute/relative->transform-ready URL transformations from the API input serializers to the model's `format()` method and replaced with a relative->absolute transform in API input serializers
- results in consistency because `.get()` on a settings model will always return an URL
- removed transform-ready->absolute transforms from the API output serializers as that is now handled at the model-layer
refs https://github.com/TryGhost/Team/issues/467
refs https://github.com/TryGhost/Ghost/pull/12731
- settings are mostly fetched directly from the settings cache rather than via the API so they aren't subject to the API-level output serializers that transform URLs meaning that URLs in the front-end ended up with raw `__GHOST_URL__` replacement strings
- added images to the Settings model's `parse()` method so they are transformed immediately when fetching from the database
closes https://github.com/TryGhost/Team/issues/467
- switches to storing "transform-ready" URLs in the database
- transform-ready URLs contain a `__GHOST_URL__` placeholder that corresponds to the configured url that gives a few benefits
- much faster and less memory intensive output transformations through not needing to parse html or markdown - the transform can be achieved using a straightforward regex find+replace
- ability to change to/from or rename subdirectory without any manual updates to the database
- modified existing 4.0 url-transformation migration rather than adding another one and repeating the transformation on posts rows
refs https://github.com/TryGhost/Team/issues/525
We expect the member event tables to be large, and they contain what is
considered metadata. For this reason we do not want to include them in
the export of sites.
refs https://github.com/TryGhost/Team/issues/469
- Cleans up response format for mrr and volume stats endpoint to more consistent pattern
- Removes `unit` attribute for now as its not used
refs: https://github.com/TryGhost/Team/issues/510
- Changed to return the full hostSettings key, not just the billing URL
- We are introducing several more settings that are needed by Admin including limits
- Passing the whole object makes this much easier to reason about as Admin has the exact same config as the server
refs: https://github.com/TryGhost/Team/issues/510
- The current member limit was implemented as a member-specific concept
- The new limit service is much more generic, here we are swapping old for new
- The updated concept here is blocking all publishing, not just email sending, when a site is over its member limit
- To determine that we are publishing a post, we must be in the model layer. The code has been moved to the permissible function which makes sense as this is a permissions error that we are throwing
- I've left the extra check for email retries in, in case there is some loophole here (but we may wish to change it)
refs: https://github.com/TryGhost/Team/issues/510
- In the case that host config is provided, keep staff users within the limiti
- The definition of a staff user is a user with a role other than Contributor, and whose status is not inactive
- Contributors don't count
- Suspended (status inactive) users don't count
- Locked users DO count
- Invited users DO count
- You can't invite more staff users whilst there are pending invites
- You can't unsuspend a user, or change the role on a user in such a way as will take you over your limit
- You can't import staff users - all imported users are automatically set to Contributors
- As part of this work, we are changing the default Ghost user to a Contributor otherwise it uses up a staff user
Note: there is one known active bug with this commit.
- Assume you have one remaining user within your limit. You send an invite, this works.
- You cannot "resend" that invite, it will think you're sending a new invite and hit the limit
- You must "revoke" that invite first, and create a new one
- This bug exists because the resend function uses the add endpoint & does a delete+add, but this hits the permission check before the delete
refs https://github.com/TryGhost/Team/issues/510
- When the host config was introduced it was incorrectly introduced as host_settings instead of hostSettings
- All other Ghost config uses camelCase, so changing this now before it becomes a problem
- Note: Also removed some rogue return awaits that don't make sense. It's not possible to hit that case, but cleaning up anyway
refs https://github.com/TryGhost/Team/issues/509
- Allows to update and read 'locale' key along with the deprecated 'lang'
- In Ghost v5 the 'lang' key will be dropped and the migration in settings table will clean up the key name to match the one exposed through the APIs
refs https://github.com/TryGhost/Team/issues/509
- During 3.x we standardised on "lang" instead of "default_locale" for the site setting, which was an assumption based on an earlier change to @site.lang to make <html lang="@site.lang"> read nicer. This was a mistake as the field represents more than "lang" its a "locale". With this changeset we introduce a transition to use "locale" name for the value instead of "lang"
- Adds `@site.locale" value in as well as new 'locale' property in Content API's response
- "lang" will be considered as deprecated starting with API v4 and will be dropped completely with API v5
refs https://github.com/TryGhost/Team/issues/509
- Flipping around key/newKey pair allows to map multiple keys to the same field in the settings
- This becomes handy when there's a need to deprecate a field. For example, we are about to introduce a 'locale' setting which would need to map to 'lang' db key, with current structure it's impossible to have many:1 mapping because it's impossible to have duplicate keys in the JS object ("hash")
refs https://github.com/TryGhost/Team/issues/513
refs https://github.com/TryGhost/Team/issues/477
- We have skipped work to improve the 'defaultTo' value when working on Ghost 4.0 release, so adding this comment while context loaded
- defaultTo should not be set to anything as it leads to more maintenance work during major version bump
- having validation might make sense but could lead to similar maintenance work unless it's linked to some global nosion of "supported API versions" used everywhere
refs https://github.com/TryGhost/Team/issues/513
- Having exports in no particular order was making it hard to spot if there's anything missing
- Having a DESC order on the exported API versions makes it very easy to modify and maintain in the future
refs: https://github.com/TryGhost/Team/issues/510
- added and wired up the new limit service, which is a lazy-loaded service
- this handles the case that there are host limits set in config, and wraps all the logic needed for detecting exceeded limits & throwing limit errors
- expects limits to be set in config under `host_settings.limits`
- supported limits are managed in the limit service, outside of core
refs https://github.com/TryGhost/Ghost/issues/12610
refs https://github.com/mailgun/mailgun-js-boland/blob/v0.22.0/lib/request.js#L285-L333
The mailgun domain is used by the mailgun API to construct the URL for
the API. e.g for a domain of "mg.example.com" the URL for the API
messages would look like:
https://api.mailgun.net/v3/mg.example.com/messages
One weird thing about the mailgun API is that if the path does not map
to an API endpoint, then instead of a 404, we get a 200, with a body of
"Mailgun Magnificent API".
The `mailgun-js` library which we use, expects a JSON response, and will
return a body of undefined if it does not get one.
This all resulted in us trying to read the property `id` of an undefined
`body` variable. The fix here is to reject the containing Promise, if
there is no body. So that the default error handling will kick in.
refs https://github.com/TryGhost/Team/issues/513
- Bumping version as this is now the latest stable API.
- The change might cause this particular side effect (acceptable for major version): if a member requests a login URL when the instance is on 3.x version and site owner upgrades do 4.x before member authenticates through login link, the login will fail and the member will have to request a new login URL.
refs https://github.com/TryGhost/Ghost/pull/12541
refs https://github.com/TryGhost/Ghost/pull/12689
- the analytics job had been switched to create it's own instance of EmailAnalyticsService to avoid requiring logging but the analytics extraction branch was created before this change and wasn't picked up when merging
- pulled `queries` option object into a separate file for re-use
- updated `fetchLatest` job to conform to extracted library interface
closes https://github.com/TryGhost/Team/issues/493
- all functionality except that directly related to Ghost's database and business logic now lives in external packages
- @tryghost/email-analytics-service
- @tryghost/email-analytics-provider-mailgun
refs https://github.com/TryGhost/Ghost/issues/12716
- This change is made to avoid using now deprecated 'v2' API anywhere in the codebase.
- Switching to 'v4' should not cause sideeffects as this parameter is always present within the URL when Admin API is used
refs https://github.com/TryGhost/Ghost/issues/12716
- The code in serializePostModel was broken and always defaulted to 'v3'! It refered to non-existent `model.get('api_version')` there's no such field in posts model! Changed the implementation so that the API version is passed in as a parameter to the method instead
- The style of providing "defaults" everywhere creates a need for future maintenance when we bump the version e.g in Ghost v5. Maybe reworking these methods to require a passed version and throwing an error instead would be more maintainable long-term?
refs https://github.com/TryGhost/Ghost/issues/12716
- The default API in Ghost v4 will be `'v4'`. When a new webhook is created or any data passes through serialization it should assume `v4` as a fallback if not specified.
refs https://github.com/TryGhost/Team/issues/498
Requests to fetch the data of the logged in member made without a
session cookie were responsing with 400 Bad Request. This was incorrect
and always should have been a 401 Unauthorized.
refs https://github.com/TryGhost/Team/issues/498
This was filling up logs with less than useful information - for every
single request made by a non-member to the frontend. Be gone!
no-issue
The slack setting in 1.x did not have a `username` property. When
updating from 1.x, the migration to move the slack setting from a JSON
blob to two individual settings, would assume that the setting in the
database would have a value for the `username` property. This resulted
in errors on SQLite "sqlite does not support inserting default values."
The fix here is to add defaults when reading from the database, meaning
that we will _never_ attempt to insert `undefined`
refs https://github.com/TryGhost/Team/issues/476
* Moved paid subscription events population migration
This migration relies on the members_stripe_customers and
members_stripe_customer_subscriptions tables having no orphaned records
in order for it to correcly generate its data.
The migration to clean up orphaned records in those tables has not been
implemented yet, moving this migration free's up the "14" slot
* Removed orphaned stripe data from SQLite3
SQLite databases do not handle removing orphaned stripe records after a
member has been deleted. Our migration to populate the paid subscription
events relies on each customer and subscription being associated with a
member.
refs https://github.com/TryGhost/Team/issues/494
The migrations in 3.1.0 which added email permissions did not add those
permissions to the roles. This means that whilst we have the permissions
in the database, only the Owner role could use any of them.
This migration ensures that the email related permissions are added to
the correct roles.
refs https://github.com/TryGhost/Ghost/issues/12602
As part of the member events, we added a third status of 'comped'.
Members with a status of 'comped' should still be considered paid, so
this fixes the definition of the paid flag to take that into account.
refs 17feb14e4a
- The original commit adding this intended to add transactions, following the pattern of always forcing a transaction when we use bookshelf-relations
- (We use bookshelf-relations here because integrations have api-keys and webhooks associated wtih them and we upsert as one)
- These add and edit methods were inadvertently added to the wrong argument object/section of bookshelf (really fucking easily done, one day we will fix bookshelf so its easier to work with)
- Bottom line: these methods have never been called
- I tried moving them to the right section, but this created test failures throughout our acceptance tests:
- Error: Transaction query already complete, run with DEBUG=knex:tx for more info
- This is likely because we need to account for integrations being used as part of the auth step in the before part of tests
- In terms of yak-shaving, fixing these tests is one step too far right now. I think not having this code here at all is a better state than having it look like it works when it doesn't
no issue
- the analytice-node v4 update contains a breaking change that throws an
exception if the message is over 32kb
- I'm pretty sure we won't hit this, but it's good to track these errors
anyway and stop Ghost from bombing out if there's an issue
- this commit wraps the tracking call and adds Sentry + logging to the
error
refs https://github.com/TryGhost/Team/issues/467
refs https://github.com/TryGhost/Team/issues/221
- we've introduced backwards-compatible changes to rendering in 3.0 such as srcset and sizes which will only have taken effect on posts created or edited since the changes were made
- 4.0 brings additional changes such as image card width/height
- re-generating the `html` field of all posts from the `mobiledoc` brings all content up to latest rendering output
refs https://github.com/TryGhost/Team/issues/469
The initial implementation was just to get us going with the dashboard
but the requirements have changed now, rather than updating the code we
allow to pass the `limit` options so the Admin can choose how many
events to display.
- There is now one true way to start Ghost - you create a server, and then call start with an express app
- We may well expand this again to improve testing pathways in future, but it will be done with a bit more clarity about expectations
refs https://github.com/TryGhost/Team/issues/221#issuecomment-759105424
- Mailgun responds to an email send with a provider id in the format `<x@y.com>` but everywhere else it's used in their API it uses the format `x@y.com`
- updates email batch save to strip the brackets, and migration removes brackets from existing records so we no longer have to add special handling for the stored id any time we use it
refs https://github.com/TryGhost/Ghost/issues/12496
- having the logging require here means that workers wanting to use the db are unable to do so without requiring logging as a side-effect
- `connection.loggingHook` does not appear to be widely used for anything outside of specific debugging scenarios when using MySQL so it should be safe to disable until a proper fix is found for workers+logging leaking file descriptors
refs https://github.com/TryGhost/Ghost/issues/12496
By requiring the models layer the shared logging util was being required as a side-effect causing the open file descriptors problem to continue. Removing logging from the models layer isn't feasible due to deep require chains spreading across the codebase, it's much quicker to remove the need for models in the analytics job.
- models layer was only needed because it's used by the session service
- updated analytics job to create it's own instance of `EmailAnalyticsService` rather than the default instance in order to pass in custom dependencies
- pass in custom `logging` object that uses `parentPort.postMessage` as a way of writing log output
- pass in custom `settings` object that returns settings that have been manually fetched and cached during job instantiation
refs https://github.com/TryGhost/Ghost/issues/12633
Adds new `browse` endpoint for emails that allows Admin to check performance of newsletters over time and show stats on dashboard as primary usecase
refs https://github.com/TryGhost/Ghost/issues/12496
- Handling logging in the main thread avoids file handle leaks which happen due to leaky implementation of bunyan logger (see referenced issue for more context)
- Bumped job-manager version to allow for `workerMessageHandler` callback funciton
refs https://github.com/TryGhost/Ghost/issues/12496
- Using ghost-ignition logging caused file handle leaks. As there is no straight-forward way to handle write streams with bunyan (ghost-ignition's underlying logging library) this method of logging was chosen as an alternative to keep the amount of open file handles to minimum
- The follow up changes will include custom formatter for jobs service which should make logging match the same format as has been used inside the jobs
refs https://github.com/TryGhost/Ghost/pull/12598
- This changeset adds idepmotence to situations where unique contraint has to be dropped or added to the table
- Note '4.0/07-alter-unique-constraint-for-posts-slug.js` was migration that was effected by lack of idempotence
refs https://github.com/TryGhost/Ghost/issues/10318
- Storing JSON object in settings has caused multiple bugs in the past and was considered an antipattern
- This is a last bit in the long process of getting rid of "object" settings in the database. At this point there should be no settings with this type. Yey!
- In the old boot the server wasn't started til we were ready
- In new boot, we start the server immediately and send the old started event
- Then, when we are ready to accept some traffic, we send a ready event
- At the moment, ready isn't quite sent at the right time:
- It _should_ be when we're ready to serve real traffic, not just send 503s
- This is after the URL generation has finished
- But this requires more refactoring work :(
- So for now we send when everything else is ready
- This really needs some tests
- Notify is a more familiar name e.g. systemd has the sd_notify system which this is similar to
- We're actually announcing the server started, it's not actually ready for traffic (will serve 503s)
- use theme.ready for loading themes instead of server.start and properly clean this up
- remove server.start and server.stop as they are no longer used (only server.start was used, and only for themes)
- we're moving away from the pattern of using global events like this as they are hard to reason about
- This commit removes the old boot process and any files that are no longer needed as a result
- Remove the duplicate event for triggering inactive themes to load
- Tidied up a few other bits
no issue
- all migrations should contain a logging message
- info for successful things, or warn if we deviated from what was
expected for some reason
- also added some spacing to make them easier to read
closes https://github.com/TryGhost/Team/issues/468
- updated post-gating
- clears excerpt if there's no access
- rebuilds excerpt from free preview if paywall card is used and there's no custom excerpt
closes https://github.com/TryGhost/Team/issues/466
- upgraded kg-default-cards to include paywall card
- extracted `htmlToPlaintext` from post model to shared util for re-use
- updated post-gating to set html+plaintext to the free preview if a paywall card has been used
- re-generates plaintext from the truncated html using `htmlToPlaintext` util
- display free content in the `{{content}}` helper via the default CTA template
refs https://github.com/TryGhost/Ghost/issues/10318
- API changes introduced:
canary/v4 Admin API
GET /settings/ (browse)
+ "unsplash" present in response as boolean value
GET /settings/:settingName (read)
+ "unsplash" present in response as boolean value
PUT /settings/ (edit)
+ "unsplash" updates setting, accepts ONLY boolean format
v3 Admin API
GET /settings/ (browse)
+ "unsplash" present in response with object value
GET /settings/:settingName (read)
+ "unsplash" present in response with object value
PUT /settings/ (edit)
+ "unsplash" updates setting, accepts either boolean or object formats
v2 Admin API
GET /settings/ (browse)
+ "unsplash" present in response with object value
GET /settings/:settingName (read)
+ "unsplash" present in response with object value
PUT /settings/ (edit)
+ "unsplash" updates setting, accepts object format
refs https://github.com/TryGhost/Ghost/pull/12631
- Initialy writen migration would break if executed twice
- Also accounted for situation if only part of the migration executed leaving only one of the values migrated
- Allow the frontend to accept post messages to generate previews of the frontend
- Created a new endpoint in admin we can use to render these previews, which is possibly not necessary
- Supports a limited group of settings, which can easily be expanded, but care should be taken if expanding to use user-provided strings
refs https://github.com/TryGhost/Ghost/issues/10318
- Because members is effectively "enabled" by default starting Ghost 4.0 have hardcoded labs setting to be such. The alternative of removing this key from labs would be equivalent to `labs.members === false` which is undesireable and would mean additional work on theme developer's side.
refs https://github.com/TryGhost/Ghost/issues/10318
- `labs` setting is dropped from setting values as the use of JSON objec
to sore settings has been deprecated
- `labs` setting is no longer accepted as a paramter in the Settings API nor the
impoprter. The value is ignored if present in the POST/PUT requests and
returns 404 in case it is requested by key at `GET /settings/:key`
refs https://github.com/TryGhost/Team/issues/332
- The last value that has been used in the code was "members"
- By default members will be always "on" starting Ghost 4.0, so there's no need for this flag anymore
- Therefore there's no real need to keep "labs" around
refs https://github.com/TryGhost/Team/issues/467
refs https://github.com/TryGhost/Team/issues/221
- we added robust url transformation in Ghost 3.0 which ensures newly created or edited content is stored with relative URLs in the database
- this migration loops over all posts (and posts_meta) in the database performing the absolute->relative transformation to normalise data for old posts
closes https://github.com/TryGhost/Ghost#12562
- From NodeJS v13 `error.errno` returns error code instead of a string. Because of that use friendly "port is already in use" message did not work anymore.
- Changed to use `error.code` which acts the same way as `error.errno` in older NodeJS versions.
refs https://github.com/TryGhost/Ghost/issues/12646
- adds `ghostVersion` property to mobiledoc of all posts so that we can pin the rendering output of existing content to protect against future breaking changes
- cleans any deprecated markdown card names from old Ghost 1.0 content
- cleans up HR cards that had an [accidental payload added](610d801bcf)
refs https://github.com/TryGhost/Ghost/issues/12646
- if the version is missing then the content was created before Ghost 4.0
- setting the version to `'3.0'` means it will continue to use the same rendering output so there are no unexpected breaking changes when migrating content
refs https://github.com/TryGhost/Ghost/issues/12602
* Updated members_status_events table
By replacing the `status` column with a `from_status` and `to_status`
column, we are able to track the changes between multiple statuses
easier, and accumulate the data. e.g. the delta of paid members in a
given time range is the sum of the `to_status` columns set to 'paid'
minus the sum of the `from_status` columns set to 'paid' within that
time range
* Updated MEGA to handle addition of 'comped' status
With the addition of the 'comped' status, we need to ensure that MEGA
will still send emails to the correct recipients. I've opted to use an
"inverse" filter, as that is the intention of the free/paid split in
MEGA - as far as MEGA is concerned, "free" is the opposite of "paid"
* Updated customQuery for MemberStatusEvent
With the `status` column replaced with `from_status` and `to_status`
this allows us to fix and update the customQuery to correctly accumulate
the data into deltas over time, broken down by day.
* Populated members_status_events table
As the table will be used to generate deltas, we need to backfill the
data so that existing sites will be able to sum up the deltas and
calculate correct data.
The assumptions used in backfilling is that a Member's current status,
is their only status.
