refs d5f03ec0b1
- underlying error message varies across node versions so the content-length can't be fixed
- applied any-content-length matcher to the right test this time
closes https://github.com/TryGhost/Team/issues/1916
closes https://github.com/TryGhost/Team/issues/1917
- Added database storage for link redirects and click events via repositories (hides away database layer) defined in the wrapper services
- Added LinkClickRepository to store click events to database
- Added LinkRedirectRepository to store link redirects to database
- Added PostLinkRepository to link LinkRedirects with posts
- Renamed link-replacement package to link-replacer, and made it dependency less (it only replaces links now, doesn't do anything else)
- The link-tracking service has a new `addTrackingToUrl` which returns a new URL that includes tracking. The new `addRedirectToUrl` method does the same but without tracking for now.
- MEGA service now uses the link-replacer to replace links in the emails using a combination of different services (member attribution + link-tracking service)
no issue
- setting the original error message to the `context` property means it's not completely lost and gives us a clue for debugging
- updated the lexical and mobiledoc validation errors to use the messages+tpl pattern
fixes https://github.com/TryGhost/Toolbox/issues/416
- this commit switches the default endpoint for our CDN-loaded assets to
a Ghost-specific jsDelivr one, with a shorter 10 min browser cache config, so
assets are refreshed quicker upon publishing
closes https://github.com/TryGhost/Team/issues/1924
- updated the Post model so the relevant `mobiledoc`/`lexical` field always has a blank document by default depending on which editor is selected via the feature flag
- ensures we don't send `{mobiledoc: null, lexical: null}` inadvertently creating a mobiledoc post (API sets the default blank document in that scenario) when the lexical editor flag is enabled
no issue
- bumped `@tryghost/url-utils` to get access to the new lexical transform utilities
- updated the Post model's `parse()` and `formatOnWrite()` methods to transform the `lexical` field contents when reading/writing to ensure any links in content point at the correct place with `site.url` config changes
refs TryGhost/Team#1907
- calculates final attribution source and medium using captured referrer information in history
- adds new referrer-translator that goes through available history and based to determine most valid referrer info
- includes referrer url, source and medium in the attribution data for storage
refs TryGhost/Team#1906
- captures referrer url and any source/medium param for referrer
- maintains history with latest referrer information when moving around pages
closes https://github.com/TryGhost/Ghost/issues/15424
refs b9ab1002b3
refs 9210ac7dc7
- When errors were throw from the model layer the HTTP API responses contained misleading information about the error - e.g.:
"message": "Unknown error - TypeError, cannot save post."
- This was due to the error handling code throwing it's own internal error resulting in compete loss of the original Ghost's error context.
no issue
- adds custom `onError` handler that both logs to console (as per default behaviour) and also logs to Sentry with the `lexical` tag added for easier filtering
- does not re-throw so Lexical will attempt to gracefully recover without losing data
no issue
- The explore endpoint needs to expose the total amount of published posts
- To be more consistent, this PR creates a PostStats class which is exposed as `stats` method within the PostService; just like it's done with the MemberService
- Moved existing method to return the date of the most recently published post into the stats service
- Updated the explore service test to reflect the new return property
refs 08b099e527
- when the use of scratch values was removed a bug was introduced where only the first character of the tag title was used to set the slug when typing
- bug was caused because previously we were checking against the original `tag.slug` value which wouldn't have been set yet because we were setting scratch values instead
- updated the code to explicitly mark the slug value as being set manually if it's edited directly so the auto-update logic can use that to skip updates instead of comparing against old/unset values
refs 58b0a1b90d
- `buildQuery` in the post/page adapters was recently removed which meant it was falling back to the `EmbeddedRelationAdapter.buildQuery` method which uses the defined Ember Data relationships to build up an `?include=` param
- the Ember Data relationships are not fully in sync with the API because we don't have models for all of the members/tiers related relationships meaning we were adding a more restrictive `?include` parameter than what the API would use internally by default, breaking parts of the app that expected the default embedded relationships
