- The delta update retry reconnection logic had an off by one error, causing the error log to occasionally say "reconnected after 0 retries".
- Previously, the delta update retry reconnection logic was hardcoded to wait 30 seconds when attempting to reconnect. Now it starts at 5 seconds and increases in multiples of 5 seconds until it gets to 60 seconds, then keeps attempting to reconnect after 60 seconds perpetually afterwards.
- In our logs, chalk.white.bgBlue() was changed to chalk.bgBlue.white() to support light coloured themes. Previously there wasn't enough contrast when a light theme was enabled.
- fixed case where schema customization might fail if a type had no subfields
- Fixed default resolver for auto aliased fields in the case that the field type is nested in ofType. GatsbyWPGutenberg/gatsby-wordpress-gutenberg#44 gatsbyjs#79
- Tested WPGraphQL v0.12.0 and adjusted the compatibility API to allow it.
- Fixed a regression where htaccess headers were not being properly passed to image file requests when they should be.
- Errors with no errorContext object would sometimes print "false", this now prints an empty string instead, followed by the proper error.
- Removed some debugging code that made it to master 😱
- Improved the error displayed when a GraphQL request is redirected or when a PHP WP filter adds additional output to a GraphQL request
- Previously when in
gatsby developor in a Preview instance, if the connection to WP went down for a moment it would fail the build. It now displays an activity timer with the number of times it's retried, and then a success message when it finally does succeed.
- Added logic to receive delta updates for non-node root fields like options and settings. Requires WPGatsby v0.4.15 to work, but nothing bad will happen if you don't update WPGatsby, you just wont get any delta updates for these fields until you do.
- htaccess auth headers were being passed to all media item file urls regardless of wether that url was the WP url or another url (like s3). This was causing 400 errors in some cases as the 3rd party server considered these headers malformed. This is now fixed.
- Updated Schema compatibility so that WPGraphQL v0.11.0 will be supported.
- Added Date resolver to
modifiedmodifiedGmtanddateGmtfields (#83)
- Null values in html image caching were sometimes throwing errors. There are now guards against null values here.
- Top level inline fragments (for WP node interfaces that are not Gatsby @nodeInterface types) were not being generated along with regular fields. This caused WooCommerce price fields (and many other fields) to not be fetched even though the schema was properly generated for these types.
- Bumped minimum WPGatsby version because the latest fixes a bug where saving a published post as a draft wouldn't delete the corresponding node.
- Updates during development were no longer automatically showing up in Gatsby due to inconsistencies in the latest version of Gatsby core
- Added a link to the wp-admin plugins page to update WPGatsby if your version is not within the accepted range.
- Split the WPGatsby and WPGraphQL upgrade reason error messages depending on wether one or both need to be updated.
- Improved the error message for GraphQL request timeouts via #86, thanks @jacobarriola!
- Some fields on the
UserandPagetypes which are by default private were being automatically excluded via default plugin options. These fields have very low overhead and in some cases devs will filter these fields to make them public. This manifests to some devs as a bug because fields they're looking for don't exist. As of this release they're no longer excluded by default.
- The root field data node was not being properly processed for html images, links, and referenced media node ids. This resulted in some media items going missing on root fields like options pages.
- Absolute path images with no image node were causing errors because the hostname was not prepended in the case that no node exists.
- Images in code blocks and pre tags were being transformed when they shouldn't be.
- Images nodes fetched by url were sometimes being double fetched
- Images with no media item node or file node were being incorrectly cached.
- Cast filenames to strings to prevent 0.jpeg from being taken as "no file name at all".jpeg 🤦♂️
- The last release introduced and error
absolutePathUrls is not defined
- Images with absolute paths were not being properly recognized. For example if you had an image with a src of
/content/2020/01/06/my.jpeg, this would resolve in vanilla WP, but the source plugin wasn't recognizing it. The solution is to find absolute paths and fetch them with the hostname of the WP url attached to the beginning. - Disallowed use of MediaItem.limit option as this option only introduces bugs and odd behaviour due to the special handling of this node type internally.
- Default exclude User.UserToMediaItemConnection field as this will cause many sites to fetch 100's more media items than needed. If you need this field you can enable this yourself by adding
type: { UserToMediaItemConnection: { exclude: false } }to your plugin options. - More file types than were required were being regexed for when searching for images in html fields. Now the types that are regexed for are
jpeg|jpg|png|gif|ico|mpg|ogv|svg|bmp|tif|tiff. - When accessing previously cached image nodes, protect against missing id's
- Connection fields to interface types that consist entirely of Gatsby node types were not being handled properly in the query generation and schema customization steps. This is due to latest WPGraphQL using interface types for node connections in places where it previously wasn't, so this bug wasn't previously discovered.
Bumped minimum WPGQL version to 0.10.3 to prevent folks from running into a WPGQL resolver regression that caused some builds to fail.
- Images fetched from HTML which had no corresponding WPGQL MediaItem and were just fetched as regular Gatsby File nodes weren't always being properly cached.
- Images fetched from HTML didn't have htaccess creds passed onto them, resulting in 401 errors.
- If all of a types child fields were excluded, that type wouldn't also be excluded
- New post draft previews and draft previews weren't working anymore. Minimum WPGatsby version has been increased in order to fix this.
- When an error path didn't exist in a GraphQL error, a property on undefined was being accessed which threw an unrelated error.
- Lists of connect Gatsby nodes resolvers weren't properly returning empty arrays and instead would return
nulleven if the remote schema returned an empty Array. - Literal
nulls in arrays of connected nodes were failing the build
- Added the error path from GraphQL errors to error output. For example: "Error path: mediaItems.nodes[77].mediaDetails.meta.focalLength"
This release adds no changes! This is the point at which we've decided to move this project into beta. 1.0.0 signifies this.
Added a new api addNodeFilter and applyNodeFilter which work similarly to filters in vanilla WP. It allows any plugin to add a function to filter some data and then any other plugin to apply all registered filters. This is currently undocumented because it's not a finished or tested API.
- Gatsby images in html do not currently lazy load properly. This will be fixed in a later release but for now this means we need to hide the placeholder on load using an inline css style.
- Images that weren't transformable by Sharp were still being transformed which was causing problems
- When using the hardCacheMediaFiles option, file names were sometimes undefined or the extension wasn't properly handled
- Html image widths weren't being properly inferred in all cases due to a problem where a variable wasn't properly cast as a Number
Changed verbose plugin option to be true by default. This is a smarter default as it's more useful to see what the plugin is doing when you're first using it. If you don't want to see all the output it's easy to turn it off.
- htaccess password and username were not being passed into the
createRemoteFileNodehelper which meant builds would fail if these files were protected.
- GATSBY_CONCURRENT_DOWNLOAD couldn't be set lower than 3 without erroring.
- Referenced MediaItem nodes were being incorrectly ignored when
html.useGatsbyImagewas set tofalse. This led to tons of images being fetched in resolvers instead of after node sourcing is complete which is problematic for some servers and causes the build to fail. - Added error context messages to every instance of
fetchGraphQL()to help give context on when a gql error occurred during the build.
Updated plugin to work with WPGraphQL 0.10.0 which provides better menu and preview support. This is listed as a minor because supported min version has been bumped and WPGraphQL has a lot of breaking changes for this release. Head to https://github.com/wp-graphql/wp-graphql/releases/tag/v0.10.0 for more information on updating.
The dateGmt field which was previously mistakenly removed has been added back.
- The non-node root query was ignoring GraphQL errors
- Referenced MediaItem node queries were missing any generated fragments
pluginOptions.html.useGatsbyImage: falsewasn't preventing files from being downloaded in all cases.
- Added an option,
pluginOptions.type.MediaItem.localFile.excludeByMimeTypesto disable fetching files associated with MediaItem nodes by mime type.
- Changed relative docs links to full URL's in the main README to work better with Gatsbyjs.org and npmjs.com
- Removed Gatsby, gatsby-source-filesystem, and gatsby-transformer-sharp from dependencies list. These were causing yarn to error complaining about yarn link when installing this package. This may also have been causing OOM issues when installing.
- Added a clearer error message about firewalls and firewall plugins when 403 errors are returned when making GraphQL requests.
- The TermNode type was not being properly recognized as a Node Interface type. Because of this, terms were being double fetched and then nodes were being created twice. The second time the node was created it would be missing data because it was fetched on the term interface the second time. This release marks TermNode as a Node Interface similar to ContentNode, and that fixes this issue.
- In some instances, using the
MediaItem.lazyNodesoption in combination withhtml.useGatsbyImagewould cause build errors.
- pathPrefix wasn't set up properly in 0.7.6. This releases fixes that. Thanks @trevorblades!
- pathPrefix option wasn't being used for inline images in html
- Url's needed to be encoded in createRemoteFileNode() to account for filenames with special characters
- Automatically excluded EnqueuedAsset, ContentNodeToEnqueuedScriptConnection, ContentNodeToEnqueuedStylesheetConnection, TermNodeToEnqueuedScriptConnection, TermNodeToEnqueuedStylesheetConnection, UserToEnqueuedScriptConnection, UserToEnqueuedStylesheetConnection types because these types can't be properly utilized yet without causing errors.
- Sometimes at the bottom of our query depth limit, fields which require a selection set were being queried without.
- Fixed an error for non-image media items where the build would fail since 0.7.1. The problem was that we were trying to access the media item by
sourceUrlbut non-image media items only have amediaItemUrl
- fixed an issue where incremental data fetching could error when some fields don't exist
- In the event that the remote schema has broken pagination which causes infinite looping between pages of remote data, there are now some safeguards to protect against this. In a way this is a bug fix, but I'm listing it as a feature because the bug is on the remote server instead of within the source plugin.
- Inline links in any node content (custom fields or in post_content) will be replaced with local relative links. https://your-beautiful-wp-site.com/page-2 will become /page-2 so that Gatsby can make sense of it.
- gatsby-plugin-catch-links is automatically installed as part of this plugin so that inline-html links work as gatsby-links out of the box.
- Inline html images anywhere in your node data are now gatsby-images. These are processed as fluid Sharp images, the media item node from WPGraphQL is fetched and added to your cache, the maxWidth of the fluid resize is inferred from html (if the img tag either has a sizes or width attribute, those are used) otherwise it falls back to a default plugin option:
html: {
// this causes the source plugin to find/replace images in html
useGatsbyImage: true,
// this adds a limit to the max width an image can be
// if the image selected in WP is smaller, or the image is smaller than this
// those values will be used instead.
imageMaxWidth: null,
// if a max width can't be inferred from html, this value will be passed to Sharp
// if the image is smaller than this, the images width will be used instead
fallbackImageMaxWidth: 100,
imageQuality: 90,
},If you delete an image in the media library which was uploaded to posts, you'll get a handy console warning telling you which post has the deleted image along with an edit url you can click to fix the problem.
- Before plugin options were being merged into default plugin options (so no fallbacks for nested options), we were trying to access nested properties on undefined which was causing errors. This release fixes that and adds an integration test for this.
- WPGraphQL and WPGatsby minimum versions have been bumped to 0.9.1 and 0.4.0 due to an oversight in how Menu Relay id's were constructed. Using WPGQL 0.9.0 and WPGatsby 0.4.0 would lead to inconsistent caching behaviour for menus.
- Added a minimum version "reason" field to supported plugin versions to add an explanation for the minimum versions.
- WPGraphQL and WPGatsby minimum versions have been bumped to 0.9.0 and 0.3.0 due to the structure of Relay id's changing in WPGraphQL. This is breaking for us because id's changing will result in inconsistent cache behavior.
pluginOptions.schema.perPagewas added to control how many nodes are fetched per-page during node sourcing. This is helpful for sites with gigantic schemas that generate very large queries. In the future queries will automatically be split into multiple queries to mitigate this automatically but for now this option will do 👍- Error context is now displayed when fetch errors occur, not just for GraphQL errors. Error context is something like "Error occurred while fetching the "Product" node type.
- The
awaiting async side effectsreporter status is now only shown once some async side effects have occurred.
- We were trying to fetch connections to WPGQL node interface types that don't have an id field by id. The schema should provide an id here, otherwise there's no way to id the connected node, but now that's protected against by checking if the field has an id before trying to fetch the id.
- The message
pluginOptions.schema.queryDepth is not a positive integer. Using default value in place of provided value.was being displayed when no plugin option for queryDepth was added. This release prevents that as we only want to display a warning if a value is provided.
- Added additional error context for GraphQL request errors to print out which node type was being sourced when the error occurred.
- Node interface types on the WPGQL side weren't being properly recognized, this release fixes that by using the __typename field to identify which type node interface types should be stored as during node sourcing
- Reusable fragments were sometimes being nested inside themselves which would throw an error. This is now fixed.
- Added plugin option
debug.timeBuildStepsto add an activity timer to all the internal build steps the plugin goes through.
- The minimum WPGatsby version has been increased to 0.2.5. This is because earlier versions were recording up to 4 duplicate content saves per content change in WordPress. This is the minimum version now because earlier versions may bloat your DB. WPGatsby does garbage collection, so any duplicate actions will be automatically removed.
- When in verbose mode, content diffs are displayed in the terminal output when content changes. When the field contained a lot of data this was really noisy. Now field values that return more than 250 characters aren't shown - instead, the field key is simply printed as
[gatsby-source-wordpress] fieldKey updated
- The generated RootQuery GraphQL query is now written to disk when using the
debug.graphql.writeQueriesToDiskoption. - Better error handling by printing out which step of the build caused an uncaught error below the stacktrace.
- Added an internal plugin options filter/validator and enforced that the
schema.queryDepthoption is a positive integer.
- Previously
schema.queryDepthdidn't work when set to 1. Now you can do that if you're so clean that you only need the top level of WPGQL fields!
- Any field named
dateis now treated as being ofDatetype and can make use of thedateFormatinput args. In the future WPGraphQL will have a Date Scalar and that will be used to determine what should be a date instead of the field name.
- Fields named
dateGmtare automatically ignored
- Fixed an issue where ommitting the new
debug.graphql.writeQueriesToDiskwould cause build errors
- Added plugin option
debug.graphql.writeQueriesToDiskwhich writes out all the internal GraphQL queries to./WordPress/GraphQL/[typname]for each node type. - Automatically generate fragments when types are infinitely nested within themselves. This makes fetching circular references more efficient and prevents running out of memory. wp-graphql-gutenberg and wp-graphql-woocommerce now appear to work!
- Increased default query depth and circular query limit since queries are more efficient now.
- Added the ability to exclude fields on the RootQuery via plugin options.
- Removed some fields that require auth by default:
RootQuery: {
excludeFieldNames: [`viewer`, `node`, `schemaMd5`],
},
Settings: {
excludeFieldNames: [`generalSettingsEmail`],
},
GeneralSettings: {
excludeFieldNames: [`email`],
},- When generating queries fields which are circular between 2 types are now disallowed and not fetched. This indicates that these are connections which can't be identified as nodes so there is no efficient way to fetch them. They are excluded and the schema author should make these connections identifiable as nodes by adding an id field to them.
- Switch from graphql-prettier to prettier since it turns out the former is not very accurate. This was a minor bug but could affect debugging accuracy when queries were prettified.
- Non node root fields which take any input arg with a type of ID are automatically ignored now. They are almost definitely unusable without input args.
- Renamed excludeFields to excludeFieldNames to keep the API consistent with the Type.excludeFieldNames option.
- Added a new plugin option for HTTP Basic authentication:
{
resolve: `gatsby-source-wordpress-experimental`,
options: {
auth: {
htaccess: {
username: `username`,
password: `password`,
}
}
}
}- Fixed regression in the generated resolver for lists of unions in
src/steps/create-schema-customization/transform-fields/transform-union.js. Thefieldvariable was being declared twice and accessed before it was initialized the second time. - Fixed a query generation / node sourcing bug where fields that should have a selection set were being queried as if they didn't which would fail the build during node sourcing. The issue was due to the new
schema.circularQueryLimitoption which limits circular query generation separately from the overallschema.queryDepthoption. Circular field references at the bottom level were sometimes missing their selectionsets.
- Added plugin option
debug.disableCompatibilityCheck. This is useful for testing the source plugin against versions of WPGraphQL outside the current accepted version range.
- Type.exclude was not removing types from inline fragments during node sourcing, that is now fixed.
- Auto aliasing of conflicting field types in inline fragments is now recursive into nested fields.
- Added proper field def to resolve non_null lists of non_null types.
[Type!]!
- Added a new plugin option
schema.circularQueryLimitwhich is used to set a limit on how many times a field type can be an ancestor of itself during query generation for node sourcing. This should help prevent out of memory issues for gigantic schemas with fields that are potentially infinitely nested. The default limit is set to 2 but this can be increased. - exclude editLock and revisionOf fields by default as these fields require authentication.
- remove reliance on WPGatsby's postTypes field and use inputFields from introspection to determine which node list queries require the temporary
where: { parent: null }input args to get a flat list of posts/pages. This slightly speeds up the sourcing process. - Added a plugin option for debugging node list query generation.
debug.graphql.copyNodeSourcingQueryAndExitexpects to be passed the type name of a WPGraphQL node such asPage. If the Gatsby site is in development mode and a valid type is passed to this option it will write the node list query to be used in node sourcing to the system clipboard and exit the build process.
MediaItem.remoteFile was deprecated and renamed to MediaItem.localFile. This more closely aligns with other Gatsby source plugins schemas.
The User.description field was mistakenly excluded by default, this release adds it back to the schema
Removed unecessary logic when fetching menu items that could prevent pulling some types of child items
-
There is now a
type.__alloption which allows you to pass options to all types instead of only to specific types. -
Because of the way menu items work in WPGraphQL, any child items in a menu need to be fetched recursively. Since this part of the build process didn't have any cli reporting the build appeared to hang for sites with a lot of menu items. This release adds logging to that part of the build step, explicitly telling you if there are async side-effects happening with a readout of how many additional nodes were created.
success gatsby-source-wordpress creating nodes - 57.784s - awaiting async side effects - 710 additional nodes fetched
In addition, this release moves recursive menu item sourcing into an async queue so we can increase concurrency and speed this up a bit. In the future this will be less of an issue when WPGraphQL moves most node queries to a flat architecture.
When fetching data for interface types, shared fields were being fetch on each inline fragment like so:
{
contentNode {
title
... on Post {
title
otherPostField
}
... on Page {
title
otherPageField
}
}
}Normally that wasn't such a big deal since it just made the queries during node sourcing larger but didn't break anything. For interfaces with particularly deeply nested fields this was a huge problem (namely for Gutenberg). This release solves this by only fetching these shared fields directly on the interface field.
{
contentNode {
title
... on Post {
otherPostField
}
... on Page {
otherPageField
}
}
}- Previously root fields that were lists of non_null built in Scalars or non_null lists of Scalars on RootQuery fields that weren't lists of nodes could throw errors in some cases. This release fixes that.
- Using the types.TypeName.lazyNodes option now works properly. Essentially what this option does is prevent fetching remote files and processing them via gatsby-image/Sharp unless they're queried for. When queried for, remote files are fetched in the gql resolver if the file doesn't exist locally. It's not recommended to use this if you're using gatsby-image a lot as fetching images this way is much slower. You might want to use this if you mostly use the original WP hosted media files 95% of the time and then just use a few gatsby-image's locally.
- Changed accepted WPGatsby version range to ~0.2.3
- Because everything is fetched in a flat list, hierarchical terms weren't being properly sourced. Code was added to specifically support the Category type, but this will be made generic for all types soon as this is a reocurring problem.
- Changed accepted WPGatsby version range to ~0.2.2
- Changed accepted WPGraphQL version range to ~0.8.3
- Removed custom WpContentTypes type and contentTypes field as WPGraphQL 0.8.3 has this built in now
- Scoped babel plugin source-map-support to just development env to prevent
warn Module not found: Error: Can't resolve 'fs' in warn Module not found: Error: Can't resolve 'module' inerrors
- The
copyQueryOnErrorplugin option was throwing cryptic errors on systems that don't support copy (namely CI). Now this is in a try/catch and the error is tossed away. This helps ensure users see relevant errors.
- Fixed normalizeUri helper to account for null uri (if a node has no uri)
- Improved fetch error messages. Some users were getting confused when they added www. to their api url setting. Visiting that URL in browser brought them to the GraphQL api endpoint. The problem is that WP seems to sometimes redirect in browser and axios can't handle this. The new error messages account for this.
- For fields that are connections to lists of nodes, default variables were added to grab the first 100, before the max was 10. In the future an API will need to be added to resolve these lists of connections on the Gatsby-side, for now this works for a good deal of use-cases
- Adding Preview support in an earlier release broke inc-builds in an effort to speed up previews. This release restores inc-builds functionality
- Fixed lists of non_null types which have their type on type.ofType.ofType instead of type.ofType
- Lists of MediaItems were not being recognized as media files that are referenced. This means those media items weren't being sourced as we only source referenced media items. This version fixes that issue!
- Fixed an error where queries return null for some posts and we were checking properties on null. TylerBarnes/using-gatsby-source-wordpress-experimental#6
- Updated Readme for npm
- In the schema, lists of non null types weren't being properly ingested. For example a NON_NULL list of Blocks. This is now fixed! Thanks Peter Pristas!