* Step 1
* Step 2
* False sense of confidence
* Couple more before dinner
* Update schema package
* Update format-title
* Upgrade specs file
* Close
* Replace ts-node-dev with tsx, and various others
* Replace lodash with lodash-es
* Add lodash-es types
* Update knex import
* More fun is had
* FSE
* Consolidate repos
* Various tweaks and fixes
* Fix specs
* Remove dependency on knex-schema-inspector
* Fix wrong imports of inspector
* Move shared exceptions to new package
* Move constants to separate module
* Move types to new types package
* Use directus/types
* I believe this is no longer needed
* [WIP] Start moving utils to esm
* ESMify Shared
* Move shared utils to @directus/utils
* Use @directus/utils instead of @directus/shared/utils
* It runs!
* Use correct schemaoverview type
* Fix imports
* Fix the thing
* Start on new update-checker lib
* Use new update-check package
* Swap out directus/shared in app
* Pushing through the last bits now
* Dangerously make extensions SDK ESM
* Use @directus/types in tests
* Copy util function to test
* Fix linter config
* Add missing import
* Hot takes
* Fix build
* Curse these default exports
* No tests in constants
* Add tests
* Remove tests from types
* Add tests for exceptions
* Fix test
* Fix app tests
* Fix import in test
* Fix various tests
* Fix specs export
* Some more tests
* Remove broken integration tests
These were broken beyond repair.. They were also written before we really knew what we we're doing with tests, so I think it's better to say goodbye and start over with these
* Regenerate lockfile
* Fix imports from merge
* I create my own problems
* Make sharp play nice
* Add vitest config
* Install missing blackbox dep
* Consts shouldn't be in types
tsk tsk tsk tsk
* Fix type/const usage in extensions-sdk
* cursed.default
* Reduce circular deps
* Fix circular dep in items service
* vvv
* Trigger testing for all vendors
* Add workaround for rollup
* Prepend the file protocol for the ESM loader to be compatible with Windows
"WARN: Only URLs with a scheme in: file and data are supported by the default ESM loader. On Windows, absolute paths must be valid file:// URLs. Received protocol 'c:'"
* Fix postgres
* Schema package updates
Co-authored-by: Azri Kahar <42867097+azrikahar@users.noreply.github.com>
* Resolve cjs/mjs extensions
* Clean-up eslint config
* fixed extension concatination
* using string interpolation for consistency
* Revert MySQL optimisation
* Revert testing for all vendors
* Replace tsx with esbuild-kit/esm-loader
Is a bit faster and we can rely on the built-in `watch` and `inspect`
functionalities of Node.js
Note: The possibility to watch other files (.env in our case) might be
added in the future, see https://github.com/nodejs/node/issues/45467
* Use exact version for esbuild-kit/esm-loader
* Fix import
---------
Co-authored-by: ian <licitdev@gmail.com>
Co-authored-by: Brainslug <tim@brainslug.nl>
Co-authored-by: Azri Kahar <42867097+azrikahar@users.noreply.github.com>
Co-authored-by: Pascal Jufer <pascal-jufer@bluewin.ch>
* Expose schema endpoints
* respond with 204
* fix payload for export flag
* allow export as YAML
* use SchemaService
* move getStringFromStream to utils
* clearer exceptions with logs
* check empty request body
* specs
* add bypassCache flag and remove flushCaches calls
* Identify integer typed fields in Oracle
* Fix uuid foreign keys type for Oracle
* Detect autoincrement for CockroachDB
* Bypass cache for fields and relations services
* Add same DB tests
* Test if working for all vendors
* Add schema snapshot testing into sequential flow
* Fix schema re-apply attempts when it's a different database vendor (#14816)
* prevent diffing on certain properties
* fix apply-snapshot test
* prevent updates to autoincrement fields' schema
* sanitize relations
* remove `colleciton.schema.sql` parsing in test
It is now being sanitized by `sanitizeCollection()` function in get-snapshot-diff.ts
* Change concurrency group to not cancel test running on main
* remove multipart for apply & add hash checking
* check incoming version & add force query param
* refine error message & stub for nested hash
* add `vendor` top-level property
as an additional safety net for cross db vendor applies
* sanitize generated schema snapshots
* snapshot joi validation
* diff joi validation
* minor cleanup
* extract applyDiff & use deep-diff applyChange
* use applyDiff in schema service
* Mark vendor as optional
* Update tests to apply diffs
* move force flag check into service
* Patch mssql vendor typo
* Set relation no action in cyclic relations
* Update mysql timestamp default value
* Oracle cannot define no action
* Update oracle timestamp type default values
* add hash checking for outdated diffs
* fix diff service & endpoint
* Add hashes when returning after
* Fix self referencing relations for Oracle
* Add temp fix for CURRENT_TIMESTAMP defaults in Oracle
* clean up driver and database client types
* only require diff hash if not kind N
* update hash comparison logic for create/delete
* Set no action constraint for self referencing M2M in MSSQL
* Add basic hash check tests
* omit default value when hashing auto increments
Specifically for CockroachDB with differing sequences
* add vendor check
* update specs
* Validate vendors with type definition
* Spread the vendors input array
* re-add Export component
* re-add js-yaml to root
* Propagate mutation options for schema apply
* Verify that snapshots differ and clear cache before hash tests
* Fix unit test
* Revert temp fix for CURRENT_TIMESTAMP defaults in Oracle
* Define and reuse type for export format
* Define and reuse list of database clients
* change `were` to `was`
* change `where` to `were`
* add some empty lines for readability
* tweak exception message
* fix test
* use object-hash
* use boolean to check whether file is included
* simplify request content type check
* throw error when multiple files were uploaded
* use nullish coalesce instead of short circuit
* Update api/src/services/schema.ts
Co-authored-by: Rijk van Zanten <rijkvanzanten@me.com>
* Remove unnecessary `Omit` on `SnapshotWithHash` type
* Revert "Remove unnecessary `Omit` on `SnapshotWithHash` type"
This reverts commit d22ac771ec.
* check empty snapshot earlier
* use allow-list logic via pick instead of omit
* Update api/src/services/schema.ts
Co-authored-by: Pascal Jufer <pascal-jufer@bluewin.ch>
* Move to own /schema endpoint controller
* Fix refs to schema endpoint
* move streamToString to utils package
* move get-versioned-hash and add test
* extract kind into an enum
* Fix mysql5 timestamp value
* Fix test collection existing on rerun
* resolve TODO comment in blackbox test
* Drop deep level hashes in diff
These hashes are used only for more accurate error reporting
but are otherwise superfluous, since changes are already
detected by the top level hash.
Therefore we remove them in favor of a simpler diff format and
implementation.
* Revert schema "fix" for createItem, add comment
* Strict diff schema validation
* Revert CrDB auto-increment detection patch in 816c998
* Clear systemCache to expose newly added fields
* Use DiffKind constants
* Extract diff & snapshot validation into own tested utils
* Apply suggestions from @azrikahar
* Update knex-schema-inspector to 3.0.1
Includes the fix for CrDB auto-increment detection (knex/knex-schema-inspector#135)
* Update knex-schema-inspector in packages
* Update lock file
* add test for schema service
* add test for export service
* add relevant tests to modified util functions
* fix csv test to account for os end of line
* fix files controller test
* dedupe test data for schema service
* Align schema specs with docs
* Update api/src/controllers/schema.ts
* Revert testing for all vendors
---------
Co-authored-by: Rijk van Zanten <rijkvanzanten@me.com>
Co-authored-by: ian <licitdev@gmail.com>
Co-authored-by: Pascal Jufer <pascal-jufer@bluewin.ch>
* Speed query up by reusing existing aliases which reduces table joins
* Use subquery in top level m2o to remove duplicates
* Fix linting
* Apply distinct on primary key field in subqueries
* Use distinct instead as there are only primary keys
* Apply subquery on top level
* Try remove sub sub query
* Test if working for all vendors
* Add support for _none and _some
* Use subquery only when field depth > 1
* Add tests
* Use original table names for columns with functions (#14690)
* Use original table names for columns with functions
* Extract filter function path parsing as shared util
* Fix filter function path when adding node
* Pass the originalCollectionName into filter functions
* Update unit test
* Replace functions within deep GraphQL
* Fix invalid operator error for _none and _some
* Add filter function tests
* Revert triggering for all vendors
* Simplify aliasMap
* Replace functions in filter within GraphQL aggregate query
* Add API support for filtering of alias field
* Mark schema as optional
* Shift logical operators upwards
* Separate recursive parseFilter
* Rework shifting of logical operators
* Error on invalid usage of _none and _some
* Use inner join to preserve sort order
* Run tests for all vendors
* Reuse aliasMap for sort and filter
* Sort on top level query
* Remove unnecessary limit on wrapper query
* Refactor applyQuery options
* Remove duplicates from nested multi relational sort
* Fix offset in MSSQL requiring OrderBy
* Disable schema cache
* Use inner query only for nested sort or multi relational filter
* Fix MSSQL duplicate order column
* Use inner query only for multi relational
* Additional integration tests
* Order within partition for multi relational sorts
* Rename to directus_row_number
* Fix unit test
* Add base sort and filter tests
* Fix Oracle uppercased rowNumber column
* Fix unit test
* Fix top level query sort with function
* Parse functions in inner query
* Increase clarity with knex.ref()
* Remove sort filter for top level primary key
* Fix unit test
* Bypass queries with groupBy
* Add collection to aliasMap to fix functions in nested sort
* Fix multi relational sort with functions
* Add tests for filter and sort with functions
* Fix accidental deletion of brackets
* Fix top level alias filter node interface
* Update M2M sort tests
* Add M2A tests
* Cast m2a primary key as varchar2 for oracle
* Enable filtering tests for M2A
* Fix prototype polluting assignment in aliasMap
* Remove unnecessary currentKey
* Simplify code to increase readability
Co-authored-by: Brainslug <br41nslug@users.noreply.github.com>
* Fix linting and missing 'this' error
* Revert optional chaining
* Add mysql5 to tests
* Fix mysql5 missing rowNumber()
* Overcome indexing delays in MySQL5
* Verify MySQL5 sorting is in order as the result count varies between runs
* Skip joining when sorting field already exists
* Simplify variable assignment
Co-authored-by: Azri Kahar <42867097+azrikahar@users.noreply.github.com>
* Fix linting
* Reduce duplicate logic with vars
* Transform _func fields in GraphQL only for valid functions
* Fix unit test
* Fix unsupported date_part() in CrDB
Co-authored-by: Brainslug <br41nslug@users.noreply.github.com>
Co-authored-by: Roger Stringer <roger@directus.io>
Co-authored-by: Azri Kahar <42867097+azrikahar@users.noreply.github.com>
Co-authored-by: Rijk van Zanten <rijkvanzanten@me.com>
* Add black box tests
* Revert docker compose file
* Update workflow
* Try use workflow from dev repo
* Increase seedDB() timeout
* Disable other checks for now
* Change DB sequence
* Update jest moduleNameMapper
* Update workflow's docker-compose.yml path
* Slice array first
* Remove differentiation of status code
* Delete field only after foreign key constraints are removed
* Add checks for different types of primary key
* Test global query filter for all field types
* Increase timeout for m2o seeding
* Add case insensitive string operators
* Update filter check to run on relational fields
* Enable time field checks
* Add seeded random and fix relational seeding
* Add casting for integer and bigInteger
* Minor fixes
* Reduce bigInt values
* Separate seeding of DB structure from values
* Add primaryKey seeding function
* Use automatic IDs except for string pk
* Try fix ci
* Update package-lock.json
* Update common.test for concealed user tokens
* Use dynamic field type for m2o.test relational fields
* Temporary disable missing nicontains for string type
* Add support for alias type filtering
* Fix relational filter operator checks
* Add initial o2m test
* Remove integer pk limit
* Add empty checks for string and uuid null
* Limit generated integer value to 4 bytes
* Patch timezone tests for MSSQL
* Remove sample query filter test
* Fix timezone test for sqlite
* Fix MSSQL uuids
* Fix MSSQL timestamp inaccuracy
* Cast datetime schema to milliseconds for comparison
* Fix MySQL / Maria timestamp inaccuracy
* Fix MySQL / Maria between operator inconsistency for float type
* Fix missing time datatype in Oracle
* Skip filter testing on Oracle
* Enable o2m filter tests for other collections
* Run tests only on SQLite for PRs unless the Full Tests label exists
* Try fix actions
* Refactor github actions
* Update tests flow setup to use getURL()
* Start postgres docker
* Reinstate package-lock
* Fix geometry test
* Remove .gitkeep files
* Add todo.md
* Rename black box to blackbox
Co-authored-by: rijkvanzanten <rijkvanzanten@me.com>
