* Start local schema registry
* Try to use latest version
* Revert "Try to use latest version"
This reverts commit 682385c0e4.
* update schema registry jenny to validate new lineages
* save kind instead of lineage
* handle plugins
* get published schemas from GH + fix plugins
* handle kind not yet published
* Add script to use in workflow + handle maturity
* first pass on publish-kinds GH workflow
* fix script path
* remove unused script
* use make gen-cue instead of script
* trigger publish-kinds on every commit (for test)
* temporary update to use specific thema commit
* wrapping errors
* remove GH token & refactor for rate limit
* Update publish-kinds.yml
* Update publish-kinds.yml
* revert go.mod and go.sum updates
* use Thema specific commit
* Kind registry v2
* fix script path
* fix second script path
* update go.mod
* update schema registry source
* test checks
* add GITHUB_TOKEN
* revert test checks
* actually write next files when publishing
* Add kind set arg
* Add comments
* clean up workflows
* update Thema
* Update .betterer.results
* few fixes after lineage flattening
* Update publish-kinds-next.yml
* add codeowners for new files
* update thema
* apply review feedback
* update go version in workflows
* clean up workflows
* Add step to generate token and test
* Update publish-kinds-next.yml
* fix script
* try with the app name
* Update publish-kinds-next.yml
* clean up and update release workflow
* add comment
* publish kinds only on cue updates
* extract kindsys
* reinstate kindsys report
This may end up living somewhere else (or not! who knows!), but the important part is that I don't get rid of it right now :)
I hate the package layout (kindsysreport/codegen) for the main function and will take pretty much any alternative suggestion, but we can change also change it later.
Note that the generated report.json is in a different location - anything using this (ops something) needs to be updated.
* kindsysreport in codeowners
* New pfs impl
* Reached codegen parity with old system
* Update all models.cue inputs
* Rename all models.cue files
* Remove unused prefixfs
* Changes Queries->DataQuery schema interface
* Recodegen
* All tests passing, nearly good now
* Add SchemaInterface to kindsys props
* Add pascal name deriver
* Relocate plugin cue files again
* Clarify use of injected fields
* Remove unnecessary aliasing
* Move DataQuery into mudball
* Allow forcing ExpandReferences on go type generation
* Move DataQuery def into kindsys, add generator to copy it to common
* Fix copy generator to replace package name correctly
* Fix duplicate type, test failure
* Fix linting issues
* Get pluginmeta mostly moved over to pkg/plugins/plugindef
* Remove dead func
* Fix up pfs, use sync.Once in plugindef
* Update to latest thema
* Chase Endec->Codec conversion in Thema
* Comments on slash header gen; use ToSlash
* Also generate JSON schema for plugindef
* Generate JSON Schema as well
* Fix slot loading from kindsys cue decls
* Remove unused vars
* skip generating plugin.schema.json for now
Co-authored-by: Marcus Efraimsson <marcus.efraimsson@gmail.com>
* Update thema to latest
* Deal with s/Library/*Runtime/
* Commit new, working results of codegen
* We like pointers now
* Always take runtime arg for NewBase()
* Sketchy handwavy pass at entity meta framework
* Little nibbles
* Update pkg/framework/coremodel/entityframework.cue
Co-authored-by: Artur Wierzbicki <wierzbicki.artur.94@gmail.com>
* Move file into new framework location
* Introduce loaders, Go code
* Complete rename to kind
* Flesh out framework, add svg/dashboard examples
* Cruft removal
* Remove generated kind go files from gitignore
* Refine maturity concept, add SlotKind
* Update embed and go deps
* Export PrefixWithGrafanaCUE
* Make the loader actually work, holy crap
* Many small tweaks to type.cue
* Add Apache 2 licensing exceptions for kinds
* Add new kinds dir, start of generator
* Roll back to earlier oapi-codegen
* Introduce new grafana-specific CUE loaders
* Introduce new tidy code generators framework
* Catch up kind framework with tinkering
* Add slices for the generators
* Add write/verify step to main generator
* Many renames
* Split up kind framework cue files
* Use kind.Decl within generated kinds
* Create kind.SomeDecl wrapper type to cache lineages
* Better names again
* Get one generated implemented, hopefully
* Copy dashboard schema into new kind.cue
* Small fixes to make the initial gen work
* Put svg kind in its new home
* Add generated Go dashboard type
* More renames and cleanups
* Add base kind registry and generator
* Stop blacklisting *_gen.go files
This is not the Go best practice, anyway. All we actually want to ignore
for enterprise is generated wire files.
* Change codegen output directories
pkg/kind -> pkg/kinds
pkg/registry/kindreg -> pkg/registry/corekind
* Rename pkg/framework/kind to pkg/kindsys
* Add core structured kind generator
* Add plural and machine names to kind spec
* Copy playlist over to kind system
* Consolidate kindsys files
* Add raw kind generator
* Update CODEOWNERS for kind framework
* Touch up comments a bit
* More docs tweaks
* Remove generated types to reduce noise for review
* Split each generator into its own file
* Rename Slot kind to Composable kind
* Add handwavy types for customkind loading
* Guard against init calls to framework loader
* First pass at doc on extending the kind system
* Improve attribute example in docs
* Fix wire imports
* Add basic TS types generator
* Fix composable kind category def
* No need for a separate file with generate directive
* Catch dashboard schema up
* Rename generator types to something saner and generic
* Make version configurable in ts/go generators
* Add CommonMeta to ease property access
* Add kindsys prop indicating whether lineage is group
* Put all kind categories back in a single file
* Finish with kindsys group props
* Refactor maturity progression per discussion
- Replace "committed" with "merged"
- All kindcats can use all maturity levels, at least for now
* Convert ts veneer index generator to modular system
* Move over to new jennywrites framework
* Strip down old coremodel generator
* Use public version of jennywrites
* Pull latest thema
* Commit generated Go types
* Add header injection postprocessor
* Move sdboyer/jennywrites to grafana/codejen
* Tweak header output
* Remove dashboard and playlist coremodels
* Fix up backend dashboards devenv test
* Fix TS import patterns to new gen filename
* Update internal imports, remove coremodel registry
* Fix compilation errors, wire generation
* Export and replace the prefix dropper
* More Go struct and field name changes
* Last name fixes, hopefully
* Fix lint errors
* Last lint error
Co-authored-by: Artur Wierzbicki <wierzbicki.artur.94@gmail.com>
* coremodels: Convert plugin-metadata schema to a coremodel
* Newer cuetsy; try quoting field name
* Add slot definitions
* Start sketching out pfs package
* Rerun codegen with fixes, new cuetsy
* Catch up dashboard with new cuetsy
* Update to go1.18
* Use new vmuxers in thema
* Add slot system in Go
* Draft finished implementation of pfs
* Collapse slot pkg into coremodel dir; add PluginInfo
* Add the mux type on top of kernel
* Refactor plugin generator for extensibility
* Change models.cue package, numerous debugs
* Bring new output to parity with old
* Remove old plugin generation logic
* Misc tweaking
* Reintroduce generation of shared schemas
* Drop back to go1.17
* Add globbing to tsconfig exclude
* Introduce pfs test on existing testdata
* Make most existing testdata tests pass with pfs
* coremodels: Convert plugin-metadata schema to a coremodel
* Newer cuetsy; try quoting field name
* Add APIType control concept, regen pluginmeta
* Use proper numeric types for schema fields
* Make pluginmeta schema follow Go type breakdown
* More decomposition into distinct types
* Add test case for no plugin.json file
* Fix missing ref to #Dependencies
* Remove generated TS for pluginmeta
* Update dependencies, rearrange go.mod
* Regenerate without Model prefix
* Use updated thema loader; this is now runnable
* Skip app plugin with weird include
* Make plugin tree extractor reusable
* Split out slot lineage load/validate logic
* Add myriad tests for new plugin validation failures
* Add test for zip fixtures
* One last run of codegen
* Proper delinting
* Ensure validation order is deterministic
* Let there actually be sorting
* Undo reliance on builtIn field (#54009)
* undo builtIn reliance
* fix tests
Co-authored-by: Will Browne <wbrowne@users.noreply.github.com>
* Remove crufty scuemata bits
Buhbye to: cue/ dir with old definitions, CI steps for checking unnecessary
things, and the original dashboard scuemata file.
* Remove grafana-cli cue subcommand
* Remove old testdata
* Don't swallow errors from codegen
* Small nits and tweaks to cuectx package
* WIP - refactor pluggen to use Thema
Also consolidate the embed.FS in the repo root.
* Finish halfway rename
* Convert all panel models.cue to thema
* Rewrite pluggen to use Thema
* Remove pkg/schema, and trim command
* Remove schemaloader service and usages
Will be replaced by coremodel-centric hydrate/dehydrate system Soon™.
* Remove schemaloader from wire
* Remove hangover field on histogram models.cue
* Fix lint errors, some vestiges of trim service
* Remove unused cuetsify cli command
* Copy over most of coremodel from intent-api branch
* Fix import paths
* Fix incorrect provider name
* Add root compgen file, fixup componentroot.yaml
* go mod tidy
* Remove compgen for now
* Add dashboard coremodel
* Remove datasource coremodel, for now
* Tweak comments on dashboard struct model
* devenv: add dashboard testing directly
* Fixup dashboard schema for openness, heatmap
* Update Thema to tip
* Fix wire/registry references
* Fix hclog version
