SDL parity tests #169
SDL parity tests #169
Conversation
WalkthroughAdd SDL parity infrastructure: embedded SDL input schema and validator, deterministic Go builders, a Go fixture generator, many new valid/invalid SDL fixtures, Go and TypeScript parity tests, TypeScript Go-format output support, output schemas/docs, Make targets, and a CI job to run SDL parity tests. Changes
Sequence Diagram(s)mermaid Estimated code review effort🎯 4 (Complex) | ⏱️ ~45 minutes
Possibly related issues
Possibly related PRs
Suggested reviewers
Poem
Pre-merge checks and finishing touches❌ Failed checks (1 warning)
✅ Passed checks (4 passed)
✨ Finishing touches
🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
vertex451
force-pushed
the
artem/sdl-parity-tests
branch
from
c464548 to
c669081
Compare
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 4
Caution
Some comments are outside the diff and can’t be posted inline due to platform limitations.
⚠️ Outside diff range comments (2)
go/sdl/go.mod (1)
7-15: gojsonschema dependency is used in production JWT schema validationThe
gojsonschemadependency is imported ingo/util/jwt/schema.gofor production JWT schema validation (initialized at package level), in addition to its use in SDL parity tests. The dependency is appropriate for these use cases, though be aware it initializes at package load time and uses reflection, which incurs startup overhead.ts/src/sdl/SDL/SDL.ts (1)
834-846: Essential change: Deterministic storage ordering for parity testing.Sorting storage keys alphabetically before mapping ensures consistent, reproducible manifest output. This is critical for cross-language parity validation.
♻️ Duplicate comments (1)
go/sdl/groupBuilder_v2.go (1)
100-121: Sorted iteration over storage volumes for deterministic manifestsUsing
storageNames+sort.Stringsbefore appendingStorageParamseliminates Go map iteration nondeterminism and keeps v2 output stable across runs and in parity tests. Implementation mirrors the v2.1 path and looks correct.
🧹 Nitpick comments (11)
go/sdl/groupBuilder_v2_1.go (1)
100-131: Deterministic storage params ordering is correctSorting
svc.Params.Storagekeys before buildingparams.Storageremoves map‑iteration nondeterminism and aligns with parity/determinism needs. Logic is sound and nil‑safe; any future refactor could share this helper with the v2 builder, but it’s not required now.make/test.mk (1)
51-63: Parity test workflow is solid; align TS path with existingTS_ROOTThe new
generate-sdl-fixturesandtest-sdl-paritytargets wire the Go and TS parity tests together cleanly, and the sequencing (fixtures → Go tests → TS tests) makes sense.For consistency with
test-ts/test-coverage-ts, consider using$(TS_ROOT)instead of a hard-codedtspath:- @cd ts && npm test -- --testPathPattern=parity.spec.ts + @cd $(TS_ROOT) && npm test -- --testPathPattern=parity.spec.tsThis avoids future drift if the TS root directory is ever renamed.
go/sdl/cmd/generate-sdl-fixtures/main.go (1)
13-13: Consider making the fixtures root path configurable.The hard-coded relative path
"../../../testdata/sdl"assumes the command is executed from a specific directory (likelygo/sdl/cmd/generate-sdl-fixtures). If the tool is run from a different location or installed elsewhere, it will fail.Consider accepting the fixtures root as a command-line argument with a default value:
+import ( + "flag" + // ... existing imports +) + func main() { - fixturesRoot := "../../../testdata/sdl" + fixturesRoot := flag.String("fixtures", "../../../testdata/sdl", "Path to SDL fixtures root directory") + flag.Parse() + versions := []string{"v2.0", "v2.1"} for _, version := range versions { - versionDir := filepath.Join(fixturesRoot, version) + versionDir := filepath.Join(*fixturesRoot, version)testdata/sdl/v2.0/persistent-storage/groups.json (1)
1-87: Add missing trailing newline.The file is missing a trailing newline at EOF. While some build systems are forgiving, most linters and Git conventions expect JSON files to end with a newline character for POSIX compliance.
Add a newline after the closing bracket at line 87.
] +testdata/sdl/v2.0/simple/groups.json (1)
1-67: Add missing trailing newline.Per the file structure, the JSON is valid but ends without a trailing newline. Add a newline after the closing bracket for POSIX compliance.
] +testdata/sdl/v2.0/storage-classes/groups.json (1)
1-87: Add missing trailing newline.The file is missing a trailing newline at EOF, violating POSIX conventions and typical linter expectations.
] +specs/sdl/sdl-input.schema.yaml (1)
10-24: Consider usingintegertype for counters instead ofnumberFields like
deployment.*.*.count(and similar count-like fields elsewhere) are conceptually integers. Using"type": "integer"instead of"type": "number"would prevent fractional values from passing schema validation while keeping current fixtures valid.Also applies to: 41-65
specs/sdl/groups.schema.yaml (1)
45-65: Tighten groups schema forcountandprice.denomTo better reflect the data model and avoid subtle mistakes:
resources[*].countshould be an integer (non-negative) rather than a genericnumber.pricecurrently requiresamountbut notdenom; in practice both are always present, so makingdenomrequired too would strengthen validation without affecting existing fixtures.go/sdl/parity_test.go (1)
17-46: Optional: reuse compiled JSON Schemas across fixtures
validateAgainstSchemareads and compiles the schema file for every manifest/groups validation. With many fixtures this adds unnecessary overhead. Consider caching compiled schemas keyed byschemaPath(e.g., package-levelmap[string]*gojsonschema.Schemaprotected bysync.Mutex/sync.Once) so each schema file is loaded and compiled only once per test run.Also applies to: 94-118
ts/src/sdl/SDL/parity.spec.ts (1)
81-101: Optional: cache compiled schemas to avoid recompiling on every fixture
validateAgainstSchemareads and compiles the YAML schema on every call. For many fixtures this adds overhead in test runs. You could cache the compiled Ajv validators byschemaPath(e.g., a simpleMap<string, ValidateFunction>) so each schema file is parsed/compiled only once per test run. This follows Ajv's recommended pattern of compiling once and reusing the validate function rather than recompiling on each invocation.specs/sdl/manifest.schema.yaml (1)
1-3: Add $schema and $id metadata for schema documentation and tooling.This is a new public schema file. It lacks standard JSON Schema metadata fields that aid documentation, versioning, and integration with schema validators.
Add
$schema(to declare the JSON Schema version) and$id(for schema identification and URL resolution) at the root level:+$schema: 'https://json-schema.org/draft/2020-12/schema' +$id: 'https://akash.network/schemas/sdl/manifest.schema.json' +title: Akash SDL Manifest Schema +description: Validates the structure of a compiled SDL manifest, including services, resources, and deployment configurations. items: additionalProperties: false properties:Also applies to: 328-328
📜 Review details
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Pro
⛔ Files ignored due to path filters (3)
go/sdl/go.sumis excluded by!**/*.sumts/package-lock.jsonis excluded by!**/package-lock.jsonts/src/sdl/SDL/__snapshots__/SDL.spec.ts.snapis excluded by!**/*.snap
📒 Files selected for processing (50)
.github/workflows/tests.yaml(1 hunks).gitignore(1 hunks)go/sdl/cmd/generate-sdl-fixtures/main.go(1 hunks)go/sdl/go.mod(2 hunks)go/sdl/groupBuilder_v2.go(1 hunks)go/sdl/groupBuilder_v2_1.go(1 hunks)go/sdl/parity_test.go(1 hunks)make/test.mk(1 hunks)specs/sdl/groups.schema.yaml(1 hunks)specs/sdl/manifest.schema.yaml(1 hunks)specs/sdl/sdl-input.schema.yaml(1 hunks)specs/sdl/validation-limitations.md(1 hunks)testdata/sdl/invalid/credentials-missing-host.yaml(1 hunks)testdata/sdl/invalid/endpoint-not-used.yaml(1 hunks)testdata/sdl/invalid/invalid-port.yaml(1 hunks)testdata/sdl/invalid/missing-deployment.yaml(1 hunks)testdata/sdl/invalid/missing-image.yaml(1 hunks)testdata/sdl/invalid/missing-version.yaml(1 hunks)testdata/sdl/invalid/negative-cpu.yaml(1 hunks)testdata/sdl/invalid/persistent-without-mount.yaml(1 hunks)testdata/sdl/v2.0/gpu-basic/groups.json(1 hunks)testdata/sdl/v2.0/gpu-basic/input.yaml(1 hunks)testdata/sdl/v2.0/gpu-basic/manifest.json(1 hunks)testdata/sdl/v2.0/http-options/groups.json(1 hunks)testdata/sdl/v2.0/http-options/input.yaml(1 hunks)testdata/sdl/v2.0/http-options/manifest.json(1 hunks)testdata/sdl/v2.0/ip-endpoint/groups.json(1 hunks)testdata/sdl/v2.0/ip-endpoint/input.yaml(1 hunks)testdata/sdl/v2.0/ip-endpoint/manifest.json(1 hunks)testdata/sdl/v2.0/multiple-services/groups.json(1 hunks)testdata/sdl/v2.0/multiple-services/input.yaml(1 hunks)testdata/sdl/v2.0/multiple-services/manifest.json(1 hunks)testdata/sdl/v2.0/persistent-storage/groups.json(1 hunks)testdata/sdl/v2.0/persistent-storage/input.yaml(1 hunks)testdata/sdl/v2.0/persistent-storage/manifest.json(1 hunks)testdata/sdl/v2.0/simple/groups.json(1 hunks)testdata/sdl/v2.0/simple/input.yaml(1 hunks)testdata/sdl/v2.0/simple/manifest.json(1 hunks)testdata/sdl/v2.0/storage-classes/groups.json(1 hunks)testdata/sdl/v2.0/storage-classes/input.yaml(1 hunks)testdata/sdl/v2.0/storage-classes/manifest.json(1 hunks)testdata/sdl/v2.1/credentials/groups.json(1 hunks)testdata/sdl/v2.1/credentials/input.yaml(1 hunks)testdata/sdl/v2.1/credentials/manifest.json(1 hunks)testdata/sdl/v2.1/ip-endpoint/groups.json(1 hunks)testdata/sdl/v2.1/ip-endpoint/input.yaml(1 hunks)testdata/sdl/v2.1/ip-endpoint/manifest.json(1 hunks)ts/package.json(1 hunks)ts/src/sdl/SDL/SDL.ts(9 hunks)ts/src/sdl/SDL/parity.spec.ts(1 hunks)
🧰 Additional context used
🧬 Code graph analysis (4)
ts/src/sdl/SDL/parity.spec.ts (1)
ts/src/sdl/SDL/SDL.ts (3)
validate(128-138)validate(246-261)SDL(88-1315)
go/sdl/cmd/generate-sdl-fixtures/main.go (1)
go/sdl/sdl.go (1)
ReadFile(83-89)
go/sdl/parity_test.go (2)
go/sdl/sdl.go (1)
ReadFile(83-89)go/testutil/deployment.go (1)
DeploymentGroups(54-62)
ts/src/sdl/SDL/SDL.ts (2)
ts/src/sdl/index.ts (1)
SdlValidationError(28-28)ts/src/sdl/types.ts (1)
v3Sdl(100-105)
🪛 GitHub Actions: lint
go/sdl/cmd/generate-sdl-fixtures/main.go
[error] 87-87: gofmt formatting issue detected. File is not properly formatted. Run 'gofmt -w' or 'go fmt ./...' to fix.
⏰ Context from checks skipped due to timeout of 90000ms. You can increase the timeout in your CodeRabbit configuration to a maximum of 15 minutes (900000ms). (1)
- GitHub Check: coverage
🔇 Additional comments (42)
.gitignore (1)
27-34: Well-organized additions supporting the testing infrastructure.The new entries appropriately capture test/build artifacts and coverage outputs introduced by the SDL parity testing changes. The naming conventions align with existing patterns, and the logical grouping with comments enhances clarity.
testdata/sdl/invalid/missing-image.yaml (1)
1-30: Invalid SDL fixture for missing image looks correctSDL structure is consistent with other v2.0 fixtures, and omitting
services.web.imageshould exercise the intended invalid case.testdata/sdl/invalid/invalid-port.yaml (1)
1-32: Invalid‑port SDL fixture is well‑formed for the negative testThe SDL is structurally valid but uses
port: 99999, which is out of the allowed port range, making it a good fixture to assert that validation rejects out‑of‑range ports.testdata/sdl/v2.0/ip-endpoint/manifest.json (1)
1-103: ip‑endpoint manifest fixture appears structurally consistentThe manifest structure (group → service → resources/endpoints → expose with
ipandendpointSequenceNumber) looks consistent with the v2.0/ip‑endpoint scenario and suitable for parity tests. Assuming this JSON was generated by the fixture tool, it should accurately represent the expected manifest; if it was edited manually, it’s worth re‑generating once to ensure it matches generator output and the companiongroups.json.testdata/sdl/invalid/negative-cpu.yaml (1)
1-32: Negative‑CPU SDL fixture correctly targets invalid resource valuesUsing
cpu.units: -100min an otherwise valid deployment is an appropriate way to exercise validation for disallowing negative CPU values.ts/package.json (1)
69-81: AJV is correctly placed as a devDependencyAJV (
^8.12.0) is used only in test files (ts/src/sdl/SDL/parity.spec.ts), confirming it should remain indevDependenciesand not be added todependencies. No runtime code imports or depends on AJV.testdata/sdl/invalid/credentials-missing-host.yaml (1)
1-35: LGTM! Invalid fixture correctly tests missing host credential validation.This fixture appropriately tests the credential validation pathway by intentionally omitting the required
hostfield while providingusernameandpassword. The validation logic ints/src/sdl/SDL/SDL.ts(lines 301-310) will correctly reject this SDL.testdata/sdl/v2.0/http-options/manifest.json (1)
1-73: LGTM! HTTP options manifest fixture is well-structured.The manifest correctly defines HTTP options with appropriate retry policies and timeout configurations for parity testing.
testdata/sdl/v2.0/storage-classes/manifest.json (1)
1-117: LGTM! Storage classes manifest correctly demonstrates multiple storage volumes.The fixture properly defines storage with varying attributes (persistent flags, storage classes) and the storage params section follows the alphabetically sorted order (cache, data, logs) as enforced by the updated SDL.ts logic.
testdata/sdl/invalid/endpoint-not-used.yaml (1)
1-35: LGTM! Invalid fixture correctly tests unused endpoint detection.This fixture appropriately validates that endpoints defined but never referenced in service exposures are detected as errors. The
validateEndpointsUtilitymethod in SDL.ts will correctly flag this scenario.testdata/sdl/v2.0/gpu-basic/manifest.json (1)
1-76: LGTM! GPU manifest fixture correctly represents GPU configuration.The fixture properly defines GPU resources with NVIDIA RTX 3080 specification, suitable for testing GPU-enabled deployment parity.
ts/src/sdl/SDL/SDL.ts (5)
2-2: LGTM! Named import improves clarity.Using the named
loadimport fromjs-yamlis more explicit and follows better practices than relying on default exports.
114-121: LGTM! Version field validation enhances robustness.Adding an explicit check for the required
versionfield with a clear error message improves SDL validation. The type cast tov3Sdlappropriately follows the validation.
285-299: LGTM! Service image and port validations strengthen SDL integrity.Both validation methods correctly enforce essential constraints:
- Empty image names are rejected
- Port values are constrained to valid range (1-65535)
The error messages are clear and helpful.
848-864: LGTM! Proper null handling and deterministic storage ordering.The method correctly handles
undefined/nullparams by returningnull, and ensures deterministic storage parameter ordering through alphabetical sorting.
899-916: LGTM! Conditional params assignment prevents unnecessary fields.Computing params via
v3ManifestServiceParamsand only assigning when not null prevents adding empty params fields to the manifest, resulting in cleaner output..github/workflows/tests.yaml (1)
57-76: LGTM! SDL parity CI job is properly configured.The new job correctly mirrors the setup steps from the coverage job and executes the SDL parity tests. The setup includes both Node.js and Go environments as required for cross-language parity validation.
testdata/sdl/v2.1/credentials/manifest.json (1)
1-75: LGTM! Credentials manifest correctly includes all required fields.This fixture appropriately tests the valid credentials scenario with all required fields (host, email, username, password) present, complementing the invalid credentials fixtures.
testdata/sdl/v2.0/gpu-basic/groups.json (1)
1-65: GPU basic group fixture looks structurally consistentThe JSON structure (requirements/resources/price) is consistent with other group fixtures, and the numeric values (CPU/memory/storage/GPU/price) look coherent for a minimal GPU case.
testdata/sdl/v2.0/multiple-services/groups.json (1)
1-130: Multiple-services group fixture is coherent and well-formedRequirements, resource variants, and pricing are internally consistent and align with the structure used in the other group fixtures in this PR.
testdata/sdl/invalid/missing-deployment.yaml (1)
1-27: Invalid “missing deployment” fixture matches its intentYAML is structurally valid, and omitting the
deploymentblock accurately represents the intended invalid case for tests.testdata/sdl/invalid/persistent-without-mount.yaml (1)
1-35: Persistent-without-mount invalid fixture is clearly definedThe compute/storage and deployment sections are well-formed, and the missing mount for a
persistent: truevolume cleanly expresses the invalid scenario the tests will exercise.testdata/sdl/v2.0/http-options/input.yaml (1)
1-50: HTTP-options SDL fixture is well-structured and expressiveThe
http_optionsblock and its parameters (timeouts, retries, next_cases) are clearly defined, and the compute/placement/deployment sections follow the established SDL v2.0 pattern, making this a good parity fixture.testdata/sdl/v2.0/ip-endpoint/input.yaml (1)
1-49: IP-endpoint SDL fixture cleanly exercises named IP exposureThe service/expose blocks correctly reference
ip: myendpoint, and the matchingendpoints.myendpointdefinition (kindip) plus standard compute/placement/deployment make this a solid parity case.testdata/sdl/v2.1/credentials/groups.json (1)
1-59: Credentials group fixture matches existing group schema usageThe requirements, resource definition, and pricing fields align with the other v2.0/v2.1 group fixtures and look appropriate for a credentials-focused parity test.
testdata/sdl/v2.0/simple/input.yaml (1)
1-47: LGTM! Well-structured SDL v2.0 test fixture.The fixture provides good coverage with multiple expose configurations (HTTP and UDP), placement attributes, and signedBy constraints with both anyOf and allOf.
testdata/sdl/v2.0/persistent-storage/manifest.json (1)
1-92: LGTM! Comprehensive manifest fixture for persistent storage.The manifest correctly captures persistent storage configuration with volume attributes (persistent: true, class: beta2) and mount parameters. The structure aligns with the expected manifest schema for v2.0.
testdata/sdl/invalid/missing-version.yaml (1)
1-31: LGTM! Valid negative test case for missing version.This fixture appropriately tests SDL validation by omitting the required
versionfield while maintaining otherwise valid structure. This ensures error handling for missing version information is properly exercised.testdata/sdl/v2.0/gpu-basic/input.yaml (1)
22-27: LGTM! Well-structured GPU resource configuration.The GPU attributes correctly specify the vendor (nvidia) and model (rtx3080) in nested format, which aligns with SDL v2.0 GPU specifications. This provides good test coverage for GPU-enabled deployments.
testdata/sdl/v2.0/http-options/groups.json (1)
1-59: LGTM! Valid groups fixture with proper structure.The groups data correctly defines requirements (signed_by, attributes), resources (compute specs, endpoints), and pricing. The 18-decimal precision for the amount field is standard for cryptocurrency denominations.
testdata/sdl/v2.1/ip-endpoint/input.yaml (1)
67-72: LGTM! Proper v2.1 IP endpoint configuration.The endpoints section correctly defines IP endpoints (endpoint1, endpoint2) which are referenced by the services' expose configurations. This provides good test coverage for the v2.1 IP endpoint feature.
testdata/sdl/v2.0/persistent-storage/input.yaml (1)
13-17: LGTM! Correct persistent storage configuration.The fixture properly links storage parameters (lines 14-17) with the compute profile's storage definition (lines 27-31). The storage volume is correctly marked as persistent with a storage class, and the mount point is properly specified.
Also applies to: 27-31
go/sdl/cmd/generate-sdl-fixtures/main.go (1)
60-60: File permissions0644are appropriate for generated test fixtures.The codebase shows a consistent pattern: test and configuration files use
0644permissions (owner read/write, group/others read-only), while sensitive files like key imports and governance data use0o600. Since the manifest and groups JSON files are test fixtures intended to be committed and read by the build system,0644is the correct choice.specs/sdl/validation-limitations.md (1)
1-12: All file references in the documentation are accurate. The schema filespecs/sdl/sdl-input.schema.yamland parser filesgo/sdl/v2.goandgo/sdl/v2_1.goall exist at the paths specified in the documentation.testdata/sdl/v2.1/ip-endpoint/manifest.json (1)
36-45: Verify if mixed endpoint structures between services are intentional.The manifest defines asymmetric endpoint configurations: the "api" service (lines 36-45) has consistent endpoints with both
kindandsequence_number, while the "web" service (lines 104-120) has mixed endpoints—some with onlysequence_number, others with both fields. Confirm whether this variation represents intentional test coverage for different endpoint types or if it's an inconsistency requiring alignment.Also applies to: 104-120
testdata/sdl/v2.1/credentials/input.yaml (1)
1-43: Schema reference and fixture consistency verified.The relative schema path
../../../../specs/sdl/sdl-input.schema.yamlis correct and the schema file exists. The input.yaml file is consistent with its corresponding groups.json and manifest.json fixtures: the service "private-app" maps to the "akash" group, and all resource specifications (CPU 500m, memory 1Gi, storage 5Gi, pricing 250 uakt) and credentials match across all three files.testdata/sdl/v2.1/ip-endpoint/groups.json (1)
1-114: ip-endpoint groups fixture shape looks consistent with schemaThe group structure (requirements.signed_by/attributes, resources with cpu/memory/storage/gpu/endpoints, count, and price) matches the new
groups.schema.yamlexpectations and provides good coverage of endpoint kinds/sequence numbers. I don’t see structural or obvious value issues here.testdata/sdl/v2.0/storage-classes/input.yaml (1)
1-59: Storage-classes SDL input aligns with the new SDL input schemaService, profiles (compute/placement), and deployment sections — including per-volume storage attributes and the
signedBy/pricing structure — all line up withsdl-input.schema.yaml. This fixture should be a solid basis for exercising storage-class handling.testdata/sdl/v2.0/multiple-services/manifest.json (1)
1-193: multiple-services manifest fixture is internally consistentService definitions (images, env, counts), inter-service expose wiring, resource IDs, and endpoint/
httpOptionsstructures are coherent and match what the v2.0 multiple-services SDL input describes. This looks like a good “golden” manifest for parity tests.testdata/sdl/v2.0/multiple-services/input.yaml (1)
1-93: multiple-services SDL input matches schema and manifest/groups fixturesThe service graph (web → api → db), compute/placement profiles, pricing, and deployment counts are coherent and line up with the generated manifest/groups fixtures. The
$schemaheader also correctly referencessdl-input.schema.yaml.specs/sdl/manifest.schema.yaml (2)
140-157: Clarify intentional structural difference between params.storage and resources.storage.Storage appears in two locations with differing structures:
- params.storage (lines 140-157): Array of objects with
{name, mount, readOnly}- resources.storage (lines 277-313): Array of objects with
{name, size, attributes}These serve different purposes, but the schema does not document this. Confirm this is intentional and document the distinction.
Verify that these storage structures represent different semantic concepts (e.g., params storage = mount points, resources storage = resource specifications) and not a schema inconsistency.
Also applies to: 277-313
67-92: httpOptions is universally required and consistently applied across all protocols in valid manifests.Verification of test fixtures confirms that all nine manifest files (v2.0 and v2.1) include httpOptions for every expose entry, regardless of protocol type (TCP, UDP, HTTP). No counter-examples exist where expose entries lack httpOptions. Non-HTTP protocols like TCP and UDP are fully supported and already accompanied by httpOptions in production fixtures, indicating this is an intentional schema design rather than an oversight. The concern that non-HTTP services would be rejected is unfounded—they are validated and expected to include httpOptions as part of the uniform expose specification.
vertex451
force-pushed
the
artem/sdl-parity-tests
branch
from
c669081 to
37393d8
Compare
vertex451
force-pushed
the
artem/sdl-parity-tests
branch
from
37393d8 to
204e70b
Compare
There was a problem hiding this comment.
Actionable comments posted: 0
♻️ Duplicate comments (1)
ts/src/sdl/SDL/parity.spec.ts (1)
163-169: Bug:finalValueis computed but immediately overwritten, making the clamping logic a no-op.This was flagged in a previous review. Line 167 assigns
finalValue(with the clampedendpointSequenceNumber), but line 169 immediately overwrites it with the originalvalue, negating allfinalValueadjustments.if (key === "endpointSequenceNumber" && typeof value === "number" && value > 1) { finalValue = 1; } normalized[normalizedKey] = finalValue; - - normalized[normalizedKey] = value;
🧹 Nitpick comments (4)
specs/sdl/sdl-input.schema.yaml (2)
211-215: Fix indentation inconsistency foraccept.items.The
itemsandtypeunderaccepthave extra indentation compared to sibling properties.accept: - items: - type: string - type: array + items: + type: string + type: array
287-291: Consider addingservicesandprofilesto required fields.Currently only
versionanddeploymentare required. While semantic validation is deferred to runtime, a valid SDL structurally requiresservicesandprofilessections. Adding them torequiredwould catch malformed SDLs earlier during IDE validation.required: - version - deployment + - services + - profilests/src/sdl/SDL/SDL.ts (1)
290-299: LGTM with minor suggestion.Port validation logic is correct. Consider a clearer error message:
- `Service "${serviceName}" has invalid port value. Port must be 0 < value <= 65535.`, + `Service "${serviceName}" has invalid port value. Port must be between 1 and 65535.`,ts/src/sdl/SDL/parity.spec.ts (1)
404-413: Consider testing invalid fixtures against both beta2 and beta3.Invalid fixtures are only tested with
beta3. If validation differs between versions (e.g., the invalid-port.yaml uses version "2.0"), testing only beta3 may miss version-specific edge cases.entries.forEach((filename) => { ["beta2", "beta3"].forEach((version) => { it(`should reject ${filename} (${version})`, () => { const input = fs.readFileSync( path.join(invalidDir, filename), "utf8", ); expect(() => SDL.fromString(input, version as "beta2" | "beta3")).toThrow(); }); }); });
📜 Review details
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Pro
⛔ Files ignored due to path filters (3)
go/sdl/go.sumis excluded by!**/*.sumts/package-lock.jsonis excluded by!**/package-lock.jsonts/src/sdl/SDL/__snapshots__/SDL.spec.ts.snapis excluded by!**/*.snap
📒 Files selected for processing (50)
.github/workflows/tests.yaml(1 hunks).gitignore(1 hunks)go/sdl/cmd/generate-sdl-fixtures/main.go(1 hunks)go/sdl/go.mod(2 hunks)go/sdl/groupBuilder_v2.go(1 hunks)go/sdl/groupBuilder_v2_1.go(1 hunks)go/sdl/parity_test.go(1 hunks)make/test.mk(1 hunks)specs/sdl/groups.schema.yaml(1 hunks)specs/sdl/manifest.schema.yaml(1 hunks)specs/sdl/sdl-input.schema.yaml(1 hunks)specs/sdl/validation-limitations.md(1 hunks)testdata/sdl/invalid/credentials-missing-host.yaml(1 hunks)testdata/sdl/invalid/endpoint-not-used.yaml(1 hunks)testdata/sdl/invalid/invalid-port.yaml(1 hunks)testdata/sdl/invalid/missing-deployment.yaml(1 hunks)testdata/sdl/invalid/missing-image.yaml(1 hunks)testdata/sdl/invalid/missing-version.yaml(1 hunks)testdata/sdl/invalid/negative-cpu.yaml(1 hunks)testdata/sdl/invalid/persistent-without-mount.yaml(1 hunks)testdata/sdl/v2.0/gpu-basic/groups.json(1 hunks)testdata/sdl/v2.0/gpu-basic/input.yaml(1 hunks)testdata/sdl/v2.0/gpu-basic/manifest.json(1 hunks)testdata/sdl/v2.0/http-options/groups.json(1 hunks)testdata/sdl/v2.0/http-options/input.yaml(1 hunks)testdata/sdl/v2.0/http-options/manifest.json(1 hunks)testdata/sdl/v2.0/ip-endpoint/groups.json(1 hunks)testdata/sdl/v2.0/ip-endpoint/input.yaml(1 hunks)testdata/sdl/v2.0/ip-endpoint/manifest.json(1 hunks)testdata/sdl/v2.0/multiple-services/groups.json(1 hunks)testdata/sdl/v2.0/multiple-services/input.yaml(1 hunks)testdata/sdl/v2.0/multiple-services/manifest.json(1 hunks)testdata/sdl/v2.0/persistent-storage/groups.json(1 hunks)testdata/sdl/v2.0/persistent-storage/input.yaml(1 hunks)testdata/sdl/v2.0/persistent-storage/manifest.json(1 hunks)testdata/sdl/v2.0/simple/groups.json(1 hunks)testdata/sdl/v2.0/simple/input.yaml(1 hunks)testdata/sdl/v2.0/simple/manifest.json(1 hunks)testdata/sdl/v2.0/storage-classes/groups.json(1 hunks)testdata/sdl/v2.0/storage-classes/input.yaml(1 hunks)testdata/sdl/v2.0/storage-classes/manifest.json(1 hunks)testdata/sdl/v2.1/credentials/groups.json(1 hunks)testdata/sdl/v2.1/credentials/input.yaml(1 hunks)testdata/sdl/v2.1/credentials/manifest.json(1 hunks)testdata/sdl/v2.1/ip-endpoint/groups.json(1 hunks)testdata/sdl/v2.1/ip-endpoint/input.yaml(1 hunks)testdata/sdl/v2.1/ip-endpoint/manifest.json(1 hunks)ts/package.json(1 hunks)ts/src/sdl/SDL/SDL.ts(9 hunks)ts/src/sdl/SDL/parity.spec.ts(1 hunks)
✅ Files skipped from review due to trivial changes (2)
- specs/sdl/validation-limitations.md
- testdata/sdl/invalid/negative-cpu.yaml
🚧 Files skipped from review as they are similar to previous changes (37)
- go/sdl/groupBuilder_v2.go
- .github/workflows/tests.yaml
- testdata/sdl/invalid/endpoint-not-used.yaml
- testdata/sdl/invalid/missing-version.yaml
- testdata/sdl/v2.0/persistent-storage/manifest.json
- go/sdl/parity_test.go
- make/test.mk
- testdata/sdl/v2.0/storage-classes/groups.json
- go/sdl/go.mod
- testdata/sdl/v2.0/multiple-services/groups.json
- testdata/sdl/v2.1/ip-endpoint/manifest.json
- testdata/sdl/v2.0/ip-endpoint/manifest.json
- testdata/sdl/v2.0/multiple-services/manifest.json
- testdata/sdl/v2.0/simple/groups.json
- testdata/sdl/v2.0/gpu-basic/groups.json
- testdata/sdl/invalid/missing-deployment.yaml
- testdata/sdl/v2.0/http-options/input.yaml
- testdata/sdl/v2.0/simple/manifest.json
- testdata/sdl/v2.0/http-options/groups.json
- testdata/sdl/invalid/missing-image.yaml
- testdata/sdl/v2.0/ip-endpoint/groups.json
- ts/package.json
- go/sdl/cmd/generate-sdl-fixtures/main.go
- testdata/sdl/v2.0/gpu-basic/manifest.json
- testdata/sdl/v2.0/storage-classes/manifest.json
- specs/sdl/manifest.schema.yaml
- testdata/sdl/v2.1/credentials/groups.json
- testdata/sdl/v2.0/simple/input.yaml
- testdata/sdl/invalid/credentials-missing-host.yaml
- testdata/sdl/v2.1/ip-endpoint/input.yaml
- testdata/sdl/v2.0/gpu-basic/input.yaml
- testdata/sdl/v2.1/ip-endpoint/groups.json
- testdata/sdl/invalid/persistent-without-mount.yaml
- testdata/sdl/v2.1/credentials/manifest.json
- specs/sdl/groups.schema.yaml
- .gitignore
- testdata/sdl/v2.0/http-options/manifest.json
🧰 Additional context used
🧬 Code graph analysis (1)
ts/src/sdl/SDL/SDL.ts (2)
ts/src/sdl/index.ts (1)
SdlValidationError(28-28)ts/src/sdl/types.ts (1)
v3Sdl(100-105)
⏰ Context from checks skipped due to timeout of 90000ms. You can increase the timeout in your CodeRabbit configuration to a maximum of 15 minutes (900000ms). (3)
- GitHub Check: coverage
- GitHub Check: test
- GitHub Check: go
🔇 Additional comments (14)
testdata/sdl/v2.0/persistent-storage/input.yaml (1)
3-48: SDL configuration is well-formed and consistent.The YAML SDL v2.0 configuration correctly defines:
- Persistent storage setup with Postgres database service
- Compute profile with appropriate resource specifications (500m CPU, 1Gi RAM, 50Gi storage)
- Placement constraints (region us-west, signed_by policy)
- Pricing and deployment mappings
Cross-referenced against the companion
groups.jsonfixture, resource specifications are consistent.testdata/sdl/v2.0/persistent-storage/groups.json (1)
1-70: JSON fixture data is accurate and consistent with input.yaml.Byte conversions are correct:
- Memory: 1,073,741,824 bytes = 1 GiB ✓
- Storage: 53,687,091,200 bytes = 50 GiB ✓
All resource specifications (CPU 500 units, storage class/attributes, pricing, count) align with the corresponding input.yaml file. The data structure appears well-suited for parity test validation.
testdata/sdl/v2.0/storage-classes/input.yaml (2)
1-1: The schema reference path is correct and resolves properly. The filespecs/sdl/sdl-input.schema.yamlexists at the expected location, and the relative path../../../../specs/sdl/sdl-input.schema.yamlcorrectly references the SDL input schema from the file's location attestdata/sdl/v2.0/storage-classes/input.yaml.
2-58: LGTM! Comprehensive storage-classes fixture.The SDL v2.0 manifest provides a well-structured test fixture that exercises the storage-classes feature with parametric storage definitions, mixed storage attributes (persistent flags and class names), and proper deployment configuration. The mixed attribute patterns (cache/logs minimal vs. data with both persistent and class) are intentional for testing edge cases. The fixture validates successfully against the sdl-input.schema.yaml schema.
go/sdl/groupBuilder_v2_1.go (1)
115-121: LGTM! Deterministic ordering ensures parity testing reliability.The implementation correctly addresses non-deterministic Go map iteration by collecting storage names, sorting them, and iterating in sorted order. This ensures consistent output across runs, which is essential for the parity tests between Go and TypeScript implementations.
testdata/sdl/v2.0/ip-endpoint/input.yaml (1)
1-49: LGTM!Well-structured SDL v2.0 test fixture for IP endpoint functionality. The configuration correctly demonstrates:
- Dual port exposure (80 and 443) with IP endpoint references
- Proper endpoint declaration matching the IP references in the service expose section
- Valid compute profile and placement configuration
testdata/sdl/v2.0/multiple-services/input.yaml (1)
1-93: LGTM!Comprehensive multi-service SDL v2.0 test fixture demonstrating:
- Service dependency chains (web → api → db)
- Multiple compute profiles with varying resource allocations
- Both
anyOfandallOfsignedBy constraints- Inter-service expose configurations
The hardcoded credentials (lines 29, 18) are appropriate for test fixtures.
testdata/sdl/invalid/invalid-port.yaml (1)
1-32: LGTM!Valid negative test fixture with intentionally invalid port value (99999 > 65535). This tests the port validation logic in
validateServicePortswhich enforcesport > 0 && port <= 65535.testdata/sdl/v2.1/credentials/input.yaml (1)
1-43: LGTM!Well-structured SDL v2.1 test fixture for private registry credentials. Correctly demonstrates the credentials block structure (host, email, username, password) for private container registry authentication.
ts/src/sdl/SDL/SDL.ts (3)
113-122: LGTM!Good addition of version field validation in
fromString. This catches malformed SDLs early and provides a clear error message.
834-864: LGTM!Good addition of deterministic storage key ordering via
.sort()in bothv2ManifestServiceParamsandv3ManifestServiceParams. This ensures consistent output for parity testing between Go and TypeScript implementations.
899-917: LGTM!Clean refactor of params handling. The conditional assignment (
if (params !== null)) prevents null params from polluting the manifest output.ts/src/sdl/SDL/parity.spec.ts (2)
24-79: LGTM!
makeSchemaLenientappropriately relaxes JSON Schema constraints for flexible parity testing. The recursive handling of nested structures (properties,items,anyOf,oneOf,allOf) is thorough.
303-315: Verify: v2.0 fixtures usev3Manifestmethod.The tests call
sdl.v3Manifest(true)for v2.0/beta2 fixtures instead ofsdl.v2Manifest(). The normalization handles GPU filtering for beta2, but verify this is the intended parity comparison approach.
There was a problem hiding this comment.
Actionable comments posted: 0
🧹 Nitpick comments (3)
specs/sdl/manifest.schema.yaml (1)
1-6: Consider adding standard JSON Schema root declarations.The schema starts directly with
items:without a root-leveltype: arraydeclaration or$schemaidentifier. While this may work with some validators, adding these improves compatibility and clarity:+$schema: "https://json-schema.org/draft/2020-12/schema" +type: array items: additionalProperties: false properties: name: type: stringts/src/sdl/SDL/parity.spec.ts (2)
68-76: Redundant processing for array items.Lines 68-76 re-check
additionalPropertiesandrequiredonitemsSchema, butmakeSchemaLenientwas already called onlenient.itemsat line 53, which would have handled these. This block is defensive but unnecessary.- if (lenient.type === "array" && lenient.items) { - const itemsSchema = lenient.items as Record<string, unknown>; - if (itemsSchema.additionalProperties === false) { - itemsSchema.additionalProperties = true; - } - if (itemsSchema.required && Array.isArray(itemsSchema.required)) { - delete itemsSchema.required; - } - } - return lenient;
81-101: Schema is recompiled on every validation call.
ajv.compile()is called each timevalidateAgainstSchemais invoked. Since the same schemas are validated multiple times across fixtures, consider caching compiled validators.+const schemaCache = new Map<string, ReturnType<typeof ajv.compile>>(); + function validateAgainstSchema(actual: unknown, schemaPath: string): { valid: boolean; errors: string[] } { if (!fs.existsSync(schemaPath)) { return { valid: true, errors: [] }; } + let validate = schemaCache.get(schemaPath); + if (!validate) { const schemaContent = fs.readFileSync(schemaPath, "utf8"); const schema = load(schemaContent) as JSONSchema; const lenientSchema = makeSchemaLenient(schema); - const validate = ajv.compile(lenientSchema as Record<string, unknown>); + validate = ajv.compile(lenientSchema as Record<string, unknown>); + schemaCache.set(schemaPath, validate); + } const valid = validate(actual);
📜 Review details
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Pro
📒 Files selected for processing (2)
specs/sdl/manifest.schema.yaml(1 hunks)ts/src/sdl/SDL/parity.spec.ts(1 hunks)
⏰ Context from checks skipped due to timeout of 90000ms. You can increase the timeout in your CodeRabbit configuration to a maximum of 15 minutes (900000ms). (5)
- GitHub Check: proto
- GitHub Check: go
- GitHub Check: sdl-parity
- GitHub Check: test
- GitHub Check: coverage
🔇 Additional comments (7)
specs/sdl/manifest.schema.yaml (2)
295-314: Good use of$defsfor reusable credentials schema.The credentials schema is now properly defined once in
$defsand referenced via$refat lines 26 and 105, eliminating the previous duplication.
33-95: LGTM!The
exposeschema is well-structured with proper nesting forhttpOptions, required fields, and nullable patterns usinganyOfwithtype: 'null'.ts/src/sdl/SDL/parity.spec.ts (5)
103-171: LGTM!The
normalizeKeysfunction correctly handles key transformations and value normalization. The previous issue withfinalValuebeing overwritten has been addressed—now onlyfinalValueis assigned tonormalized[normalizedKey].
179-188: Numeric string normalization may have edge cases.The logic at lines 181-186 only normalizes strings containing
.and checks for specific trailing zero patterns. Integer strings (e.g.,"100") won't be converted, and strings like"1.0"might not match the.000000000000000000pattern.If this is intentional to handle specific Go/TS output differences, consider adding a brief comment explaining the expected scenarios.
251-280: LGTM!The
loadFixturesfunction correctly discovers and validates fixture directories, ensuring only complete fixtures (with input.yaml, manifest.json, and groups.json) are returned.
402-411: Invalid fixtures only tested against beta3.The invalid fixtures test uses hardcoded
"beta3"version. If v2.0/beta2 has different validation rules, consider also testing invalid fixtures against beta2, or documenting that invalid fixtures are version-specific.
295-329: LGTM!The v2.0 test suite correctly loads fixtures, normalizes both actual and expected outputs, compares them, and validates against schemas. The error logging on validation failure aids debugging.
…sdl-parity-tests Signed-off-by: Artem Shcherbatiuk <artem@akash.network> Co-authored-by: Artem Shcherbatiuk <vertex451@gmail.com>
There was a problem hiding this comment.
Actionable comments posted: 3
🧹 Nitpick comments (7)
specs/sdl/sdl-input.schema.yaml (1)
386-388: Consider addingservicesto required fields.The
requiredarray includesversionanddeployment, butservicesis not required. Since deployments reference services by name, an SDL without services would likely fail at runtime. Addingserviceshere would catch this earlier during schema validation.required: - version - deployment + - servicests/src/sdl/SDL/SDL.ts (1)
1305-1313: Consider extracting duplicate amount formatting logic.This amount formatting logic is duplicated in both
v2Groups(lines 1305-1312) andv3Groups(lines 1210-1217). Consider extracting to a helper method:private formatAmount(amount: number | string | undefined): string { return typeof amount === "number" ? amount.toFixed(18) : (amount?.toString() || "0"); }specs/sdl/groups-output.schema.yaml (1)
1-5: Consider adding$schemadeclaration and metadata.Unlike
sdl-input.schema.yaml, this schema lacks the$schemadeclaration and descriptive metadata. Adding these would improve documentation and IDE support:$schema: http://json-schema.org/draft-07/schema# title: Akash SDL Groups Output Schema description: Schema for validating SDL groups output structure. items: # ... rest of schemaspecs/sdl/manifest-output.schema.yaml (1)
1-6: Consider adding$schemadeclaration and metadata.Similar to the groups output schema, adding metadata would improve documentation:
$schema: http://json-schema.org/draft-07/schema# title: Akash SDL Manifest Output Schema description: Schema for validating SDL manifest output structure. items: # ... rest of schemats/src/sdl/SDL/parity.spec.ts (3)
22-35: Consider caching compiled schema validators.
ajv.compile(schema)is called for each validation, recompiling the same schemas multiple times (once per fixture × 3 schemas). AJV caches compiled validators internally by schema identity, but since the schema is re-read and re-parsed each time, caching is bypassed.Consider preloading and compiling schemas once:
const schemaCache = new Map<string, AjvModule.ValidateFunction>(); function validateAgainstSchema(name: string, data: unknown, schemaPath: string): void { let validate = schemaCache.get(schemaPath); if (!validate) { const schemaContent = fs.readFileSync(schemaPath, "utf8"); const schema = load(schemaContent); validate = ajv.compile(schema as Record<string, unknown>); schemaCache.set(schemaPath, validate); } // ... rest of validation }
71-122: Consider parameterizing v2.0 and v2.1 test blocks.The two describe blocks are nearly identical. You could reduce duplication with parameterized tests:
describe.each([ { version: "v2.0", sdlVersion: "beta2" as const }, { version: "v2.1", sdlVersion: "beta3" as const }, ])("$version", ({ version, sdlVersion }) => { const fixtures = loadFixtures(version); fixtures.forEach((fixture) => { it(fixture.name, () => { // ... test body using sdlVersion }); }); });However, the current explicit form is also fine for readability.
124-142: Early return silently skips invalid SDL tests.If
testdata/sdl/invaliddoesn't exist, the describe block executes but defines no tests. This might hide a missing test fixture issue. Consider either:
- Making the directory required (throw an error)
- Using
describe.skipwith a message- At minimum, logging a warning
describe("invalid SDLs rejected", () => { const invalidDir = path.join(fixturesRoot, "invalid"); if (!fs.existsSync(invalidDir)) { it.skip("invalid fixtures directory not found", () => {}); return; } // ... });
📜 Review details
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Pro
📒 Files selected for processing (7)
go/sdl/parity_test.go(1 hunks)specs/sdl/README.md(1 hunks)specs/sdl/groups-output.schema.yaml(1 hunks)specs/sdl/manifest-output.schema.yaml(1 hunks)specs/sdl/sdl-input.schema.yaml(1 hunks)ts/src/sdl/SDL/SDL.ts(19 hunks)ts/src/sdl/SDL/parity.spec.ts(1 hunks)
✅ Files skipped from review due to trivial changes (1)
- specs/sdl/README.md
🚧 Files skipped from review as they are similar to previous changes (1)
- go/sdl/parity_test.go
🧰 Additional context used
🧬 Code graph analysis (1)
ts/src/sdl/SDL/parity.spec.ts (1)
ts/src/sdl/SDL/SDL.ts (3)
validate(129-139)validate(318-333)SDL(89-1428)
🪛 GitHub Check: ts
ts/src/sdl/SDL/SDL.ts
[failure] 1212-1212:
Trailing spaces not allowed
[failure] 1307-1307:
Trailing spaces not allowed
⏰ Context from checks skipped due to timeout of 90000ms. You can increase the timeout in your CodeRabbit configuration to a maximum of 15 minutes (900000ms). (5)
- GitHub Check: test
- GitHub Check: coverage
- GitHub Check: sdl-parity
- GitHub Check: go
- GitHub Check: proto
🔇 Additional comments (8)
specs/sdl/sdl-input.schema.yaml (2)
117-128: Verify the storage conditional validation logic.The second
if/thenblock appears to enforce that whenpersistentisfalse, theclassproperty must NOT exist. This seems inverted from the typical constraint (usually you'd want to restrict whenpersistent: truewithclass: ram). The first block (lines 105-116) already handles theclass: ramcase.Could you confirm this is the intended behavior? If it's meant to prevent non-persistent storage from having a class attribute, consider adding a comment explaining this constraint.
182-184: Verify denom pattern case sensitivity.The pattern
^(uakt|ibc/.*)$uses lowercaseibc. In the Cosmos ecosystem, IBC denoms are typically uppercase (e.g.,ibc/27394FB092D2ECCD...). If both cases are valid, consider making the pattern case-insensitive:- pattern: '^(uakt|ibc/.*)$' + pattern: '^(uakt|[iI][bB][cC]/.*)$'Or simply use uppercase if that's the canonical form.
ts/src/sdl/SDL/SDL.ts (3)
247-316: LGTM -convertToGoFormathandles key transformations comprehensively.The method handles Go-specific key naming conventions (snake_case, special cases like
HTTPOptions→httpOptions), empty array to null conversion forall_of/any_of, and properly skips undefined values. The recursion for arrays and nested objects is correct.
906-918: LGTM - Sorted storage keys ensure deterministic output.Sorting storage keys before iteration ensures consistent ordering between Go and TypeScript implementations, which is essential for parity testing.
1170-1179: LGTM - GPU default handling matches Go behavior.Always including GPU in groups output (defaulting to units
"0"when absent) ensures parity with Go's output format.specs/sdl/groups-output.schema.yaml (1)
49-50: LGTM - Schema structure is comprehensive.The groups output schema properly validates the nested structure including resources with CPU, GPU, memory, storage, and endpoints. The strict
additionalProperties: falseensures no unexpected fields pass validation.specs/sdl/manifest-output.schema.yaml (1)
295-314: LGTM - Good use of$defsfor reusable credentials schema.The credentials definition is properly extracted and referenced in two places, reducing duplication and ensuring consistency.
ts/src/sdl/SDL/parity.spec.ts (1)
37-69: LGTM - Fixture loading with helpful error messages.The
loadFixturesfunction provides clear error messages guiding developers to runmake generate-sdl-fixtureswhen fixtures are missing. This is good developer experience.
There was a problem hiding this comment.
Actionable comments posted: 1
🧹 Nitpick comments (3)
ts/src/sdl/SDL/SDL.ts (3)
247-316: Consider a key transformation map for maintainability.The
convertToGoFormatmethod handles key transformations with a long chain of conditionals. While functional, this approach can be error-prone as the number of special cases grows. Lines 279 and 281 mapendpointSequenceNumberandexternalPortto themselves—verify these keys don't require snake_case transformation for Go compatibility.Consider replacing the conditional chain with a transformation map:
private readonly GO_KEY_MAP: Record<string, string> = { sequenceNumber: "sequence_number", signedBy: "signed_by", allOf: "all_of", anyOf: "any_of", quantity: "size", HTTPOptions: "httpOptions", IP: "ip" // add other mappings }; private convertToGoFormat(obj: unknown): unknown { // ... existing null/array/primitive checks ... for (const key in objRecord) { // ... existing undefined checks ... let convertedKey = this.GO_KEY_MAP[key] || key; // Then apply fallback transformations (camelCase, snake_case) if not in map // ... } }This makes the transformations explicit and easier to audit.
1028-1061: Update comment to reflect deduplication behavior.The comment at line 1029 states "Collect all IPs (including duplicates)" but line 1050 explicitly deduplicates via
new Set(endpointNames). The implementation correctly deduplicates, but the comment is misleading.Update the comment to:
- // Collect all IPs (including duplicates) to match Go behavior + // Collect all IPs from service endpoints, then deduplicate and sort for stable sequence assignment
1210-1218: Consider extracting amount formatting to reduce duplication.The amount formatting logic (lines 1210-1214) is duplicated in
v2Groups(lines 1305-1309). Extracting this to a helper method would improve maintainability and ensure consistent precision handling.Add a helper method:
private formatAmountForGo(amountValue: number | string | undefined): string { return typeof amountValue === "number" ? amountValue.toFixed(18) : (amountValue?.toString() || "0"); }Then replace both occurrences with:
const amountStr = this.formatAmountForGo(pricing.amount as number | string | undefined);
📜 Review details
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Pro
📒 Files selected for processing (1)
ts/src/sdl/SDL/SDL.ts(19 hunks)
🧰 Additional context used
🧬 Code graph analysis (1)
ts/src/sdl/SDL/SDL.ts (2)
ts/src/sdl/index.ts (1)
SdlValidationError(28-28)ts/src/sdl/types.ts (3)
v3Manifest(4-4)v3ComputeResources(218-224)v3DeploymentGroup(270-285)
⏰ Context from checks skipped due to timeout of 90000ms. You can increase the timeout in your CodeRabbit configuration to a maximum of 15 minutes (900000ms). (5)
- GitHub Check: sdl-parity
- GitHub Check: coverage
- GitHub Check: test
- GitHub Check: proto
- GitHub Check: go
🔇 Additional comments (15)
ts/src/sdl/SDL/SDL.ts (15)
2-2: LGTM: Updated to named import from js-yaml.The change from
YAML.load(...)toload(...)aligns with modern js-yaml API usage.
57-57: LGTM: Clean type definition for output format switching.The
OutputFormattype provides clear intent for toggling between TypeScript and Go-compatible output formats.
129-139: LGTM: Updated YAML parsing to use load().The change aligns with the updated import and maintains existing validation logic.
318-333: LGTM: Extracted validation methods improve modularity.The new calls to
validateServiceImageandvalidateServicePortsmake the validation logic more organized and testable.
357-361: LGTM: Runtime validation ensures non-empty image names.This validation complements the JSON schema checks used in tests and provides runtime protection against empty or whitespace-only image names.
362-371: LGTM: Port range validation is correct.The validation ensures ports fall within the valid TCP/UDP range (1-65535), providing runtime protection consistent with the JSON schema constraints.
906-918: LGTM: Storage key sorting ensures deterministic output.Sorting storage keys before iteration ensures consistent, reproducible manifest output that matches Go implementation behavior.
920-936: LGTM: Proper null handling and deterministic storage ordering.The method correctly returns
nullfor undefined/null params and sorts storage keys for deterministic output. The error check at line 928 is appropriate for catching malformed configurations.
961-989: LGTM: Conditional params assignment matches Go's optional field handling.Only assigning
paramswhen non-null ensures the output matches Go's behavior of omitting optional fields that are null/undefined.
998-1010: LGTM: Format parameter enables Go-compatible output.The method cleanly supports both TypeScript and Go output formats by conditionally applying
convertToGoFormatand propagating the format parameter to dependent methods.
1080-1096: LGTM: Dynamic key naming supports format compatibility.Switching between
"quantity"and"size"based on theasStringparameter ensures the output matches expected formats for both TypeScript and manifest representations.
1099-1110: LGTM: Consistent key switching for storage units.The dynamic key naming matches the pattern used in
resourceUnitMemoryand ensures format compatibility.
1170-1179: LGTM: GPU always included to match Go behavior.The code correctly defaults to GPU with
units: "0"when not present, ensuring output parity with the Go implementation. The comment clearly explains this design decision.
1194-1287: LGTM: Format parameter and precision handling ensure Go parity.The method correctly handles the format parameter and applies 18-decimal precision to amounts (lines 1212-1214), which is critical for matching Go's numeric output. The
convertToGoFormatconversion is appropriately applied at the end.
1190-1192: Thegroups()method passes theformatparameter only tov3Groups(), notv2Groups()(line 1191).This is a factual difference in how the two code paths handle the format parameter. If v2 is deprecated and no longer in active use, this design is appropriate. However, if v2 deployments are still supported or tested for parity with v3, the v2Groups() method should also accept and use the format parameter to maintain feature parity.
There was a problem hiding this comment.
Actionable comments posted: 4
🧹 Nitpick comments (2)
go/sdl/schema_validator.go (1)
114-120: Consider usingerrors.Joinor structured error for validation failures.The current approach concatenates error strings into a slice and formats with
%v. This works but loses structure for downstream error handling.+import "strings" + if !result.Valid() { - var errors []string + var errMsgs []string for _, desc := range result.Errors() { - errors = append(errors, desc.String()) + errMsgs = append(errMsgs, desc.String()) } - return fmt.Errorf("schema validation failed: %v", errors) + return fmt.Errorf("schema validation failed: %s", strings.Join(errMsgs, "; ")) }ts/src/sdl/SDL/SDL.ts (1)
276-281: Redundant key mappings preserve original values unnecessarily.Lines 278-281 map keys to themselves (
sequenceNumber→sequenceNumber,externalPort→externalPort), which has no effect. These could be removed unless they're placeholders for future mappings.if (key === "sequenceNumber") { convertedKey = "sequence_number"; - } else if (key === "endpointSequenceNumber") { - convertedKey = "endpointSequenceNumber"; - } else if (key === "externalPort") { - convertedKey = "externalPort"; } else if (key === "signedBy") {If these are intentionally preserved to document "no change needed", consider adding a comment.
📜 Review details
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Pro
📒 Files selected for processing (5)
go/sdl/cmd/generate-sdl-fixtures/main.go(1 hunks)go/sdl/parity_test.go(1 hunks)go/sdl/schema_validator.go(1 hunks)go/sdl/sdl.go(2 hunks)ts/src/sdl/SDL/SDL.ts(19 hunks)
🚧 Files skipped from review as they are similar to previous changes (2)
- go/sdl/cmd/generate-sdl-fixtures/main.go
- go/sdl/parity_test.go
🧰 Additional context used
🧬 Code graph analysis (2)
ts/src/sdl/SDL/SDL.ts (1)
ts/src/sdl/types.ts (4)
v3Sdl(100-105)v3Manifest(4-4)v3ComputeResources(218-224)v3DeploymentGroup(270-285)
go/sdl/schema_validator.go (1)
go/sdl/sdl.go (1)
ReadFile(83-89)
⏰ Context from checks skipped due to timeout of 90000ms. You can increase the timeout in your CodeRabbit configuration to a maximum of 15 minutes (900000ms). (5)
- GitHub Check: sdl-parity
- GitHub Check: test
- GitHub Check: coverage
- GitHub Check: go
- GitHub Check: proto
🔇 Additional comments (6)
go/sdl/sdl.go (1)
92-136: Schema validation integration is consistent and well-structured.The pattern of capturing
schemaErrupfront and callingcheckSchemaValidationResultat each exit point (both error and success) provides comprehensive coverage for comparing schema validation results against Go's runtime validation. All error paths are covered.One consideration: since
schemaErris captured once at the start butcheckSchemaValidationResultis called at different validation stages, any logged mismatch may not distinguish which validation stage disagreed with the schema. This is likely acceptable for parity testing purposes, but worth noting if debugging becomes difficult.ts/src/sdl/SDL/SDL.ts (5)
362-371: Port validation logic is correct.The validation correctly ensures ports are in the valid range (1-65535). The error message accurately describes the constraint.
1170-1179: GPU default handling ensures Go parity.The logic to always include GPU with a default
units: { val: "0" }when not present ensures consistent output structure matching Go's behavior. The comment is helpful.
906-936: Sorted storage keys ensure deterministic manifest output.The changes to sort storage keys (
Object.keys(params?.storage ?? {}).sort()) in both v2 and v3 manifest service params functions ensure consistent ordering across runs, which is essential for parity testing.
1053-1058: Verify if duplicate IP overwrite behavior matches Go implementation.The code collects all IPs including duplicates (line 39) but assigns sequence numbers via object key assignment (line 1057), which means duplicate IPs will have only their last sequence number retained. The comments indicate this matches Go behavior, but this should be confirmed against the Go implementation to ensure the overwrite semantics are intentional rather than a bug.
1210-1217:toFixed(18)with very large numbers can produce scientific notation, but pricing amounts in fixtures are well within safe range.JavaScript's
toFixed(18)does produce scientific notation for values ≥ 1e21 (e.g.,(1e21).toFixed(18)returns"1e+21"), but all pricing amounts in SDL fixtures are substantially smaller (ranging from 25 to 1,000,000), making this a non-issue in practice. The current implementation is safe for the actual use cases in this codebase.
There was a problem hiding this comment.
Actionable comments posted: 0
🧹 Nitpick comments (2)
go/sdl/cmd/generate-sdl-fixtures/main.go (2)
13-15: Make fixtures root and version discovery less cwd‑dependent
fixturesRoot := "../../testdata/sdl"and the hard‑codedversions := []string{"v2.0", "v2.1"}assume a specific working directory and require code edits when new versions are added.Consider:
- Deriving
fixturesRootrelative to the executable (or accepting it as a flag/env), sogo run ./go/sdl/cmd/generate-sdl-fixturesfrom repo root still works reliably.- Optionally discovering versions by listing subdirectories under
fixturesRootinstead of keeping a manual slice, unless you explicitly want to gate which versions are generated.
49-81: Permissions and overall generation loop look good, with a minor mode nitThe manifest/groups generation flow and error handling look solid and idempotent: each failure is logged and the loop continues with other fixtures.
One minor nit:
os.WriteFile(..., 0600)will create new JSON fixtures as user‑read/write only. For repo fixtures that are meant to be shared,0644is more typical; consider switching to that unless you intentionally want stricter permissions.
📜 Review details
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Pro
📒 Files selected for processing (4)
go/provider/client/client.go(1 hunks)go/sdl/cmd/generate-sdl-fixtures/main.go(1 hunks)go/sdl/go.mod(2 hunks)go/util/jwt/jwt_test.go(1 hunks)
🧰 Additional context used
🧬 Code graph analysis (1)
go/sdl/cmd/generate-sdl-fixtures/main.go (2)
go/sdl/sdl.go (1)
ReadFile(83-89)go/testutil/deployment.go (1)
DeploymentGroups(54-62)
⏰ Context from checks skipped due to timeout of 90000ms. You can increase the timeout in your CodeRabbit configuration to a maximum of 15 minutes (900000ms). (5)
- GitHub Check: proto
- GitHub Check: test
- GitHub Check: coverage
- GitHub Check: sdl-parity
- GitHub Check: go
🔇 Additional comments (3)
go/provider/client/client.go (1)
99-103: LGTM!Clean stylistic improvement. Omitting names for unused receiver and parameters is idiomatic Go and equivalent to using
_placeholders.go/util/jwt/jwt_test.go (1)
189-189: LGTM! Clean simplification.The variadic append is more idiomatic and concise than the previous nested loop approach while preserving identical behavior.
go/sdl/go.mod (1)
8-8: Consider the maintenance posture of gojsonschema and its dependencies.The
gojsonschema v1.2.0dependency is Go 1.25-compatible. However, the library hasn't had a release since October 2019 (5+ years), and its transitive dependencies (gojsonpointerandgojsonreference) were last maintained in January 2018 and are now effectively abandoned. While no CVEs are published, the lack of maintenance creates a future security posture risk. If production use requires more actively-maintained alternatives, consider evaluating other JSON schema validators likegoogle/jsonschema-goorsanthosh-tekuri/jsonschema.Also applies to: 13-13, 166-167
There was a problem hiding this comment.
Actionable comments posted: 1
📜 Review details
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Pro
⛔ Files ignored due to path filters (1)
go/cli/go.sumis excluded by!**/*.sum
📒 Files selected for processing (1)
go/cli/go.mod(1 hunks)
⏰ Context from checks skipped due to timeout of 90000ms. You can increase the timeout in your CodeRabbit configuration to a maximum of 15 minutes (900000ms). (5)
- GitHub Check: go
- GitHub Check: proto
- GitHub Check: sdl-parity
- GitHub Check: coverage
- GitHub Check: test
There was a problem hiding this comment.
Actionable comments posted: 1
🧹 Nitpick comments (1)
go/sdl/sdl-input.schema.yaml (1)
259-262: Inconsistent indentation foracceptproperty.The
acceptproperty's children have 8 extra spaces of indentation compared to sibling properties (as,port,proto). While valid YAML, this visual inconsistency may cause confusion during maintenance.accept: - items: - type: string - type: array + items: + type: string + type: array
📜 Review details
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Pro
📒 Files selected for processing (4)
go/sdl/schema_validator.go(1 hunks)go/sdl/sdl-input.schema.yaml(1 hunks)make/test.mk(1 hunks)ts/src/sdl/SDL/SDL.ts(19 hunks)
🚧 Files skipped from review as they are similar to previous changes (1)
- make/test.mk
🧰 Additional context used
🧬 Code graph analysis (1)
ts/src/sdl/SDL/SDL.ts (3)
ts/src/sdl/index.ts (1)
SdlValidationError(28-28)ts/src/sdl/types.ts (5)
v2Sdl(93-98)v3Sdl(100-105)v3Manifest(4-4)v3ComputeResources(218-224)v3DeploymentGroup(270-285)ts/src/network/index.ts (1)
NetworkId(2-2)
⏰ Context from checks skipped due to timeout of 90000ms. You can increase the timeout in your CodeRabbit configuration to a maximum of 15 minutes (900000ms). (5)
- GitHub Check: go
- GitHub Check: proto
- GitHub Check: coverage
- GitHub Check: sdl-parity
- GitHub Check: test
🔇 Additional comments (8)
go/sdl/sdl-input.schema.yaml (1)
1-390: LGTM! Comprehensive schema for SDL input validation.The schema covers all essential SDL structures with appropriate constraints including version enumeration, port validation, GPU vendor/model configuration, and storage attribute handling. The documented separation of structural (schema) vs semantic (Go runtime) validation is a good approach.
go/sdl/schema_validator.go (1)
1-203: LGTM! Well-structured schema validation implementation.The implementation addresses previous review concerns:
- Thread-safe logger access with
sync.RWMutex- Robust fallback chain (embedded → env var → file system) with actionable error messages
- Compile-time interface assertion for
noOpLogger- Lazy, one-time schema compilation with
sync.Oncets/src/sdl/SDL/SDL.ts (6)
249-318: Approve Go format conversion logic.The
convertToGoFormatmethod handles the necessary transformations for Go output parity:
- Key name conversions (camelCase ↔ snake_case)
- Special case handling (HTTPOptions → httpOptions, quantity → size)
- Empty
allOf/anyOfarrays converted to null- Recursive processing with undefined filtering
359-373: LGTM! Validation methods for service image and ports.The new validations enforce:
- Non-empty, non-whitespace-only image names
- Port values in the valid range (1-65535)
These complement the schema-level validations and provide clearer runtime errors.
1192-1194:v2Groupsdoesn't receive format parameter.The
groups()method passesformattov3Groupsbut not tov2Groups. If Go-format output is needed for v2 deployments, this will silently use TypeScript format instead.If this is intentional (v2 is legacy and doesn't need Go parity), consider adding a comment. Otherwise,
v2Groupsshould accept and use the format parameter.groups(format: OutputFormat = "typescript") { - return this.version === "beta2" ? this.v2Groups() : this.v3Groups(format); + return this.version === "beta2" ? this.v2Groups(format) : this.v3Groups(format); }
1030-1063: Approve rewritten endpoint sequence computation.The implementation correctly:
- Collects all IPs (including duplicates) to match Go behavior
- Skips global endpoints without IP
- Sorts for stable, deterministic ordering
- Assigns incrementing sequence numbers
The approach intentionally matches Go's behavior where duplicate IPs result in later assignments overwriting earlier ones.
1172-1181: Approve default GPU handling for Go parity.Always including GPU with units "0" when not specified matches Go's output behavior. The comment clearly documents this design decision.
1212-1219: Approve amount formatting with 18 decimal precision.Using
toFixed(18)for numeric amounts matches Go's precision requirements for parity testing. The fallback totoString()or"0"handles edge cases appropriately.
There was a problem hiding this comment.
Actionable comments posted: 0
♻️ Duplicate comments (1)
go/cli/go.mod (1)
215-216: Remove transitive xeipuuv dependencies from go/cli/go.mod.This is a duplicate of the previous review feedback: these packages are only used in
go/sdl(schema_validator.go, parity_test.go) and are transitive through the directpkg.akt.dev/go/sdl => ../sdldependency. Remove lines 215–216 and rungo mod tidyto restore only truly necessary dependencies.
🧹 Nitpick comments (2)
ts/src/sdl/SDL/SDL.ts (2)
1000-1012: Update JSDoc to document the format parameter.The implementation correctly supports format selection, but the JSDoc comment for
v3Manifestshould document the newformatparameter (purpose, valid values, default).
1192-1194: Update JSDoc for the format parameter.The format parameter support is well-implemented with consistent threading through the methods and proper 18-decimal precision for amount values. Consider updating the JSDoc comments for
groups()andv3Groups()to document the newformatparameter.Also applies to: 1196-1289
📜 Review details
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Pro
📒 Files selected for processing (2)
go/cli/go.mod(1 hunks)ts/src/sdl/SDL/SDL.ts(19 hunks)
⏰ Context from checks skipped due to timeout of 90000ms. You can increase the timeout in your CodeRabbit configuration to a maximum of 15 minutes (900000ms). (5)
- GitHub Check: sdl-parity
- GitHub Check: coverage
- GitHub Check: test
- GitHub Check: proto
- GitHub Check: go
🔇 Additional comments (9)
ts/src/sdl/SDL/SDL.ts (9)
2-2: LGTM!The switch to
js-yaml'sloadfunction and the addition of theOutputFormattype are both clean and appropriate for supporting Go/TypeScript output parity.Also applies to: 57-57
114-125: LGTM!The null/undefined guard and version-based casting correctly address the previous review comments. The parsing logic is now more robust and type-safe.
359-373: LGTM!The new validation methods correctly enforce non-empty image names and valid port ranges (1-65535), addressing the previous review comments about adding runtime validation.
908-920: LGTM!The deterministic sorting of storage keys and explicit null handling ensure consistent, predictable output for parity testing. The changes maintain backward compatibility while improving clarity.
Also applies to: 922-938, 973-973, 986-988
1082-1099: LGTM!The dynamic key selection (
quantityvssize) based on theasStringparameter correctly supports format compatibility and is consistent with theconvertToGoFormatlogic.Also applies to: 1101-1112
1172-1181: LGTM!Always including the GPU field (defaulting to units "0" when absent) ensures consistent structure and matches Go behavior as documented in the comment.
1307-1315: LGTM!The 18-decimal precision formatting is consistent with
v3Groupsand ensures Go-compatible numeric representation.
1030-1063: Review comment is incorrect—TypeScript implementation properly matches Go behavior.The Go implementation in
go/sdl/v2.go(lines 519–548) andgo/sdl/v2_1.go(lines 332–373) demonstrates that duplicate endpoint IPs are intentionally collected and allowed to overwrite in the result map. The assignment loop iterates through the sortedendpointNamesarray (which contains duplicates) and assigns sequence numbers without deduplication:for _, name := range endpointNames { endpointSeqNumber++ seqNo := endpointSeqNumber res[name] = seqNo }This produces the exact same behavior as the TypeScript code—duplicates overwrite previous assignments. The TypeScript implementation at lines 1054–1060 correctly mirrors this Go behavior. No changes are required.
Likely an incorrect or invalid review comment.
249-318: Parity tests already verify the conversion logic—no action needed unless refactoring is desired.The
convertToGoFormatmethod is already covered by comprehensive parity tests (make test-sdl-parity) that compare TypeScript output against Go fixtures for v2.0 and v2.1 SDLs. The field name conversions are correct and match the Go struct JSON serialization (e.g.,sequenceNumber→sequence_number, whileendpointSequenceNumberandexternalPortremain unchanged per their Go struct JSON tags).If maintainability is a concern, extracting a
convertFieldNamehelper could reduce the 70-line method's complexity, but this is optional and not required for correctness.
There was a problem hiding this comment.
Actionable comments posted: 1
🧹 Nitpick comments (4)
ts/src/sdl/SDL/parity.spec.ts (2)
23-36: Consider caching compiled schemas for better performance.The
validateAgainstSchemafunction compiles the schema on every call. For repeated validations against the same schema (e.g.,inputSchemaPath), consider caching the compiled validators.+const schemaCache = new Map<string, ReturnType<typeof ajv.compile>>(); + function validateAgainstSchema(name: string, data: unknown, schemaPath: string): void { - const schemaContent = fs.readFileSync(schemaPath, "utf8"); - const schema = load(schemaContent); - const validate = ajv.compile(schema as Record<string, unknown>); + let validate = schemaCache.get(schemaPath); + if (!validate) { + const schemaContent = fs.readFileSync(schemaPath, "utf8"); + const schema = load(schemaContent); + validate = ajv.compile(schema as Record<string, unknown>); + schemaCache.set(schemaPath, validate); + } const valid = validate(data);
125-143: Silent skip when invalid fixtures directory is missing.Unlike the Go implementation which uses
t.Skip()with a message, this silently returns when the directory doesn't exist. Consider adding visibility for debugging:describe("invalid SDLs rejected", () => { const invalidDir = path.join(fixturesRoot, "invalid"); if (!fs.existsSync(invalidDir)) { + it.skip("Invalid fixtures directory does not exist yet", () => {}); return; }go/sdl/parity_test.go (1)
161-171: Consider filtering by file extension in invalid SDL tests.The loop processes all non-directory entries. If non-YAML files (e.g.,
.gitkeep, README) exist in theinvaliddirectory, they would be tested and likely fail unexpectedly.for _, entry := range entries { if entry.IsDir() { continue } + if !strings.HasSuffix(entry.Name(), ".yaml") && !strings.HasSuffix(entry.Name(), ".yml") { + continue + } fixturePath := filepath.Join(invalidDir, entry.Name())Note: Add
"strings"to imports if applying this change.go/sdl/schema_validator.go (1)
122-133: Improve schema validation error formatting for readability (optional)Right now all schema errors are rendered via
%von a string slice, which tends to produce a single dense line. Joining them with newlines or commas would make logs and test failures easier to read, especially for complex SDLs.- if !result.Valid() { - var errors []string - for _, desc := range result.Errors() { - errors = append(errors, desc.String()) - } - return fmt.Errorf("schema validation failed: %v", errors) - } + if !result.Valid() { + var msgs []string + for _, desc := range result.Errors() { + msgs = append(msgs, desc.String()) + } + return fmt.Errorf("schema validation failed:\n - %s", strings.Join(msgs, "\n - ")) + }(Remember to add
stringsto the imports if you adopt this.)
📜 Review details
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Pro
📒 Files selected for processing (21)
go/sdl/parity_test.go(1 hunks)go/sdl/schema_validator.go(1 hunks)make/test.mk(1 hunks)testdata/sdl/invalid/credentials-missing-host.yaml(1 hunks)testdata/sdl/invalid/endpoint-not-used.yaml(1 hunks)testdata/sdl/invalid/invalid-port.yaml(1 hunks)testdata/sdl/invalid/missing-deployment.yaml(1 hunks)testdata/sdl/invalid/missing-image.yaml(1 hunks)testdata/sdl/invalid/missing-version.yaml(1 hunks)testdata/sdl/invalid/negative-cpu.yaml(1 hunks)testdata/sdl/invalid/persistent-without-mount.yaml(1 hunks)testdata/sdl/v2.0/gpu-basic/input.yaml(1 hunks)testdata/sdl/v2.0/http-options/input.yaml(1 hunks)testdata/sdl/v2.0/ip-endpoint/input.yaml(1 hunks)testdata/sdl/v2.0/multiple-services/input.yaml(1 hunks)testdata/sdl/v2.0/persistent-storage/input.yaml(1 hunks)testdata/sdl/v2.0/simple/input.yaml(1 hunks)testdata/sdl/v2.0/storage-classes/input.yaml(1 hunks)testdata/sdl/v2.1/credentials/input.yaml(1 hunks)testdata/sdl/v2.1/ip-endpoint/input.yaml(1 hunks)ts/src/sdl/SDL/parity.spec.ts(1 hunks)
✅ Files skipped from review due to trivial changes (1)
- testdata/sdl/invalid/invalid-port.yaml
🚧 Files skipped from review as they are similar to previous changes (9)
- testdata/sdl/invalid/missing-image.yaml
- testdata/sdl/v2.1/credentials/input.yaml
- testdata/sdl/invalid/missing-version.yaml
- testdata/sdl/invalid/persistent-without-mount.yaml
- testdata/sdl/v2.1/ip-endpoint/input.yaml
- testdata/sdl/v2.0/http-options/input.yaml
- testdata/sdl/invalid/negative-cpu.yaml
- testdata/sdl/v2.0/storage-classes/input.yaml
- testdata/sdl/v2.0/ip-endpoint/input.yaml
🧰 Additional context used
🧬 Code graph analysis (2)
go/sdl/parity_test.go (2)
go/sdl/sdl.go (1)
ReadFile(83-89)go/testutil/deployment.go (1)
DeploymentGroups(54-62)
ts/src/sdl/SDL/parity.spec.ts (1)
ts/src/sdl/SDL/SDL.ts (1)
SDL(89-1430)
⏰ Context from checks skipped due to timeout of 90000ms. You can increase the timeout in your CodeRabbit configuration to a maximum of 15 minutes (900000ms). (4)
- GitHub Check: coverage
- GitHub Check: sdl-parity
- GitHub Check: test
- GitHub Check: go
🔇 Additional comments (16)
testdata/sdl/invalid/endpoint-not-used.yaml (1)
32-34: Test fixture correctly represents the endpoint-not-used invalid case.The endpoint
myendpointis properly defined but intentionally not referenced by any service or deployment, making this a valid test case for validating SDL rejection of unused endpoints.testdata/sdl/v2.0/gpu-basic/input.yaml (3)
12-27: GPU resource specification looks correct.The gpu-compute profile properly defines:
- CPU units in millicores (1000m = 1 core)
- Memory and storage in standard units (Gi)
- GPU count with vendor/model attributes (NVIDIA RTX3080)
This is a realistic test case for GPU workloads.
1-1: Schema reference path is correct and schema file exists.The yaml-language-server directive at line 1 correctly references the SDL input schema. The schema file exists at
go/sdl/sdl-input.schema.yaml, and the relative path../../../../go/sdl/sdl-input.schema.yamlfrom this fixture location properly resolves to it.
3-43: Test fixture is properly paired and schema-compliant.The fixture structure is well-formed and properly organized. Verification confirms:
- ✓
manifest.jsonexists and validates againstspecs/sdl/manifest-output.schema.yaml- ✓
groups.jsonexists and validates againstspecs/sdl/groups-output.schema.yaml- ✓ GPU configuration (RTX3080, 1 unit) correctly maps from input to both outputs
- ✓ Resource specifications (CPU: 1000m, Memory: 2Gi, Storage: 10Gi) accurately converted
- ✓ Placement attributes and pricing preserved in generated outputs
The AI-generated test data is accurate and ready for parity testing.
testdata/sdl/v2.0/persistent-storage/input.yaml (1)
1-48: Well-structured persistent storage fixture.The SDL v2.0 persistent storage test fixture correctly demonstrates a database service with persistent volume, placement constraints, and signedBy policies. The hardcoded
POSTGRES_PASSWORD=secretis acceptable for test fixture purposes.testdata/sdl/invalid/missing-deployment.yaml (1)
1-27: Valid invalid-SDL fixture for missing deployment case.The fixture correctly omits the
deploymentsection while providing validservicesandprofiles, which should trigger validation errors in both implementations.make/test.mk (1)
50-63: Well-structured Makefile targets for SDL parity testing.The targets establish a clear workflow:
generate-sdl-fixturescreates the expected outputs, thentest-sdl-parityruns both Go and TypeScript tests against them. The dependency on$(AKASH_TS_NODE_MODULES)ensures npm dependencies are available.testdata/sdl/v2.0/simple/input.yaml (1)
1-47: Good coverage of multiple expose rules and signedBy constraints.The fixture exercises both TCP and UDP expose rules, hostname acceptance, and signedBy constraints with
anyOfandallOf. The placeholder values for signedBy are acceptable for testing structural parity.testdata/sdl/invalid/credentials-missing-host.yaml (1)
1-35: Valid invalid-SDL fixture for missing credentials host.The fixture correctly defines a
credentialsblock withusernameandpasswordbut omits the requiredhostfield, which should trigger validation errors in both implementations.ts/src/sdl/SDL/parity.spec.ts (2)
72-97: v2.0 parity tests are well-structured.The test flow correctly validates input against schema, parses with
beta2, compares manifest/groups against fixtures, and validates outputs against their schemas. The use oftoEqualfor deep comparison is appropriate.
99-123: v2.1 parity tests mirror v2.0 structure correctly.The v2.1 tests use
beta3for parsing, matching the expected version-to-network mapping. The structure is consistent with v2.0 tests.go/sdl/parity_test.go (3)
19-27: Good use of sync.Once for schema caching.Caching the compiled schemas with
sync.Onceensures schemas are compiled only once across all test runs, improving performance without sacrificing thread safety.
37-67: testParity logic is comprehensive and well-structured.The function correctly iterates fixtures, validates existence of required files, and runs both schema validation and fixture comparison as subtests. The
t.Fatalffor missing fixtures provides clear actionable guidance.
118-149: Fixture validation uses proper JSON unmarshaling for comparison.Unmarshaling both expected and actual to
anybefore comparison ensures JSON equivalence regardless of Go struct field ordering, which is correct for parity testing.testdata/sdl/v2.0/multiple-services/input.yaml (1)
3-92: Multi‑service SDL fixture looks consistent and realisticThe SDL models a clear web→api→db dependency chain, scoped exposure rules, and distinct compute/placement/deployment profiles, which is ideal for exercising parity between implementations. I don’t see structural issues in this fixture for the stated test purpose.
go/sdl/schema_validator.go (1)
14-88: Embedded schema + lazy compilation design looks solidUsing go:embed with a byte slice and compiling the YAML schema once via
sync.Onceis clean and thread-safe. Error paths for missing/invalid schema (load, YAML parse, JSON marshal, compile) are preserved viaschemaCompileError, so callers see a stable failure once initialization fails. No concurrency or correctness issues spotted in this part.
There was a problem hiding this comment.
Actionable comments posted: 0
♻️ Duplicate comments (1)
ts/src/sdl/SDL/SDL.ts (1)
1030-1063: Duplicate IPs cause sequence number assignments to be overwritten with non-contiguous numbering.The Go implementation has identical logic without deduplication. When the same IP appears multiple times in the sorted array, successive assignments overwrite previous ones, resulting in gaps in sequence numbers. For example, with duplicates
["10.0.0.1", "10.0.0.1", "10.0.0.2"], the final map would be{"10.0.0.1": 2, "10.0.0.2": 3}instead of the intended{"10.0.0.1": 1, "10.0.0.2": 2}.Apply deduplication before assignment:
- // Make the assignment stable (matching Go implementation) - endpointNames.sort(); + // Make the assignment stable with deduplication + const uniqueNames = [...new Set(endpointNames)].sort(); // Assign sequence numbers: start at 0, so first is 1 (matching Go) const res: Record<string, number> = {}; let endpointSeqNumber = 0; - for (const name of endpointNames) { + for (const name of uniqueNames) { endpointSeqNumber++; res[name] = endpointSeqNumber; }
🧹 Nitpick comments (1)
ts/src/sdl/SDL/parity.spec.ts (1)
13-14: Replaceanywith more specific type for Ajv validator.Static analysis flagged the use of
anyin the Ajv type cast. Use a more specific type for better type safety.Apply this change:
-const Ajv = AjvModule as typeof AjvModule & { new (options?: { allErrors?: boolean }): any }; +const Ajv = AjvModule as typeof AjvModule & { new (options?: { allErrors?: boolean }): AjvModule };
📜 Review details
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Pro
📒 Files selected for processing (13)
go/sdl/schema_validator.go(1 hunks)testdata/sdl/v2.0/placement/groups.json(1 hunks)testdata/sdl/v2.0/placement/input.yaml(1 hunks)testdata/sdl/v2.0/placement/manifest.json(1 hunks)testdata/sdl/v2.0/port-ranges/groups.json(1 hunks)testdata/sdl/v2.0/port-ranges/input.yaml(1 hunks)testdata/sdl/v2.0/port-ranges/manifest.json(1 hunks)testdata/sdl/v2.0/pricing/groups.json(1 hunks)testdata/sdl/v2.0/pricing/input.yaml(1 hunks)testdata/sdl/v2.0/pricing/manifest.json(1 hunks)ts/src/sdl/SDL/SDL.ts(18 hunks)ts/src/sdl/SDL/parity.spec.ts(1 hunks)ts/src/sdl/types.ts(1 hunks)
✅ Files skipped from review due to trivial changes (3)
- testdata/sdl/v2.0/placement/groups.json
- testdata/sdl/v2.0/port-ranges/manifest.json
- testdata/sdl/v2.0/pricing/groups.json
🚧 Files skipped from review as they are similar to previous changes (1)
- go/sdl/schema_validator.go
🧰 Additional context used
🧬 Code graph analysis (2)
ts/src/sdl/SDL/parity.spec.ts (2)
ts/src/sdk/provider/auth/jwt/validate-payload.ts (1)
schema(8-394)ts/src/sdl/SDL/SDL.ts (5)
validate(131-141)validate(320-335)SDL(89-1427)manifest(1014-1016)groups(1192-1194)
ts/src/sdl/SDL/SDL.ts (1)
ts/src/sdl/types.ts (5)
v2Sdl(93-98)v3Sdl(100-105)v3Manifest(4-4)v3ComputeResources(218-224)v3DeploymentGroup(270-285)
🪛 GitHub Check: ts
ts/src/sdl/SDL/parity.spec.ts
[failure] 13-13:
Unexpected any. Specify a different type
⏰ Context from checks skipped due to timeout of 90000ms. You can increase the timeout in your CodeRabbit configuration to a maximum of 15 minutes (900000ms). (5)
- GitHub Check: coverage
- GitHub Check: test
- GitHub Check: sdl-parity
- GitHub Check: proto
- GitHub Check: go
🔇 Additional comments (17)
testdata/sdl/v2.0/placement/input.yaml (2)
1-1: No issues found. The schema reference path../../../../go/sdl/sdl-input.schema.yamlcorrectly resolves from the fixture location togo/sdl/sdl-input.schema.yaml, and the file exists with proper permissions. IDE YAML validation will work as intended.
12-32: Profile and pricing configuration is already validated by the parity test.The placement fixture is automatically tested by
TestParityV2_0()inparity_test.go, which performs deep equality comparison between the generated manifest and groups against the expected fixture files. This covers all the verification points mentioned: theanyOfconstraint in signedBy, the pricing denomuaktand amount1000, resource specifications, and count consistency. All values ininput.yamlcorrectly matchgroups.jsonandmanifest.json.testdata/sdl/v2.0/placement/manifest.json (2)
1-70: Confirm AI-generated test fixture accuracy.Per the PR description, this test data was generated using AI. While the JSON structure appears valid and memory/storage byte conversions are correct (512 MiB → 536870912 bytes), the CPU unit transformation and overall parity with the corresponding input.yaml and groups.json requires manual validation to ensure test expectations are accurate.
#!/bin/bash # Verify schema validation and fixture consistency # Check if the schema file referenced in input.yaml exists echo "=== Checking schema file existence ===" SCHEMA_PATH="go/sdl/sdl-input.schema.yaml" if [ -f "$SCHEMA_PATH" ]; then echo "Schema found at: $SCHEMA_PATH" echo "Schema exists: YES" else echo "Schema not found at: $SCHEMA_PATH" echo "This may cause IDE validation errors in input.yaml" fi # Verify JSON validity of manifest echo -e "\n=== Validating manifest.json structure ===" if command -v jq &> /dev/null; then if jq empty testdata/sdl/v2.0/placement/manifest.json 2>/dev/null; then echo "Manifest JSON is valid" else echo "Manifest JSON has validation errors" jq empty testdata/sdl/v2.0/placement/manifest.json fi fi
13-16: No unit conversion issue detected.The CPU value "500m" in input.yaml is correctly converted to "500" in the manifest. Akash's cpuQuantity parser strips the "m" suffix and stores the integer value directly, where each unit represents 1/1000 of a CPU (so 500 units = 0.5 cores). This matches the standard Kubernetes milliCPU notation and is consistent across all test fixtures.
ts/src/sdl/types.ts (1)
279-283: LGTM! Nullable types align with schema and Go output.The nullable adjustments for
attributes,allOf, andanyOfcorrectly match the Go implementation's behavior and the JSON schemas defined in specs/sdl. This ensures type safety when these fields are absent or explicitly null.testdata/sdl/v2.0/pricing/input.yaml (1)
1-41: LGTM! Valid SDL v2.0 pricing test fixture.The fixture correctly defines a two-provider deployment scenario with different pricing configurations. The schema reference enables IDE validation, and the structure aligns with v2.0 SDL specifications.
testdata/sdl/v2.0/port-ranges/input.yaml (1)
1-43: LGTM! Valid port-ranges test fixture.The fixture correctly defines multiple port exposures including TCP and UDP protocols, providing good coverage for port-range scenarios in parity testing.
testdata/sdl/v2.0/pricing/manifest.json (1)
1-138: LGTM! Valid manifest fixture for pricing test.The fixture correctly represents the expected manifest output for a two-provider deployment, with both providers exposing identical service configurations. The structure is consistent with the manifest-output.schema.yaml.
ts/src/sdl/SDL/SDL.ts (7)
114-125: LGTM! Version-aware type casting correctly implemented.The addition of the null/undefined check and version-based casting to v2Sdl or v3Sdl properly addresses previous review comments. This ensures type consistency between the parsed data and the internal representation.
249-318: LGTM! Comprehensive Go format conversion implemented.The
convertToGoFormatmethod correctly handles the transformations needed for Go parity:
- Recursive conversion of nested structures
- Key name normalization (camelCase ↔ snake_case)
- Empty array → null conversion for signedBy fields
- GPU field preservation matching Go behavior
359-373: LGTM! Essential service validations added.The new
validateServiceImageandvalidateServicePortsmethods add critical validation:
- Ensuring non-empty image names prevents runtime failures
- Port range validation (1-65535) prevents invalid service configurations
908-938: LGTM! Deterministic storage key ordering implemented.The sorting of storage keys in both v2 and v3 manifest service params ensures consistent, reproducible output across runs. This is essential for manifest version hashing and reliable parity testing.
1000-1012: LGTM! Format parameter enables Go parity while preserving compatibility.The addition of the
formatparameter with a default value of"typescript"maintains backward compatibility for existing consumers while enabling Go-format output for parity tests. The conditional conversion at the end ensures efficient processing.
1172-1181: LGTM! GPU field always included matching Go behavior.The implementation correctly ensures GPU is always present in groups output with units defaulting to "0" when not specified. This matches the Go implementation's behavior and is necessary for schema conformance.
1212-1220: LGTM! Consistent 18-decimal precision for amounts.Both
v3Groupsandv2Groupscorrectly format amounts with 18 decimal places usingtoFixed(18), matching Go's decimal precision. This ensures consistent numeric representation across implementations and prevents floating-point precision issues in parity tests.Also applies to: 1306-1314
testdata/sdl/v2.0/port-ranges/groups.json (1)
38-50: The duplicate endpoint entries are intentional and correct. The input defines three ports (8080 TCP, 10000 UDP, 20000 UDP), and the groups.json output correctly includes three endpoint entries—one for each port. The two UDP ports both havekind: 1, but they are separate endpoint entries because they represent distinct port exposures. This matches the behavior of the authoritative Go implementation (sdl.DeploymentGroups()), confirming the fixture is accurately generated.ts/src/sdl/SDL/parity.spec.ts (1)
85-97: Theversionparameter controls SDL parsing but not manifest/groups output format.Lines 90-91 always call
sdl.v3Manifest(true, "go")andsdl.v3Groups("go")regardless of theversionparameter ("beta2" | "beta3"). While the version parameter correctly controls SDL creation viaSDL.fromString(inputBytes, version), the output methods bypass version-aware dispatch.This behavior appears intentional because:
- Go format is a v3-only feature (v2Manifest lacks format parameter and convertToGoFormat)
- The
groups()method supports format parameter butmanifest()does not, creating asymmetryHowever, this design means v2.0 (beta2) fixtures are output as v3 format, which may warrant clarification:
- Is this intentional for parity testing?
- Or should the test use version-aware dispatch (
sdl.manifest(true)andsdl.groups("go")when available)?Consider adding a comment explaining why v3 output is always used for both beta2 and beta3 fixtures despite the version parameter being provided.
There was a problem hiding this comment.
Actionable comments posted: 2
🧹 Nitpick comments (2)
ts/src/sdl/SDL/parity.spec.ts (2)
13-14: Consider simplifying AJV instantiation.The double cast through
unknownis verbose. A common pattern for handling ESM/CJS interop:const Ajv = (AjvModule as { default?: typeof AjvModule }).default || AjvModule; const ajv = new Ajv({ allErrors: true });This avoids the
@ts-expect-errorand is more readable.
85-97:sdlis returned but unused by callers.
validateFixturesonly destructures{ manifest, groups }. Either removesdlfrom the return object or use it in the caller.- return { sdl, manifest, groups }; + return { manifest, groups };
📜 Review details
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Pro
📒 Files selected for processing (2)
go/sdl/schema_validator.go(1 hunks)ts/src/sdl/SDL/parity.spec.ts(1 hunks)
⏰ Context from checks skipped due to timeout of 90000ms. You can increase the timeout in your CodeRabbit configuration to a maximum of 15 minutes (900000ms). (5)
- GitHub Check: test
- GitHub Check: sdl-parity
- GitHub Check: coverage
- GitHub Check: proto
- GitHub Check: go
🔇 Additional comments (12)
go/sdl/schema_validator.go (8)
14-19: Excellent use ofembedfor schema portability.The compile-time embedding approach eliminates the previous runtime path resolution fragility and makes binaries fully portable. The documentation clearly explains the benefits.
21-31: Well-designed validator with proper synchronization.The combination of
sync.Oncefor lazy compilation,RWMutexfor logger access, and a no-op logger default ensures thread-safe operation with minimal contention.
33-54: Thread-safe logger injection with nil safety.All previous concerns addressed: the mutex ensures thread safety, nil loggers are safely handled, and the documentation clearly explains usage and thread-safety guarantees.
56-63: LGTM: Straightforward embedded schema access.The empty-check provides a clear error if embedding fails, though in practice the build would fail if the file were missing.
65-94: Solid lazy compilation with comprehensive error handling.The
sync.Oncepattern ensures thread-safe, one-time compilation. The YAML→JSON conversion via intermediate Go data structures is the standard approach forgojsonschema, and error caching is appropriate since the embedded schema cannot change at runtime.
96-106: Complete no-op logger implementation with interface assertion.The compile-time assertion at line 106 (addressing a previous comment) ensures interface compliance, and the no-op behavior is correctly implemented throughout.
108-142: Standard validation pipeline with clear error reporting.The YAML→JSON conversion correctly handles arbitrary input structures, and validation errors are comprehensively collected and formatted for actionable feedback.
144-172: Effective mismatch detection for observability.The logic correctly identifies divergence between schema and Go validation, using read locks for performant concurrent access. The asymmetric field naming (
schema_errorvsgo_error) appropriately highlights which validator failed.ts/src/sdl/SDL/parity.spec.ts (4)
1-8: LGTM!Imports are appropriate for the parity test requirements.
16-36: LGTM!Clean interface definition and effective schema caching pattern.
51-83: LGTM!Good error messages with actionable hints for missing fixtures.
99-108: LGTM!Clean fixture validation with proper deep equality comparison.
There was a problem hiding this comment.
Actionable comments posted: 0
🧹 Nitpick comments (1)
ts/src/sdl/SDL/parity.spec.ts (1)
123-140: LGTM! Test organization is clear and addresses previous feedback.The explicit
it.skipwhen the invalid fixtures directory is missing properly addresses the previous review concern about silent failures. The test structure is clean and maintainable.One optional enhancement: The invalid SDL tests currently expect any throw (line 137). If specific error types or messages are important for parity, consider asserting on
SdlValidationErrorspecifically. However, for broad rejection testing, the current approach is acceptable.
📜 Review details
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Pro
⛔ Files ignored due to path filters (1)
ts/src/sdl/SDL/__snapshots__/SDL.spec.ts.snapis excluded by!**/*.snap
📒 Files selected for processing (3)
ts/src/sdl/SDL/SDL.spec.ts(10 hunks)ts/src/sdl/SDL/parity.spec.ts(1 hunks)ts/test/fixtures/sdl/groups-basic-snapshot.json(3 hunks)
🧰 Additional context used
🧬 Code graph analysis (1)
ts/src/sdl/SDL/parity.spec.ts (1)
ts/src/sdl/SDL/SDL.ts (5)
validate(131-141)validate(320-335)SDL(89-1427)manifest(1014-1016)groups(1192-1194)
⏰ Context from checks skipped due to timeout of 90000ms. You can increase the timeout in your CodeRabbit configuration to a maximum of 15 minutes (900000ms). (5)
- GitHub Check: proto
- GitHub Check: test
- GitHub Check: sdl-parity
- GitHub Check: go
- GitHub Check: coverage
🔇 Additional comments (6)
ts/src/sdl/SDL/SDL.spec.ts (1)
613-682: LGTM! Test expectations updated consistently.The changes correctly update test expectations to match the new SDL schema:
- Field names:
quantity→sizefor memory and storage- Added GPU blocks with zero units for completeness
- Price precision increased to 18 decimal places
allOfset tonullinstead of empty arrayThese changes are consistent with the type system updates described in the PR.
ts/test/fixtures/sdl/groups-basic-snapshot.json (1)
1-63: LGTM! Fixture updated to match new schema.The snapshot correctly reflects the updated SDL output format:
- Empty arrays changed to
nullforattributes,allOf, andanyOf(semantically clearer)- GPU units simplified from object representation to string
"0"- Field renames:
quantity→sizefor memory- Price precision increased to 18 decimal places
All changes are consistent with the schema updates across the codebase.
ts/src/sdl/SDL/parity.spec.ts (4)
23-49: LGTM! Well-implemented schema validation with caching.The schema compilation caching is a good performance optimization, and the error formatting provides clear diagnostics with path and keyword information. The fallback to
"(root)"for empty instance paths properly addresses the previous review feedback.
51-83: LGTM! Robust fixture loading with clear error messages.The function properly validates fixture structure and provides actionable error messages (e.g., suggesting
make generate-sdl-fixtureswhen files are missing). This will prevent silent failures in CI and help developers diagnose issues quickly.
85-108: LGTM! Clean separation of validation concerns.The two-function approach cleanly separates:
- Schema validation: Ensures input/output conform to JSON schemas
- Parity validation: Ensures TS implementation matches Go expectations
The deliberate lack of error handling in
validateSchemasallows SDL parsing errors to surface naturally, which is appropriate for these tests.
13-14: Documentation for AJV instantiation is already present; no action needed.The code includes an explanatory comment (
@ts-expect-error - AjvModule has non-standard export, cast needed for instantiation) that adequately documents why the complex type casting is necessary. This pattern is a standard workaround for AJV v8.12.0's CommonJS/ESM interop behavior and is the sole instance in the codebase. No additional changes required.
vertex451
force-pushed
the
artem/sdl-parity-tests
branch
from
bd7411b to
eb3c877
Compare
There was a problem hiding this comment.
Actionable comments posted: 0
Caution
Some comments are outside the diff and can’t be posted inline due to platform limitations.
⚠️ Outside diff range comments (1)
ts/src/sdl/SDL/SDL.ts (1)
1196-1288: Critical bug in v3Groups: endpoints property accessed at wrong nesting level and uses+=for array concatenationThe aggregation path (line 1278) incorrectly accesses
group!.dgroup.resources[location].endpoints, but endpoints are actually nested atgroup!.dgroup.resources[location].resource.endpoints. This is evident from the creation path (line 1267) where the resource object is wrapped as{ resource: res, ... }andres.endpointswas set at line 1260.Additionally, using the
+=operator on arrays causes string concatenation instead of merging. For example, concatenating an undefined or array value with the endpoints array produces a corrupted string like"undefined[object Object]"rather than properly combining arrays.This bug corrupts endpoint data whenever multiple services share the same placement and profile. Update the aggregation logic to access the nested
resource.endpointsproperty and use array spreading instead of+=:} else { const endpoints = this.v3ServiceResourceEndpoints(service) || []; const resource: any = group!.dgroup.resources[location].resource; const existing = (resource.endpoints as any[]) || []; group!.dgroup.resources[location].count += svcdepl.count; resource.endpoints = [...existing, ...endpoints]; }
🧹 Nitpick comments (2)
go/node/client/v1beta3/tx.go (1)
535-551: Broadened sequence-resync trigger is good; consider also handling ErrInvalidSequence ABCI codeExpanding the
syncSequencecondition to resync whenrErrmatches eithersdkerrors.ErrWrongSequenceorsdkerrors.ErrInvalidSequenceis a solid improvement and should reduce stuck sequence issues. The ABCI-code branch, however, still only checkstxResp.Code == sdkerrors.ErrWrongSequence.ABCICode(). If Cosmos can surfaceErrInvalidSequencevia a distinct ABCI code, you may also want to treat that as a resync signal:- if (rErr != nil && (sdkerrors.ErrWrongSequence.Is(rErr) || sdkerrors.ErrInvalidSequence.Is(rErr))) || (valid && (txResp.Code == sdkerrors.ErrWrongSequence.ABCICode())) { + if (rErr != nil && (sdkerrors.ErrWrongSequence.Is(rErr) || sdkerrors.ErrInvalidSequence.Is(rErr))) || + (valid && (txResp.Code == sdkerrors.ErrWrongSequence.ABCICode() || txResp.Code == sdkerrors.ErrInvalidSequence.ABCICode())) {This is optional and only needed if such a code path actually exists.
It would be worth double-checking against the Cosmos SDK error definitions (and your existing tests) whether
ErrInvalidSequenceis ever surfaced via ABCI code in your flows.go/sdl/sdl.go (1)
92-135: Schema validation is observational only; consider documenting fail‑open behavior
Readalways runsvalidateInputAgainstSchemabut never returnsschemaErr—it only logs mismatches viacheckSchemaValidationResult. This means production callers will accept SDLs that fail the schema as long as Go validation passes. That seems intentional but is a policy choice worth calling out (e.g., in package docs) so future changes don’t accidentally flip behavior to fail‑closed without explicit decision.
📜 Review details
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Pro
⛔ Files ignored due to path filters (2)
go/cli/go.sumis excluded by!**/*.sumts/src/sdl/SDL/__snapshots__/SDL.spec.ts.snapis excluded by!**/*.snap
📒 Files selected for processing (43)
go/cli/go.mod(1 hunks)go/node/client/v1beta3/tx.go(1 hunks)go/provider/client/client.go(1 hunks)go/sdl/cmd/generate-sdl-fixtures/main.go(1 hunks)go/sdl/go.mod(2 hunks)go/sdl/parity_test.go(1 hunks)go/sdl/schema_validator.go(1 hunks)go/sdl/sdl-input.schema.yaml(1 hunks)go/sdl/sdl.go(2 hunks)go/util/jwt/jwt_test.go(1 hunks)make/test.mk(1 hunks)specs/sdl/README.md(1 hunks)testdata/sdl/invalid/credentials-missing-host.yaml(1 hunks)testdata/sdl/invalid/endpoint-not-used.yaml(1 hunks)testdata/sdl/invalid/invalid-port.yaml(1 hunks)testdata/sdl/invalid/missing-deployment.yaml(1 hunks)testdata/sdl/invalid/missing-image.yaml(1 hunks)testdata/sdl/invalid/missing-version.yaml(1 hunks)testdata/sdl/invalid/negative-cpu.yaml(1 hunks)testdata/sdl/invalid/persistent-without-mount.yaml(1 hunks)testdata/sdl/v2.0/gpu-basic/input.yaml(1 hunks)testdata/sdl/v2.0/http-options/input.yaml(1 hunks)testdata/sdl/v2.0/ip-endpoint/input.yaml(1 hunks)testdata/sdl/v2.0/multiple-services/input.yaml(1 hunks)testdata/sdl/v2.0/persistent-storage/input.yaml(1 hunks)testdata/sdl/v2.0/placement/groups.json(1 hunks)testdata/sdl/v2.0/placement/input.yaml(1 hunks)testdata/sdl/v2.0/placement/manifest.json(1 hunks)testdata/sdl/v2.0/port-ranges/groups.json(1 hunks)testdata/sdl/v2.0/port-ranges/input.yaml(1 hunks)testdata/sdl/v2.0/port-ranges/manifest.json(1 hunks)testdata/sdl/v2.0/pricing/groups.json(1 hunks)testdata/sdl/v2.0/pricing/input.yaml(1 hunks)testdata/sdl/v2.0/pricing/manifest.json(1 hunks)testdata/sdl/v2.0/simple/input.yaml(1 hunks)testdata/sdl/v2.0/storage-classes/input.yaml(1 hunks)testdata/sdl/v2.1/credentials/input.yaml(1 hunks)testdata/sdl/v2.1/ip-endpoint/input.yaml(1 hunks)ts/src/sdl/SDL/SDL.spec.ts(10 hunks)ts/src/sdl/SDL/SDL.ts(18 hunks)ts/src/sdl/SDL/parity.spec.ts(1 hunks)ts/src/sdl/types.ts(1 hunks)ts/test/fixtures/sdl/groups-basic-snapshot.json(3 hunks)
✅ Files skipped from review due to trivial changes (1)
- testdata/sdl/invalid/credentials-missing-host.yaml
🚧 Files skipped from review as they are similar to previous changes (25)
- testdata/sdl/invalid/invalid-port.yaml
- testdata/sdl/v2.0/port-ranges/manifest.json
- ts/test/fixtures/sdl/groups-basic-snapshot.json
- testdata/sdl/v2.0/placement/input.yaml
- make/test.mk
- go/sdl/cmd/generate-sdl-fixtures/main.go
- testdata/sdl/v2.0/port-ranges/groups.json
- ts/src/sdl/SDL/SDL.spec.ts
- specs/sdl/README.md
- testdata/sdl/invalid/missing-image.yaml
- go/sdl/sdl-input.schema.yaml
- testdata/sdl/v2.0/placement/groups.json
- testdata/sdl/v2.1/ip-endpoint/input.yaml
- testdata/sdl/v2.0/placement/manifest.json
- go/cli/go.mod
- testdata/sdl/v2.0/http-options/input.yaml
- testdata/sdl/invalid/endpoint-not-used.yaml
- testdata/sdl/invalid/negative-cpu.yaml
- testdata/sdl/v2.0/pricing/manifest.json
- testdata/sdl/v2.1/credentials/input.yaml
- testdata/sdl/v2.0/simple/input.yaml
- testdata/sdl/v2.0/storage-classes/input.yaml
- testdata/sdl/v2.0/port-ranges/input.yaml
- testdata/sdl/invalid/missing-deployment.yaml
- testdata/sdl/v2.0/pricing/groups.json
🧰 Additional context used
🧬 Code graph analysis (3)
go/sdl/sdl.go (1)
go/testutil/deployment.go (1)
DeploymentGroups(54-62)
go/sdl/parity_test.go (3)
go/testutil/base.go (1)
Name(28-30)go/sdl/sdl.go (1)
ReadFile(83-89)go/testutil/deployment.go (1)
DeploymentGroups(54-62)
ts/src/sdl/SDL/parity.spec.ts (1)
ts/src/sdl/SDL/SDL.ts (5)
validate(131-141)validate(320-335)SDL(89-1427)manifest(1014-1016)groups(1192-1194)
⏰ Context from checks skipped due to timeout of 90000ms. You can increase the timeout in your CodeRabbit configuration to a maximum of 15 minutes (900000ms). (3)
- GitHub Check: coverage
- GitHub Check: test
- GitHub Check: go
🔇 Additional comments (14)
go/util/jwt/jwt_test.go (1)
188-190: Slice expansion change is correct and simplifies control flowUsing
append(res, s.initClaims(testCases[i])...)correctly flattens the returned[]jwtTestCaseintoresand is equivalent to the previous manual loop over the slice, while being clearer and less error‑prone. No behavioral issues here.testdata/sdl/v2.0/persistent-storage/input.yaml (1)
3-47: Well-structured SDL v2.0 test fixture.The YAML configuration is syntactically correct and well-organized. It effectively demonstrates a practical persistent-storage scenario with:
- Postgres service with port exposure and parameterized storage mounts
- Compute profile with CPU, memory, and persistent volume declarations
- Placement profile with regional attributes and pricing tiers
- Deployment referencing the defined profiles
This is appropriate for parity testing.
go/sdl/go.mod (1)
7-16: gojsonschema/log additions look consistent with new SDL schema validationDirectly requiring
cosmossdk.io/logandgithub.com/xeipuuv/gojsonschemaplus pinninggojsonpointer/gojsonreferenceas indirects is aligned with the new schema validator usage and keeps versions explicit; nothing stands out as problematic.Please ensure
go mod tidyand CI’s module step remain clean after these changes and that there are no unexpected version conflicts with other Cosmos SDK components.Also applies to: 166-167
testdata/sdl/v2.0/pricing/input.yaml (1)
1-41: Multi-provider pricing fixture looks structurally soundThe v2.0 SDL input is well-formed: shared
services.app, a singleprofiles.compute.app-computeused by bothplacement.provider-a/provider-bpricing blocks, and matching entries underdeployment.app. This should be a good parity case for cross-provider pricing differences.testdata/sdl/v2.0/ip-endpoint/input.yaml (1)
1-49: IP endpoint fixture cleanly models multi-port, single-endpoint usageThis v2.0 SDL cleanly expresses two global TCP ports both bound to the same named IP endpoint (
myendpoint), with matchingendpoints.myendpoint.kind: ipand a plausible placement/signedBy section. It should exercise endpoint resolution and manifest/group generation nicely.go/provider/client/client.go (1)
99-103: Receiver rename on defaultCertQuerier is purely cosmeticDropping the blank receiver name (
_) from(*defaultCertQuerier) GetAccountCertificate(...)is a style-only change; method set and behavior are unchanged, and the default implementation still consistently reports on-chain certs as unavailable.A quick
go test ./...in modules that depend on this client should confirm there are no interface-satisfaction regressions aroundCertQuerierusage.testdata/sdl/invalid/persistent-without-mount.yaml (1)
1-35: Invalid fixture correctly captures “persistent volume without service mount”This SDL marks storage as
persistent: trueunderprofiles.compute.web.resources.storagebut never declares a matching service-level mount, which is exactly the invalid scenario you want to exercise. It should be a useful negative test for both Go and TS SDL validators.ts/src/sdl/types.ts (1)
270-285: Nullability for v3DeploymentGroup requirements matches schema/fixturesAllowing
requirements.attributesandrequirements.signedBy.allOf/anyOfto benullbetter mirrors the actual group-output shape and schemas where these fields are explicitly nullable when no constraints exist. This should eliminate type/fixture mismatches; just ensure consuming code treats these asnull-or-array rather than assuming[].A quick
tscrun and parity test pass (TS + Go) will confirm there are no remaining call sites that assume these properties are always arrays.testdata/sdl/invalid/missing-version.yaml (1)
1-31: Missing-version invalid fixture is appropriate and minimalThis fixture omits the top-level
versionwhile keeping the rest of the SDL minimal but valid-looking, which is ideal for asserting that both implementations hard-reject SDLs without an explicit version.testdata/sdl/v2.0/gpu-basic/input.yaml (1)
3-43: GPU v2.0 fixture looks schema‑compatible and parity‑readyService, GPU, placement, and pricing structure all match the expected v2.0 SDL shape and GPU validation rules; good coverage for basic GPU parity.
ts/src/sdl/SDL/parity.spec.ts (1)
10-140: Parity harness is well‑structured and aligned with Go testsShared schema compilation, fixture loading, and the v2.0/v2.1 + invalid‑SDL flows mirror the Go parity tests and exercise the new
v3Manifest("go")/v3Groups("go")paths effectively.go/sdl/parity_test.go (1)
29-149: Go parity tests thoroughly cover fixtures and schemasThe parity tests correctly enforce three things per fixture: input schema validity, Go SDL processing (manifest + groups) validity, and JSON equality against generated fixtures. The structure looks robust and maintainable.
testdata/sdl/v2.0/multiple-services/input.yaml (1)
3-92: Multi‑service fixture nicely exercises complex SDL shapesThis SDL covers multiple services, service dependencies, internal exposes, and multi‑profile pricing under a single placement, which is ideal for stressing manifest/groups parity.
go/sdl/schema_validator.go (1)
14-172: Embedded schema validator is thread‑safe and production‑friendlyEmbedding
sdl-input.schema.yaml, caching compilation withsync.Once, and guarding logger access with an RWMutex gives a safe, self‑contained validator. ThevalidateInputAgainstSchemaandcheckSchemaValidationResultAPIs integrate cleanly withReadwithout changing its public contract.
There was a problem hiding this comment.
Actionable comments posted: 1
🧹 Nitpick comments (1)
go/sdl/parity_test.go (1)
54-60: Consider usingt.Errorfto report all missing fixtures at once.Using
t.Fatalfstops the test run immediately on the first missing fixture. If multiple fixtures are missing, the developer must fix them one at a time. Consider collecting all missing fixtures and reporting them together.+ var missingFixtures []string for _, entry := range entries { if !entry.IsDir() { continue } fixtureName := entry.Name() fixtureDir := filepath.Join(fixturesDir, fixtureName) - inputPath := filepath.Join(fixtureDir, "input.yaml") manifestPath := filepath.Join(fixtureDir, "manifest.json") groupsPath := filepath.Join(fixtureDir, "groups.json") if _, err := os.Stat(manifestPath); os.IsNotExist(err) { - t.Fatalf("manifest.json not generated for %s (run: make generate-sdl-fixtures)", fixtureName) + missingFixtures = append(missingFixtures, fixtureName+"/manifest.json") } if _, err := os.Stat(groupsPath); os.IsNotExist(err) { - t.Fatalf("groups.json not generated for %s (run: make generate-sdl-fixtures)", fixtureName) + missingFixtures = append(missingFixtures, fixtureName+"/groups.json") } + } + + if len(missingFixtures) > 0 { + t.Fatalf("Missing fixtures (run: make generate-sdl-fixtures): %v", missingFixtures) + } + + for _, entry := range entries { + if !entry.IsDir() { + continue + } + + fixtureName := entry.Name() + fixtureDir := filepath.Join(fixturesDir, fixtureName) + inputPath := filepath.Join(fixtureDir, "input.yaml") + manifestPath := filepath.Join(fixtureDir, "manifest.json") + groupsPath := filepath.Join(fixtureDir, "groups.json") t.Run(fixtureName, func(t *testing.T) {
📜 Review details
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Pro
📒 Files selected for processing (20)
go/sdl/parity_test.go(1 hunks)go/sdl/schema_validator.go(1 hunks)specs/sdl/README.md(1 hunks)testdata/sdl/invalid/count-zero.yaml(1 hunks)testdata/sdl/invalid/empty-image.yaml(1 hunks)testdata/sdl/invalid/invalid-endpoint-name.yaml(1 hunks)testdata/sdl/invalid/invalid-protocol.yaml(1 hunks)testdata/sdl/invalid/ip-without-global.yaml(1 hunks)testdata/sdl/invalid/missing-memory-size.yaml(1 hunks)testdata/sdl/invalid/missing-pricing.yaml(1 hunks)testdata/sdl/invalid/port-too-high.yaml(1 hunks)testdata/sdl/invalid/port-zero.yaml(1 hunks)testdata/sdl/invalid/ram-persistent-true.yaml(1 hunks)testdata/sdl/invalid/unknown-field.yaml(1 hunks)testdata/sdl/schema-only-invalid/email-too-short.yaml(1 hunks)testdata/sdl/schema-only-invalid/invalid-denom.yaml(1 hunks)testdata/sdl/schema-only-invalid/invalid-gpu-vendor.yaml(1 hunks)testdata/sdl/schema-only-invalid/invalid-version.yaml(1 hunks)testdata/sdl/schema-only-invalid/password-too-short.yaml(1 hunks)ts/src/sdl/SDL/SDL.ts(18 hunks)
✅ Files skipped from review due to trivial changes (1)
- testdata/sdl/schema-only-invalid/password-too-short.yaml
🚧 Files skipped from review as they are similar to previous changes (1)
- specs/sdl/README.md
🧰 Additional context used
🧬 Code graph analysis (1)
ts/src/sdl/SDL/SDL.ts (3)
ts/src/sdl/index.ts (1)
SdlValidationError(28-28)ts/src/sdl/types.ts (5)
v2Sdl(93-98)v3Sdl(100-105)v3Manifest(4-4)v3ComputeResources(218-224)v3DeploymentGroup(270-285)ts/src/network/index.ts (1)
NetworkId(2-2)
⏰ Context from checks skipped due to timeout of 90000ms. You can increase the timeout in your CodeRabbit configuration to a maximum of 15 minutes (900000ms). (5)
- GitHub Check: go
- GitHub Check: proto
- GitHub Check: sdl-parity
- GitHub Check: coverage
- GitHub Check: test
🔇 Additional comments (42)
testdata/sdl/invalid/port-too-high.yaml (3)
8-8: Port value correctly triggers validation error.Port 99999 appropriately exceeds the valid TCP port range (1–65535), making this a well-chosen invalid test case. The YAML structure is valid and complete.
1-1: The schema reference path is correct and resolves properly. The filego/sdl/sdl-input.schema.yamlexists at the expected location.
1-33: Verify companion fixture files exist.Per the PR description, each SDL test fixture should include
input.yaml,manifest.json(Go parser output), andgroups.json(Go parser output). Confirm that the correspondingmanifest.jsonandgroups.jsonfiles exist for this test case, or clarify if input-only fixtures are intentional for invalid test cases.Additionally, since the PR notes that input and output fixtures were AI-generated, manually verify or regenerate the expected error outputs in
manifest.jsonandgroups.jsonto ensure they correctly represent how both parsers should reject this invalid port value.#!/bin/bash # Description: Check for companion fixture files for this test case test_case="testdata/sdl/invalid/port-too-high" echo "Checking for fixture files:" ls -la "${test_case}".* 2>/dev/null || echo "No companion files found" ls -la "${test_case}"/ 2>/dev/null || echo "No directory found" # List all files in the invalid directory to understand structure echo -e "\n--- Files in testdata/sdl/invalid/ ---" fd -t f . testdata/sdl/invalid/ | head -20testdata/sdl/invalid/ram-persistent-true.yaml (2)
3-36: YAML structure is valid and correctly represents an invalid SDL scenario.The fixture properly defines a v2.0 SDL with a RAM storage declared as persistent (lines 24–25), which should be rejected by both parsers since RAM storage is inherently ephemeral. The structure aligns with the PR's goal of testing parity in rejecting invalid configurations.
Verify that both Go and TypeScript SDL parsers reject this configuration with consistent error messages. If reference output files (manifest.json, groups.json, or error logs) are expected for this fixture, confirm they are present.
1-1: The schema reference path is correct and the schema file exists at the expected location. The filego/sdl/sdl-input.schema.yamlis reachable from the fixture file using the relative path../../../go/sdl/sdl-input.schema.yaml, and the file is 12K in size with proper permissions.testdata/sdl/invalid/count-zero.yaml (1)
1-33: LGTM!This fixture correctly tests the zero-count deployment validation. The SDL structure is valid except for
count: 0at line 32, which should be rejected by the parser validation added inSDL.ts(lines 128-137).testdata/sdl/schema-only-invalid/invalid-version.yaml (1)
1-33: LGTM!This fixture correctly tests schema validation of unsupported SDL version "3.0". The placement in
schema-only-invaliddirectory indicates the schema should reject this while the Go parser accepts it, which is appropriate for testing schema strictness.testdata/sdl/invalid/invalid-protocol.yaml (1)
1-34: LGTM!This fixture correctly tests rejection of invalid protocol value "HTTP". The
parseServiceProtomethod in SDL.ts only accepts "TCP", "UDP", or undefined, so this should trigger the "ErrUnsupportedServiceProtocol" error.testdata/sdl/invalid/empty-image.yaml (1)
1-33: LGTM!This fixture correctly tests rejection of empty image strings. The
validateServiceImagemethod (lines 377-380) validates that the image is non-empty usingimage.trim().length > 0.testdata/sdl/invalid/missing-memory-size.yaml (1)
1-33: LGTM!This fixture correctly tests rejection of missing memory size. The memory block at line 18-19 only has
attributes: {}without asizefield, which should be caught during resource validation.testdata/sdl/invalid/port-zero.yaml (1)
1-33: LGTM!This fixture correctly tests rejection of port value 0. The
validateServicePortsmethod (lines 382-391) validates thatport > 0 && port <= 65535, so port 0 should be rejected.testdata/sdl/schema-only-invalid/invalid-gpu-vendor.yaml (1)
1-39: LGTM!This fixture tests GPU vendor configuration that the schema should reject while Go accepts. The GPU attributes structure (lines 24-27) uses a specific vendor/model nesting pattern that exercises schema validation boundaries.
ts/src/sdl/SDL/SDL.ts (7)
114-143: LGTM!The
fromStringmethod improvements address previous review feedback:
- Null/undefined check for parsed YAML (line 117) prevents cryptic errors on empty input
- Top-level key validation (lines 121-126) catches typos early
- Deployment count validation (lines 128-137) enforces
count >= 1- Version-based type casting (lines 139-141) now correctly respects the version parameter
267-336: LGTM!The
convertToGoFormathelper correctly handles recursive conversion of TypeScript objects to Go-compatible format. The key renaming logic (lines 294-321) properly handles:
- Snake_case conversions (
signedBy→signed_by,allOf→all_of)- Special cases (
quantity→size,HTTPOptions→httpOptions)- Empty arrays to null for
all_of/any_of(lines 327-330) to match Go output
377-391: LGTM!The new validation methods correctly enforce:
- Non-empty image names with whitespace trimming (line 379)
- Valid port range
0 < port <= 65535(line 387-388)These runtime validations complement the schema validations discussed in past reviews.
1048-1081: LGTM!The endpoint sequence number computation correctly implements Go parity:
- Collects all IPs including duplicates (line 1049 comment)
- Skips global endpoints without IP (lines 1056-1058) matching Go's
to.Global && len(to.IP) == 0- Sorts for stable ordering (line 1070)
- Assigns incremental sequence numbers starting at 1
Note: The duplicate IP overwrite behavior is intentional to match Go (per past review discussion).
1190-1202: LGTM!The GPU handling now always includes GPU in groups output to match Go behavior. The default
{ units: { val: "0" } }(lines 1196-1198) ensures deterministic output when GPU is not specified.
1230-1238: LGTM!The amount formatting with
toFixed(18)(line 1233) ensures Go-compatible precision for pricing amounts. The type checking handles numbers, strings, and undefined values appropriately.
1018-1030: LGTM!The
v3Manifestmethod correctly accepts and propagates theformatparameter, applyingconvertToGoFormatwhenformat === "go"(line 1029).testdata/sdl/schema-only-invalid/invalid-denom.yaml (1)
1-33: LGTM!This fixture correctly tests schema-only validation with an invalid denomination value (
invalidtoken). The SDL structure is complete and will pass Go parsing while being rejected by the schema validator.testdata/sdl/invalid/missing-pricing.yaml (1)
1-31: LGTM!This fixture correctly tests the missing pricing validation scenario. The placement section lacks the required
pricingblock, which should cause both Go parser and schema validation to fail.testdata/sdl/invalid/unknown-field.yaml (1)
1-34: LGTM!This fixture correctly tests rejection of unknown fields at the root level. The
unknown_fieldproperty should trigger validation failure if the schema enforcesadditionalProperties: false.testdata/sdl/invalid/ip-without-global.yaml (1)
1-36: LGTM!This fixture correctly tests the invalid scenario where an IP endpoint is referenced without the required
global: truesetting. The exposure target at line 14 only specifiesip: myipwithout the necessary global flag.testdata/sdl/invalid/invalid-endpoint-name.yaml (1)
1-37: LGTM!This fixture correctly tests endpoint name validation. The endpoint name
Invalid-Nameviolates naming conventions (uppercase letters not typically allowed), and the reference at line 15 ensures the validation covers both definition and usage.testdata/sdl/schema-only-invalid/email-too-short.yaml (1)
1-38: LGTM!This fixture correctly tests schema-only validation for email length constraints. The email
a@bis syntactically valid but likely fails aminLengthrequirement in the schema, while the Go parser should accept it.go/sdl/parity_test.go (8)
16-17: LGTM!The constants for fixture and schema paths are appropriately defined with relative paths from the test package location.
19-27: LGTM!Good use of
sync.Oncefor lazy schema compilation. This avoids redundant compilation across test runs while properly preserving any initialization errors.
69-90: LGTM!The validation flow is well-structured: input schema validation → SDL parsing → manifest/groups schema validation. Using
require.NoErrorappropriately fails fast on validation errors.
92-116: LGTM!Good use of lazy initialization with
sync.Oncefor schema compilation. The marshal-then-validate pattern works correctly with gojsonschema.
118-149: LGTM!The JSON round-trip comparison approach correctly normalizes types for deep equality checking. This ensures the generated output matches the expected fixtures regardless of field ordering.
151-172: LGTM!The test correctly ensures all files in the
invalid/directory causeReadFileto return an error. The directory existence check witht.Skipallows incremental development.
174-201: LGTM!Good test design that validates the distinction between schema-only validation and Go parser acceptance. The assertions at lines 195 and 198 correctly verify that schema rejects but Go accepts these inputs.
203-243: LGTM!The helper functions are well-implemented. The YAML-to-JSON conversion correctly bridges YAML schema definitions with the JSON-only gojsonschema library, and error aggregation provides useful debugging output.
go/sdl/schema_validator.go (9)
1-12: LGTM! Clean imports.The package imports are appropriate for the schema validation functionality, with correct use of the embed package for compile-time schema inclusion.
14-18: Excellent resolution of the previous schema path issue.Using
go:embedeliminates the fragileruntime.Caller()approach that was flagged in earlier reviews. The schema is now reliably embedded at compile time, making this production-safe.
20-26: Well-structured with proper synchronization primitives.The struct correctly uses
sync.Oncefor lazy schema compilation andsync.RWMutexfor thread-safe logger access, addressing the race condition concern from previous reviews.
32-53: Excellent implementation of thread-safe logger configuration.This properly addresses both previous critical concerns:
- Thread-safety is ensured with
sync.RWMutex- Nil logger is gracefully handled by falling back to
noOpLoggerThe documentation clearly explains usage and thread-safety guarantees.
55-62: LGTM! Defensive programming with clear error message.The empty check provides a safety net with an informative error message, though it should never occur since the schema is embedded at compile time.
64-93: LGTM! Proper lazy initialization pattern.The
sync.Onceensures thread-safe, one-time compilation with error caching. The YAML-to-JSON conversion is necessary for compatibility with gojsonschema, and error wrapping follows best practices.
95-105: Perfect implementation with compile-time safety.The no-op logger correctly implements all
log.Loggermethods, and the compile-time assertion on line 105 ensures interface compliance, addressing the previous minor issue.
107-141: LGTM! Clean validation implementation.The validation logic correctly:
- Parses YAML input
- Converts to JSON for gojsonschema compatibility
- Collects and formats all validation errors
- Uses proper error wrapping throughout
143-171: LGTM! Correct mismatch detection with thread-safe logging.The implementation correctly:
- Detects mismatches between schema and Go validation
- Uses
RLockto safely retrieve the logger pointer- Logs only when there's a mismatch (avoiding noise)
- Provides clear, distinct warning messages for each mismatch type
There was a problem hiding this comment.
Actionable comments posted: 1
Caution
Some comments are outside the diff and can’t be posted inline due to platform limitations.
⚠️ Outside diff range comments (2)
ts/src/sdl/SDL/SDL.ts (2)
1210-1305:v3Groups: bug inlocationcheck prevents proper resource merging; endpoint sort lacks comparatorTwo issues in the
v3Groupscore loop:
locationfalsy check breaks merging for first resource indexconst location = group!.boundComputes[placementName][svcdepl.profile]; if (!location) { // create new resource, then: group!.boundComputes[placementName][svcdepl.profile] = group!.dgroup.resources.length - 1; } else { // merge into existing resource }When the first resource for a given
(placementName, profile)is stored at index0,locationwill be0, which is falsy. On the next occurrence of the same(placementName, profile),!locationis stilltrue, so a new resource is created instead of merging with index 0. Only when the index becomes1or higher does theelsebranch ever trigger.This prevents deterministic aggregation of counts/endpoints for the first matching resource and will diverge from the Go behavior that actually reuses the same resource slot.
Use an explicit
undefinedcheck:- const location = group!.boundComputes[placementName][svcdepl.profile]; - - if (!location) { + const location = group!.boundComputes[placementName][svcdepl.profile]; + + if (location === undefined) { // ... - group!.boundComputes[placementName][svcdepl.profile] = group!.dgroup.resources.length - 1; + group!.boundComputes[placementName][svcdepl.profile] = group!.dgroup.resources.length - 1; } else { const endpoints = this.v3ServiceResourceEndpoints(service); - group!.dgroup.resources[location].count += svcdepl.count; - group!.dgroup.resources[location].endpoints = (group!.dgroup.resources[location].endpoints as any).concat(endpoints); - group!.dgroup.resources[location].endpoints.sort(); + group!.dgroup.resources[location].count += svcdepl.count; + const existing = (group!.dgroup.resources[location].endpoints || []) as any[]; + group!.dgroup.resources[location].endpoints = existing.concat(endpoints); + // Optionally: provide an explicit comparator if Go sorts endpoints + // group!.dgroup.resources[location].endpoints.sort((a, b) => + // a.kind - b.kind || a.sequence_number - b.sequence_number + // ); }
- Sorting endpoints without a comparator is a no-op on objects
group!.dgroup.resources[location].endpoints.sort();operates on an array of endpoint objects. Without a comparator, JavaScript sorts by stringifying elements (all are"[object Object]"), which is effectively a no-op for order and may not match Go’s sort semantics if Go sorts endpoints deterministically.If Go does sort endpoints, add an explicit comparator that matches that ordering (e.g., by
kindthensequence_number). If Go doesn’t sort them, the call to.sort()here is misleading and can be removed.
114-152: AlignfromStringand deprecatedvalidatearound empty/invalid YAML handling
fromStringcorrectly guards againstload(yaml)returningundefined(lines 2–4 check!parsed), but the deprecatedvalidatemethod assumesload(yaml)always succeeds and immediately accessesdata.profileswithout a null check (line 40). For empty or whitespace-only YAML,load()returnsundefined, causing aTypeErrorinstead of the structuredSdlValidationErrorthatfromStringproduces.Although
validateis deprecated, consider adding the same guard to prevent confusing error messages:static validate(yaml: string) { console.warn("SDL.validate is deprecated. Use SDL.constructor directly."); - const data = load(yaml) as v3Sdl; + const parsed = load(yaml) as any; + if (!parsed) { + throw new SdlValidationError("SDL invalid: no content"); + } + const data = parsed as v3Sdl; // existing GPU/storage validation… }
🧹 Nitpick comments (2)
ts/src/sdl/SDL/SDL.ts (2)
639-668:quantityvssizekey toggling is correct but duplicates conversion logicSwitching between
"quantity"and"size"based onasStringinserviceResourceMemory/serviceResourceStorageandresourceUnitMemory/resourceUnitStoragealigns with how you later post-process manifests (e.g.,manifestSortedJSONand Go-format output).However, you now have two independent mechanisms converting
"quantity"→"size": the explicitreplaceAllinmanifestSortedJSONand thequantity→sizerule inconvertToGoFormat. To reduce drift, consider centralizing this transformation viaconvertToGoFormat(or a small shared helper) and removing the string-levelreplaceAll, so there is a single source of truth for this mapping.Also applies to: 1100-1130
1048-1081:computeEndpointSequenceNumberssignature vs usageThe implementation now very explicitly matches the Go behavior (including duplicates and skip-logic for global endpoints without IPs), which is good for parity.
Given that you pass
this.data(typed asv2Sdl | v3Sdl) intocomputeEndpointSequenceNumbers, but the parameter is declared assdl: v2Sdl, it may be clearer to widen the type tov2Sdl | v3Sdl(or a shared SDL interface) so the signature reflects actual usage and avoids future confusion when v3-specific fields are added.- computeEndpointSequenceNumbers(sdl: v2Sdl) { + computeEndpointSequenceNumbers(sdl: v2Sdl | v3Sdl) {This is purely a typing / readability tweak; runtime behavior is already correct.
📜 Review details
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Pro
📒 Files selected for processing (1)
ts/src/sdl/SDL/SDL.ts(18 hunks)
⏰ Context from checks skipped due to timeout of 90000ms. You can increase the timeout in your CodeRabbit configuration to a maximum of 15 minutes (900000ms). (5)
- GitHub Check: proto
- GitHub Check: go
- GitHub Check: sdl-parity
- GitHub Check: test
- GitHub Check: coverage
🔇 Additional comments (3)
ts/src/sdl/SDL/SDL.ts (3)
338-353: New service-level validations (image/ports) look goodThe added checks for non-empty
imageand port range0 < port <= 65535integrate cleanly intovalidate()and mirror the schema-level constraints. This should catch a useful class of invalid SDLs early without changing any successful flows.Also applies to: 377-392
1018-1034:v3Manifest/v3Groupsintegration for Go format looks consistentUsing
v3Groups(format)to derive stableserviceIds and then conditionally wrapping the final manifest viaconvertToGoFormatkeeps the ID assignment consistent between TypeScript and Go-formatted outputs. BecauseconvertToGoFormatdoes not alter theidfield, readinggroups[pIdx].resources[sIdx].resource.idcontinues to work even whenformat === "go".No functional issues spotted here.
1171-1201: GPU field is already always-emitted in groups fixtures—no behavior changeThe test fixture at
ts/test/fixtures/sdl/groups-basic-snapshot.jsonalready shows thegpufield present in group resources even when GPU is not configured, withunits.val = "0". This confirms the current behavior is not a new change; thegpufield has been consistently emitted. No downstream code in the codebase treatsgpuas optional in groups output. The concern about changing the TS-facing shape is not applicable here since the shape already includesgpu.
There was a problem hiding this comment.
Actionable comments posted: 1
♻️ Duplicate comments (1)
ts/src/sdl/SDL/SDL.ts (1)
1049-1081: computeEndpointSequenceNumbers: behavior matches Go, including duplicatesThe new implementation (skipping global endpoints without IP, collecting all others, sorting, then assigning sequence numbers) is clear and intentionally mirrors the Go behavior per the comments. Note that duplicate IPs will still end up with the highest sequence number assigned to that IP, which is fine for parity but might be surprising semantically; your inline comment documents this adequately.
🧹 Nitpick comments (4)
ts/src/sdl/SDL/SDL.ts (4)
115-152: fromString/validate: stronger YAML handling looks good; consider version consistency checkSwitching to
load(yaml)with the explicit!parsed || !parsed.versionguard and the top-level key allow‑list /deployment.*.count >= 1checks is a solid improvement in early failure behavior and error messages. One follow‑up you may want (not required for this PR) is to assert thatparsed.versionis compatible with theNetworkVersionparameter, so users don’t accidentally pass a 2.0 SDL while asking for"beta3"semantics and only discover it later via more obscure errors.
927-956: Manifest service params: sorted storage keys and null handling are sensibleSorting
params.storagekeys for both v2 and v3 manifest params gives deterministic output, which is important for parity tests, and returningnullfromv3ManifestServiceParamswhenparamsis absent keeps the manifest lean. The additional runtime guard forparams.storagein the mappers is also reasonable defensive code.Also applies to: 991-1006
1018-1034: Manifest/groups format parameterization: API shape is mostly consistentAdding
OutputFormatand threading it throughv3Manifestandv3Groupsmakes the Go‑compatible output path explicit and keeps the default"typescript"behavior backward‑compatible. One thing to keep in mind is thatmanifest()itself doesn’t expose theformatyet (unlikegroups(format)), so callers that need Go‑format manifests must reach forv3Manifest(…, "go")directly; if you want full symmetry later, you might extendmanifest(asString, format?)in a future iteration.Also applies to: 1210-1215
1100-1129: Resource unit key switch and GPU defaulting look intentional for paritySwitching memory/storage units to use
"quantity"whenasStringis true and"size"otherwise, plus always emitting agpuunit (defaulting to"0"when missing), lines up with the described Go JSON shapes for groups. This does mean that in groups,price.amountand GPU units are now consistently stringified, so any existing TS consumers that expected numbers should be treated as acceptingstring | number(or updated) if they aren’t already.Also applies to: 1171-1202
📜 Review details
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Pro
📒 Files selected for processing (1)
ts/src/sdl/SDL/SDL.ts(18 hunks)
⏰ Context from checks skipped due to timeout of 90000ms. You can increase the timeout in your CodeRabbit configuration to a maximum of 15 minutes (900000ms). (5)
- GitHub Check: go
- GitHub Check: proto
- GitHub Check: coverage
- GitHub Check: sdl-parity
- GitHub Check: test
🔇 Additional comments (4)
ts/src/sdl/SDL/SDL.ts (4)
259-265: Constructordatatype widened tov2Sdl | v3Sdlaligns with fromStringUpdating the constructor to accept
v2Sdl | v3Sdlmatches thefromStringcasting and avoids the earlier mismatch between stored data and version-specific methods. This looks correct and keeps the internal API coherent.
267-336: convertToGoFormat: key/shape normalization aligns with Go outputThe recursive normalization (Uint8Array → string, camelCase ↔ snake_case,
quantity→size, emptyall_of/any_of→null) is consistent with the Go JSON expectations and the parity goals of this PR. The implementation is straightforward and guarded againstundefinedvalues, so I don’t see functional issues here.
338-352: New image/port validations integrate cleanly into validate()Wiring
validateServiceImageandvalidateServicePortsintovalidate()adds useful sanity checks (non‑empty image strings and port range enforcement) without changing existing flows, and the failure messages are clear. This should catch a good class of SDL mistakes early while staying consistent with otherSdlValidationErrorchecks.Also applies to: 377-391
1324-1351: v2Groups price formatting and requirements normalization look correctStringifying
pricing.amountto 18‑decimal strings, normalizing attributes into sorted key/value pairs, and explicitly settingsignedBy.allOf/anyOftonullwhen absent should help the v2 groups output line up with the Go fixtures. The shape is consistent with the v3Groups requirements block, which is good for maintainability.
vertex451
force-pushed
the
artem/sdl-parity-tests
branch
from
e6a1969 to
6dd94ad
Compare
There was a problem hiding this comment.
Actionable comments posted: 0
🧹 Nitpick comments (1)
go/sdl/parity_test.go (1)
54-60: Consider usingt.Skipfinstead oft.Fatalffor missing fixtures.Using
t.Fatalfstops all subsequent tests if fixtures are missing. If this is intentional (fixtures should always exist in CI), this is fine. Otherwise,t.Skipfwould allow other tests to continue.if _, err := os.Stat(manifestPath); os.IsNotExist(err) { - t.Fatalf("manifest.json not generated for %s (run: make generate-sdl-fixtures)", fixtureName) + t.Skipf("manifest.json not generated for %s (run: make generate-sdl-fixtures)", fixtureName) } if _, err := os.Stat(groupsPath); os.IsNotExist(err) { - t.Fatalf("groups.json not generated for %s (run: make generate-sdl-fixtures)", fixtureName) + t.Skipf("groups.json not generated for %s (run: make generate-sdl-fixtures)", fixtureName) }
📜 Review details
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Pro
📒 Files selected for processing (20)
go/sdl/parity_test.go(1 hunks)go/sdl/schema_validator.go(1 hunks)specs/sdl/README.md(1 hunks)testdata/sdl/invalid/count-zero.yaml(1 hunks)testdata/sdl/invalid/empty-image.yaml(1 hunks)testdata/sdl/invalid/invalid-endpoint-name.yaml(1 hunks)testdata/sdl/invalid/invalid-protocol.yaml(1 hunks)testdata/sdl/invalid/ip-without-global.yaml(1 hunks)testdata/sdl/invalid/missing-memory-size.yaml(1 hunks)testdata/sdl/invalid/missing-pricing.yaml(1 hunks)testdata/sdl/invalid/port-too-high.yaml(1 hunks)testdata/sdl/invalid/port-zero.yaml(1 hunks)testdata/sdl/invalid/ram-persistent-true.yaml(1 hunks)testdata/sdl/invalid/unknown-field.yaml(1 hunks)testdata/sdl/schema-only-invalid/email-too-short.yaml(1 hunks)testdata/sdl/schema-only-invalid/invalid-denom.yaml(1 hunks)testdata/sdl/schema-only-invalid/invalid-gpu-vendor.yaml(1 hunks)testdata/sdl/schema-only-invalid/invalid-version.yaml(1 hunks)testdata/sdl/schema-only-invalid/password-too-short.yaml(1 hunks)ts/src/sdl/SDL/SDL.ts(18 hunks)
✅ Files skipped from review due to trivial changes (2)
- testdata/sdl/schema-only-invalid/invalid-denom.yaml
- testdata/sdl/schema-only-invalid/invalid-version.yaml
🚧 Files skipped from review as they are similar to previous changes (10)
- testdata/sdl/invalid/ip-without-global.yaml
- testdata/sdl/invalid/count-zero.yaml
- specs/sdl/README.md
- go/sdl/schema_validator.go
- testdata/sdl/invalid/empty-image.yaml
- testdata/sdl/invalid/unknown-field.yaml
- testdata/sdl/schema-only-invalid/password-too-short.yaml
- testdata/sdl/invalid/ram-persistent-true.yaml
- testdata/sdl/invalid/missing-pricing.yaml
- testdata/sdl/schema-only-invalid/invalid-gpu-vendor.yaml
🧰 Additional context used
🧬 Code graph analysis (1)
go/sdl/parity_test.go (2)
go/sdl/sdl.go (1)
ReadFile(83-89)go/testutil/deployment.go (1)
DeploymentGroups(54-62)
⏰ Context from checks skipped due to timeout of 90000ms. You can increase the timeout in your CodeRabbit configuration to a maximum of 15 minutes (900000ms). (4)
- GitHub Check: test
- GitHub Check: sdl-parity
- GitHub Check: coverage
- GitHub Check: go
🔇 Additional comments (23)
testdata/sdl/invalid/invalid-endpoint-name.yaml (1)
1-37: Verify endpoint name validation is the isolated failure point.The YAML structure is valid and well-formed. The fixture is correctly placed in the
invalid/directory and references the endpoint consistently (line 5 definition, line 15 reference). The endpoint nameInvalid-Namewith uppercase characters appears to be the intended invalid element.Please confirm that:
- Endpoint names must use lowercase-only format (reject
Invalid-Namedue to uppercase)- Both the schema validator and Go parser reject this fixture specifically for the endpoint name violation
- No other unintended invalid elements exist in the fixture
testdata/sdl/invalid/port-too-high.yaml (2)
1-1: Schema reference path resolves correctly.The YAML schema reference at line 1 (
../../../go/sdl/sdl-input.schema.yaml) correctly resolves to the schema file located atgo/sdl/sdl-input.schema.yamlin the repository. The relative path navigation fromtestdata/sdl/invalid/up three levels to the repository root and then intogo/sdl/is valid.
8-9: Validate port range constraint is properly defined in schema.Port 99999 exceeds the valid TCP/UDP port range (0–65535), making this a correct invalid test case. The
sdl-input.schema.yamlproperly defines and enforces this constraint withmaximum: 65535on bothportandasfields in the service expose configuration, ensuring this fixture is correctly rejected by schema validation.testdata/sdl/invalid/missing-memory-size.yaml (1)
18-19: LGTM - Correctly tests missing memory size.The fixture properly demonstrates the invalid case with
attributes: {}instead of a requiredsizefield in the memory block. This will validate that both Go and schema parsers reject SDLs missing memory size specification.testdata/sdl/invalid/invalid-protocol.yaml (1)
10-10: LGTM - Correctly tests invalid protocol value.The fixture uses
HTTPas the protocol value, which is invalid since theprotofield expects transport-layer protocols liketcporudp, not application-layer protocols.testdata/sdl/invalid/port-zero.yaml (1)
8-8: LGTM - Correctly tests invalid port zero.Port 0 is invalid for service exposure, and this fixture properly validates that the parsers reject this edge case.
testdata/sdl/schema-only-invalid/email-too-short.yaml (1)
7-11: LGTM - Correctly tests schema-only email validation.The fixture properly demonstrates a case where the schema is stricter than the Go parser. The email
a@bis syntactically valid but fails minimum length requirements in the schema, while the Go parser accepts it.go/sdl/parity_test.go (7)
16-17: Consider using constants consistently for paths.Both constants use relative paths, which is correct for test fixtures. The separation of fixture root from schema root is good for clarity.
19-27: LGTM - Good use of sync.Once for schema caching.The lazy initialization pattern with
sync.Onceis appropriate for expensive schema compilation and works correctly with parallel test execution.
69-116: LGTM - Clean validation flow.The validation sequence (input schema → parse → output schemas) is well-structured, and the
sync.Oncepattern correctly handles schema compilation caching with proper error checking.
118-149: LGTM - Robust fixture comparison.The JSON round-trip comparison pattern correctly normalizes formatting differences between expected and actual outputs. The duplicate SDL parsing (once in
validateSchemas, once here) provides good test isolation.
151-172: LGTM - Correct invalid SDL rejection testing.The test properly verifies that the Go parser rejects all files in the
invalid/directory. The graceful skip when the directory doesn't exist is appropriate.
174-201: LGTM - Correctly tests schema-only validation divergence.The test properly validates that schema validation is stricter than the Go parser for certain edge cases, with clear assertion messages explaining the expected behavior.
203-243: LGTM - Well-implemented helper functions.The YAML-to-JSON conversion for schema loading is necessary since
gojsonschemaexpects JSON. Error wrapping with%wand the collection of validation errors into a descriptive slice are good practices.ts/src/sdl/SDL/SDL.ts (9)
2-2: LGTM! Import refactored and format type added.The switch to the named
loadexport from js-yaml is correct, and the newOutputFormattype properly supports dual-mode output for Go/TypeScript parity.Also applies to: 57-57
114-142: Excellent validation improvements!The enhanced
fromStringmethod now includes:
- Null/undefined guard (addresses past issue)
- Allowed-keys validation to catch typos
- Deployment count validation (must be >= 1)
- Version-aware type casting to v2Sdl or v3Sdl (fixes past type mismatch)
These changes significantly improve input validation and type safety.
267-336: Well-structured Go format converter.The
convertToGoFormatmethod comprehensively handles the TypeScript-to-Go JSON transformation:
- Properly converts empty arrays to
nullfor all_of/any_of (as noted in past review)- Handles key renaming (camelCase ↔ snake_case)
- Recursively processes nested structures
- Preserves GPU fields per Go's behavior
The implementation correctly addresses the earlier comment about
anyOf/any_ofempty-array handling.
377-391: Good addition of focused validation helpers.The new
validateServiceImageandvalidateServicePortsmethods provide clear, early validation of service configuration with descriptive error messages.
940-956: Good deterministic storage parameter handling.The sorted
storageKeysensures consistent output order. The null-check and defensive error on line 948 are appropriate for type safety.
1018-1030: Excellent format parameter integration and past issue resolution.The format parameter is properly threaded through
v3Manifest→v3Groups→convertToGoFormat, enabling dual-mode output. Key improvements:
- Line 1230-1238: Amount formatting with
toFixed(18)ensures Go-compatible precision- Lines 1243-1248: Sorted attributes guarantee deterministic output
- Lines 1257-1259:
signedByfields default tonullfor empty arrays (addresses past comment)- Line 1275:
location === undefinedcheck fixes the critical past issue where index 0 was treated as falsy- Lines 1289-1296: Proper array concatenation replaces the broken
+=operator (addresses past major issue)- Lines 1190-1199: GPU always included with units "0" when absent, matching Go behavior
All referenced past issues have been correctly addressed.
Also applies to: 1210-1212, 1214-1304
1307-1398: v2Groups updated with same deterministic improvements.Amount precision, attribute sorting, and
signedBynull-defaults mirror the v3Groups improvements—good consistency across versions.
640-640: Good format-aware key selection for memory/storage.Dynamically choosing
"quantity"vs"size"based on theasStringparameter ensures output compatibility with Go's expected field names.Also applies to: 653-653, 1102-1102, 1121-1121
1048-1081: No action required—endpoint sequence numbering is implemented correctly.The current implementation matches the Go behavior exactly. When duplicate IPs exist in
expose.todefinitions (e.g., two services sharing the same IP endpoint), the code collects all occurrences, sorts them, and assigns sequence numbers by iterating the sorted list. Duplicate IPs in the collection cause the final map value to be overwritten with the later sequence number—this is intentional, not a bug.The Go test
TestV2Parse_SharedIPvalidates this behavior: it confirms that two services referencing the same IP "meow" both receive the sameSequenceNumber, which is the expected and correct outcome. The comment in the code accurately describes the behavior as "matching Go implementation."Deduplicating with a
Setbefore assignment would change the sequence number assignments and break parity with the Go implementation. The previous review suggestion about deduplication was likely a refactoring consideration, not a correctness requirement—your decision to not apply it was correct.
Parity Testing
./testdata/sdl/contains test scenarios. Each directory has:input.yaml— SDL inputmanifest.json— Generated manifest (from Go parser)groups.json— Generated deployment groups (from Go parser)Tests validate both parsers(Golang and TS) produce identical output from the same input.
Schema Validation
Three schemas validate the SDL pipeline:
sdl-input.schema.yaml— Validates input YAML (required fields, formats, enums, patterns)manifest-output.schema.yaml— Validates manifest JSON outputgroups-output.schema.yaml— Validates groups JSON outputTest Fixtures
testdata/sdl/v2.0/,v2.1/— Valid fixtures for parity teststestdata/sdl/invalid/— Both schema and parsers rejecttestdata/sdl/schema-only-invalid/— Schema rejects, Go accepts (stricter schema rules)IDE validation
IDE supports YAML validation against the schema when creating manifests in the editor.
Important note:
🔧 Purpose of the Change
📌 Related Issues
✅ Checklist