feat(tmpl): template renderer + BaseData (bookshelf-qqz.7)

zombor commented

2026-05-20 17:41:46 +00:00

Owner

Summary

Adds internal/tmpl package with Renderer (buffer-first html/template rendering), BaseData struct, and BuildBaseData helper
Adds templates/layouts/base.html (HTML5 skeleton with {{block "content" .}} slot) and templates/pages/home.html
Adds templates/templates.go with //go:embed FS declaration for the production binary
Exports middleware.NewContextWithTraceID so cross-package tests can inject a trace ID without running the full middleware chain
100% statement coverage on internal/tmpl via Ginkgo v2 + fstest.MapFS

Test plan

make test passes (all internal packages green, race detector clean)
make coverage reports 100% on internal/
Layout-missing → error returned, zero bytes written to ResponseWriter
Page-missing → error returned, zero bytes written to ResponseWriter
Execution error (nil pointer in data) → error returned, zero bytes written, Content-Type not set
Happy path: DOCTYPE rendered, Content-Type set to text/html; charset=utf-8
BuildBaseData propagates TraceID from context (present and absent cases)

Closes bead bookshelf-qqz.7 on merge.

## Summary - Adds `internal/tmpl` package with `Renderer` (buffer-first `html/template` rendering), `BaseData` struct, and `BuildBaseData` helper - Adds `templates/layouts/base.html` (HTML5 skeleton with `{{block "content" .}}` slot) and `templates/pages/home.html` - Adds `templates/templates.go` with `//go:embed` FS declaration for the production binary - Exports `middleware.NewContextWithTraceID` so cross-package tests can inject a trace ID without running the full middleware chain - 100% statement coverage on `internal/tmpl` via Ginkgo v2 + `fstest.MapFS` ## Test plan - [ ] `make test` passes (all internal packages green, race detector clean) - [ ] `make coverage` reports 100% on `internal/` - [ ] Layout-missing → error returned, zero bytes written to ResponseWriter - [ ] Page-missing → error returned, zero bytes written to ResponseWriter - [ ] Execution error (nil pointer in data) → error returned, zero bytes written, Content-Type not set - [ ] Happy path: DOCTYPE rendered, Content-Type set to `text/html; charset=utf-8` - [ ] `BuildBaseData` propagates TraceID from context (present and absent cases) Closes bead bookshelf-qqz.7 on merge.

zombor added 1 commit

2026-05-20 17:41:46 +00:00

feat(tmpl): add template Renderer + BaseData (bookshelf-qqz.7) 1a7e062013

- internal/tmpl: NewRenderer(fs.FS), Render(w, r, layout, page, data) with
  buffer-first rendering so parse/execute errors never write partial bytes
- internal/tmpl: BaseData{AssetVersion, TraceID, Title} + BuildBaseData helper
  reading trace ID via middleware.TraceIDFromContext
- internal/middleware: export NewContextWithTraceID for use in cross-package tests
- templates/layouts/base.html: full HTML5 skeleton with {{block "content" .}} slot
- templates/pages/home.html: defines the content block with welcome message
- templates/templates.go: //go:embed FS declaration for production binary
- 100% test coverage on internal/tmpl (Ginkgo v2 + fstest.MapFS)

zombor force-pushed bd-bookshelf-qqz.7 from 1a7e062013 to a96a7b6aa1

/ Test (pull_request) Failing after 2s

Details

/ Coverage (pull_request) Has been skipped

Details

/ Lint (pull_request) Successful in 1m11s

Details

2026-05-20 17:51:44 +00:00

Compare

zombor referenced this pull request from a commit

2026-05-20 18:01:39 +00:00

Add code-style rules: var-at-top + functional args over interfaces

zombor referenced this pull request

2026-05-20 18:01:49 +00:00

Add code-style rules: var-at-top + functional args #10

zombor added 1 commit

2026-05-20 18:03:19 +00:00

ci: drop fixed host port on MySQL service container

/ Lint (pull_request) Successful in 1m2s

Details

/ Test (pull_request) Successful in 1m32s

Details

/ Coverage (pull_request) Failing after 1m2s

Details

3fc1d87af0

Two concurrent PR runs both tried to bind 0.0.0.0:3307 and the second one
failed in "Set up job" with "Bind for 0.0.0.0:3307 failed: port is already
allocated". Drop the host port mapping and have the test step talk to the
service container by its network alias (mysql:3306) so each run gets its own
isolated MySQL on its own workflow network.

Discovered while shipping bookshelf-qqz.7 — same failure mode hit every open
PR (qqz.4, qqz.7, qqz.8) while the main-branch runs (which never run
concurrently with each other) stayed green.

zombor added 1 commit

2026-05-20 18:08:07 +00:00

ci(coverage): force fresh test runs via -count=1

/ Lint (pull_request) Successful in 1m4s

Details

/ Test (pull_request) Successful in 1m36s

Details

/ Coverage (pull_request) Successful in 1m22s

Details

491913343f

The Coverage job in CI ran `go test -coverprofile=... -coverpkg=./internal/...`
after the Test job in the same workflow had already populated the test-result
cache without `-coverpkg`. Several packages then returned cached results whose
profiles were missing cross-package coverage data, and the gate reported 95.9%
even though every internal package is fully covered.

-count=1 disables the test result cache for this invocation so the
coverprofile is always regenerated against the requested -coverpkg set.

Discovered while shipping bookshelf-qqz.7 — local `make coverage` consistently
passes at 100%, but the first PR run on a fixed-MySQL CI dropped to 95.9%
because of cache reuse. The build cache is not the problem; the test-result
cache is.

zombor referenced this pull request from a commit

2026-05-20 18:14:03 +00:00

Add code-style rules: var-at-top + functional args over interfaces

zombor force-pushed bd-bookshelf-qqz.7 from 491913343f

/ Lint (pull_request) Successful in 1m4s

Details

/ Test (pull_request) Successful in 1m36s

Details

/ Coverage (pull_request) Successful in 1m22s

Details

to e637ce1ec9

/ Lint (pull_request) Successful in 53s