Ports frontend/ (Jinja2/FastAPI, ~1180 LOC) to Go with html/template.
No climate math, no DB, no auth here -- every route fetches from the
backend's /content/* API, so this is I/O-bound glue with no hard-porting
wall; the risk was always in reproducing the rendering exactly, not the
language.
Verified with a golden-HTML diff, not just unit tests: both the Python
original and the Go rewrite were run against the same committed fixtures
(frontend/tests/fixtures/) and every one of the 11 routes compared
byte-for-byte. The only surviving differences after that process are
insignificant inter-tag whitespace and one attribute where Go's stricter
escaper HTML-encodes an apostrophe Jinja left literal (functionally
identical in every browser) -- confirmed programmatically by normalizing
whitespace and unescaping before diffing, not by eyeballing.
That process caught defects unit tests alone would have missed, because
map[string]any has no compile-time field check:
- Render-context keys were snake_case throughout (content.py's Jinja
convention, ported verbatim) while the templates -- written
independently -- read PascalCase fields. A missing map key doesn't
error in html/template, it silently renders empty, so this was invisible
in every status code and every "it built" signal: title, meta
description, canonical URL, OpenGraph tags, the homepage's entire ranked
list, and the brand-tag/nav-active state were all blank across every
page. Fixed by renaming every key to match each template's own header
comment (the authoritative per-page field contract) and, where an
API struct's exported fields already matched what a template needed
(contentapi.CityInfo, Crumb, HomeRanked, HubCountry, ...), passing the
struct straight through instead of hand-rewrapping it in a map --
removes a whole layer of future drift risk, not just this instance of it.
- Three pages 500'd outright: `.ToolHref` needed a fully-composed href
string, not the bare "lat,lon" fragment the handlers were building; the
all-time-records table needed the raw contentapi.AllTimeRecords struct,
not a re-wrapped map.
- JSON-LD was being double-encoded: `<script type="application/ld+json">`
is JAVASCRIPT context to html/template's contextual escaper regardless
of the script's `type` attribute, so a template.HTML-typed value placed
there gets re-escaped as a quoted JS string instead of emitted raw --
the entire structured-data payload shipped as a JSON string containing
JSON, which no crawler would parse as the intended object. Needed
template.JS instead, the type that actually means "trusted JS source."
The glossary term page's JSON-LD was simply never built at all (the
Jinja original assembled it inline in the template rather than through
content.py's context dict, and that got lost in translation) -- added.
- html/template silently strips literal HTML comments AND JavaScript
comments from the parsed output (verified in isolation, zero template
actions involved) -- confirmed as real engine behavior, not a bug in
either port, so both need a FuncMap function returning template.HTML /
template.JS respectively to survive parsing rather than a literal
`<!-- -->` or `//` in the template source.
Packaging: multi-stage Go build, final image alpine (not distroless -- the
Swarm stack's env-entrypoint.sh shim needs bash), 187MB -> 22.6MB. Two
defects caught before they reached a host:
- The Swarm stack overrides `entrypoint:` with no `command:`, which drops
the image's own CMD entirely (Docker/Swarm semantics, not merged) --
env-entrypoint.sh then fell through to its hardcoded `exec uvicorn
app:app` fallback, which doesn't exist in this image. Every deploy
would have exited 127. Fixed with an explicit `command:` on the stack's
frontend service, and corrected the shim's stale comment claiming CMD
passes through automatically.
- `COPY --chown=thermograph` resolves the group by NAME at copy time;
Alpine's `adduser -S` with no `-G` doesn't create a same-named group, so
the classic (non-BuildKit) Docker builder -- which this CI runner falls
back to, since it installs plain `docker.io` with no buildx plugin --
failed outright. Fixed with an explicit group and numeric --chown.
Verification: go build/vet/test -race clean across all packages; the
Docker image builds and passes its embedded go test step under both
BuildKit and the classic builder; shellcheck 0 findings on the one script
touched; rebased onto current main (the ERA5 lake stack landed on both
main and dev during this work -- confirmed additive, no overlap with
frontend/daemon).
main took two direct PRs (#19 shell-lint, #21 the thermograph-daemon Go
service) while dev accumulated the ERA5 lake stack; both sides added compose/
stack services and touched the same seams. Resolutions keep both worlds:
- compose + Swarm stack: lake AND daemon are sibling services; backend env
carries THERMOGRAPH_LAKE_URL and THERMOGRAPH_INTERNAL_TOKEN.
- deploy.sh: backend deploys roll (backend lake daemon); all adds frontend.
- requirements.txt: websockets/apscheduler stay removed (moved to the Go
daemon), duckdb stays (the lake role's engine).
Merged tree: backend suite green, both compose configs validate.
The gateway bot and APScheduler were long-lived stateful I/O loops running
inside the async web app under a leader election. They move into a single Go
binary that owns ONLY that I/O -- websocket, RESUME, heartbeat, backoff, timers.
It owns no grading logic. Anything needing data calls back over a new
internal-only surface (/internal/discord/grade, /internal/jobs/*). Grading
depends on polars and the parquet cache; reimplementing it in Go would let the
bot's grades drift from the API's. The grade route returns gateway-ready JSON
and Go relays the bytes verbatim.
The binary ships in the backend image and runs as a second compose service off
the same tag, so the two ends of the /internal/* contract can never skew.
deploy.sh rolls daemon alongside backend -- without that the service would never
be created, since a single-service deploy uses --no-deps. It also probes the
image first and skips the daemon when rolling a tag that predates the binary:
infra tracks main while image tags are env-staged, so a host can legitimately be
asked to roll an older backend image, and creating the service anyway would
leave a container crash-looping on a missing binary.
replicas: 1 with order: stop-first replaces the leader election -- Discord
permits one gateway connection per bot token.
THERMOGRAPH_INTERNAL_TOKEN is optional: both ends derive it from
THERMOGRAPH_AUTH_SECRET via HMAC under a domain-separation label, so this needs
no new vault entry. The derivation is pinned to a shared cross-language test
vector asserted on both sides, so drift fails CI instead of 401ing every call.
Fail closed when neither secret is set.
Improvements over the Python: a close intended for RESUME uses 4000 rather than
1000 (Discord invalidates a session closed 1000, so the old default defeated its
own resume); MESSAGE_CREATE runs on a bounded worker pool; and a malformed HELLO
returns an error rather than a clean reconnect, which would otherwise reset
backoff and hot-loop against the gateway.
365 Python tests pass; Go build/vet/test -race clean; shellcheck 0 findings.
The cutover moved the host checkout's deploy tree to /opt/thermograph/infra/deploy/,
but the stack file's absolute bind sources (db init, env-entrypoint.sh, autoscale.sh)
still pointed at the old /opt/thermograph/deploy/ -- a service update kept the stale
mounts and new tasks failed with 'bind source path does not exist'.
/opt/thermograph becomes a monorepo checkout: deploy.sh gains INFRA_DIR (git
ops at the root, compose work cd'd into infra/), lock/tags/render paths move
under infra/deploy/, image-path defaults become emi/thermograph/backend|
frontend across compose, stack, and the tag-prune. docker-compose.yml pins
name: thermograph -- without it compose run from .../infra derives project
"infra" and recreates the whole stack beside the running one;
deploy-dev.sh pins COMPOSE_PROJECT_NAME=thermograph-dev (env wins over the
file key) to keep LAN dev's separate project.
