MapColonies / shigola / 7ff6cedd7

44%

Build Type

push

github

Committed by web-flow

Commit Message

refactor(provider): serve PostGIS only, through ST_AsMVT (MAPCO-11487, MAPCO-11488, MAPCO-11490) (#11)

* fix(config): name the offending type in the unknown-provider error

ErrUnknownProviderType.Error passed Name where the format string says type and
Type where it says provider, so a config naming a type this build does not serve
was rejected with the two halves swapped:

    config: invalid type (provider1) for provider hana

That message is the only thing an operator sees -- an unknown type is rejected
by Validate before anything is registered -- and it is about to be the whole of
what three provider removals leave behind, so it has to read the right way
round.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

* refactor(provider): remove the SAP HANA provider

SAP HANA is not a backend this fork serves. The provider, the mvt_hana
placeholder package, the blank import in atlas and the noHanaProvider build tag
go, and the generated build-tag registry is regenerated without them.

The CI half is the part worth reading. The workflow set RUN_HANA_TESTS with a
connection string carrying a username and password in plaintext, pointed at a
third-party SAP HANA Cloud instance inherited from upstream tegola. Those
credentials were committed on purpose upstream, but they are still credentials
in a public tree reaching a host we do not control, and nothing here can rotate
them. Both go, along with the gate in ci/coverage that recorded whether the
suite had run.

github.com/SAP/go-hdb was the provider's only consumer, so go mod tidy drops it
and golang.org/x/exp with it -- hence the vendor churn.

atlas.TestRemovedProviderTypes is what holds this. A provider registers itself
from an init function reached through a blank import in atlas, so a deletion
breaks nothing anywhere else and nothing would notice the type coming back;
config.Validate rejects what is not in provider.Drivers, so absence from that
list is precisely what makes a config na... (continued)

Coverage Stats

57 of 69 new or added lines in 7 files covered. (82.61%)

24 existing lines in 3 files now uncovered.

7889 of 17787 relevant lines covered (44.35%)

39.88 hits per line