deno-libs / gql / 31175399800

100%

push

github

upgrade graphql to v17 (breaking)

graphql refuses to work with a schema built by a different copy of itself, so
requiring v17 is breaking for every consumer: a v16 schema fails to type-check
against GQLOptions and is rejected at runtime by assertSchema. Hence 5.0.0.
Use gql v4 with graphql v16, v5 with v17 — documented in the README.

Dependencies: graphql 16.10.0 -> 17.0.2, @std/http 0.224.5 -> 1.1.3,
@std/testing 0.225.3 -> 1.0.20, superfetch 3.0.0 -> 4.0.0,
@graphql-tools/schema 10.0.3 -> 10.0.38, oak (deno.land/x) -> jsr:@oak/oak
17.2.0, graphql_tag (deno.land/x) -> jsr:@deno-libs/graphql-tag 0.2.0. The
graphiql render test moved off deno.land/std 0.221.0 and deno.land/x/expect.

graphql-http 1.22.4 still declares `graphql >=0.11 <=16`, but that range simply
predates v17 (see graphql/graphql-http#155 and the PR resolving it, which
widens the range without touching src). All 60 server audits pass on v17.

Fix `deno task test:audit`, which never ran anything: test.exclude filtered
audit_test.ts even when named explicitly, so the whole spec suite was silently
dead. Renamed to audit.ts, which `deno test` does not auto-discover but an
explicit path still runs.

Improve the catch-all error path, which returned a bare 500 with an empty body
and logged the cause only to console:
- respond with {"errors":[{"message":"Internal server error"}]} and a JSON
  content type, keeping details server-side
- apply GQLOptions.headers, which were dropped on this path only
- add a hint when the failure looks like a graphql copy mismatch, the error
  this upgrade is most likely to provoke

Centralize dependencies in a deno.json import map with ^ ranges so consumers
can dedupe rather than being pinned to an exact graphql, which also clears 19
no-import-prefix lint errors. Exclude tests, examples, audit.ts and editor and
CI config from the published package; it was shipping 28 files including
.claude/settings.local.json. Flatten the deprecated fmt.options nesting.

Co-Au... (continued)

35 of 35 branches covered (100.0%)

Branch coverage included in aggregate %.

22 of 22 new or added lines in 3 files covered. (100.0%)

164 of 164 relevant lines covered (100.0%)

10.72 hits per line