tamada / fauxrest / 30511224395
83%
main: 86%

LAST BUILD BRANCH: ci/force_fmt_in_CI
DEFAULT BRANCH: main
Ran
Jobs 1
Files 9
Run time 1min
Badge
Embed ▾
README BADGES
x

If you need to use a raster PNG badge, change the '.svg' to '.png' in the link

Markdown

Textile

RDoc

HTML

Rst
coverage: 82.571% (+0.5%) from 82.117%
30511224395

Pull #12

github

tamada 
feat!: narrow $derive.type to string and int

$derive.type shipped in 0.0.3 with five variants: string, int, float, bool and
auto. A field-type survey across the datasets then in testdata found 9 int
fields usable as grouping keys, 3 bool fields (all used as $filter predicates,
never as path segments), and zero float fields.

auto also has a verified sharp edge: a value derived from a string field and
inferred as a number stops matching that field under eq, silently emptying the
endpoint. That is the same fail-quietly mode as the $filter bugs in #1 and #2,
and it is the variant that would be most awkward to withdraw once anyone
depends on it.

A derived value becomes a path segment, and string and int are the kinds that
makes sense for. type is an additive enum, so float and bool can return later
without invalidating any configuration that works today — there is no lock-in
cost to leaving them out now.

This is a breaking change, taken deliberately while it is nearly free: 0.0.3
was released two days ago, the crate is absent from crates.io, and the GitHub
release has no meaningful uptake. Waiting makes it expensive.

An unsupported type is rejected when the configuration loads, so a config
written against the wider set fails instead of quietly changing behaviour.
The message is unhelpfully generic because DeriveSource is #[serde(untagged)],
which discards the inner variant error — that affects every $derive mistake,
not just this one, and is left for separate work.

infer_scalar_value is removed along with auto, its only caller. Docs updated
in docs/content/{config,features}.md and both language versions of
.github/docs_for_ai/spec/specifications.md.

The auto-specific integration test is replaced by one that pins both sides of
the narrowed contract: auto/float/bool are rejected, string/int still load.
Verified end to end as well — with a pattern-extracting $derive, int and string
both produce /papers/2024 while the other three exit 1.

Co-Authored-By... (continued)
Pull Request #12: feat!: narrow $derive.type to string and int

1137 of 1377 relevant lines covered (82.57%)

14.13 hits per line

Jobs
ID Job ID Ran Files Coverage
1 30511224395.1 9
82.57
 GitHub Action Run
Source Files on build 30511224395