feat: read the format keyword as part of the type

[alexander-akait]

Salvages what is still useful from #59 and #68, both of which this supersedes.

What is left in those two

#59 (feat: format object properties with types, issue #42) — the feature it adds landed in #224, so nothing here comes from it. Worth recording why it is not just rebased: it names the property type from a bare schema.properties[x].type, which prints foo: string,number for a union and nothing at all for a property described by anyOf/$ref/const. #224 formats the type through formatInnerSchema and drops it when it is composite or when the shape stops fitting on a line. The two open questions in its thread stay open — breaking long objects over lines (also raised in #42 (comment)) and sorting properties alphabetically.

#68 (feat: humanize format, issue #66) — two separate things. The absolutePath-as-a-format half is breaking and vankop flagged it for discussion; #223 has since covered that reporting as a keyword, so it is dropped. The other half is not on main and is what this PR takes.

The change

A format was reported as a hint sitting next to the constraints of the string:

configuration.strictFormat should be a string (should match format "date", should be < "2016-02-06").

It names the string, so it now reads as part of the type:

configuration.strictFormat should be a date string (should be < "2016-02-06").

src/util/humanize.js turns the name into words, so date-time, snake_case, PascalCase and camelCase formats all read as words.

Two cases #68 did not cover, both of which the existing fixtures caught:

• A format is any string, not only a name. The test schema has format: "[0-9]*", which humanizing renders as [0 9]* string. A format that is not a name is left as a hint.
• A negated format has no type to be part of — should not match format "date" stays a hint.

The format name goes next to the word it names rather than in front of the whole type, so a format: "email" with minLength: 1 is a non-empty email string, not an email non-empty string.

Tests

test/humanize.test.js carries over vankop's word list from #68, plus the empty/one-character and date-time cases. test/hints.test.js gains the composition and negation cases; its existing [0-9]* case covers the not-a-name guard. Two snapshots move, both to the shorter message above.

npm test (503 tests, 338 snapshots), npm run lint and npm run build all pass.

Co-authored with @vankop, whose commit in #68 this is taken from.

🤖 Generated with Claude Code

https://claude.ai/code/session_013wqLmVHkAGBsWQgXknEQCK

---

Generated by Claude Code

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 96.72%. Comparing base (4baf113) to head (e86c29f).
⚠️ Report is 2 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #225      +/-   ##
==========================================
+ Coverage   96.66%   96.72%   +0.06%     
==========================================
  Files           9       10       +1     
  Lines         989     1009      +20     
  Branches      408      414       +6     
==========================================
+ Hits          956      976      +20     
  Misses         31       31              
  Partials        2        2              

Flag	Coverage Δ
integration	96.72% <100.00%> (+0.06%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.