Honor quote preference for legacy-quoted scalars

[oguzkilcan]

stringv() method of Representer hardcoded single-quote style whenever a string needed quoting to avoid being misread as an old-style bool, base60 float, or merge key, bypassing the configured QuotePreference entirely. Marshal() always requests WithV3Defaults() (QuoteLegacy) for backward compatibility with the legacy encoder, so every caller of plain Marshal() silently got single-quoted output no matter what preference it asked for.

[Copilot]

Pull request overview

This PR fixes quoting behavior for legacy-quoted scalar strings by ensuring the representer respects the configured QuotePreference when it must force quotes for YAML 1.1-ambiguous values (e.g., old-style bools like off, base60-float-like values like 1:1, and merge-key-like <<). This aligns Marshal() output (which uses WithV3Defaults() / QuoteLegacy for backward compatibility) with the actual legacy quoting expectations instead of always producing single-quoted scalars in these cases.

Changes:

• Update Representer.stringv() to select single vs double quotes based on r.quotePreference when needsQuoting is true.
• Update golden expectations for affected legacy-quoted outputs in testdata/encode.yaml.
• Add explicit encode-opts test coverage for the "off" YAML 1.1 bool-like string across QuoteSingle, QuoteDouble, and QuoteLegacy.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated no comments.

File	Description
testdata/encode.yaml	Updates expected YAML output quoting and adds new quote-preference test cases for YAML 1.1 bool-like strings (off).
internal/libyaml/representer.go	Ensures representer-forced quoting (YAML 1.1 compatibility) honors the configured QuotePreference rather than hardcoding single quotes.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[ccoVeille]

There is something strange to me here.

There is no reason to use the double quote unless we are in v3 compatibility, right?

Or is it our test suite that forces this setting.

What do you think @colinjlacy ?

[ccoVeille]

@oguzkilcan ?

[oguzkilcan]

Sorry @ccoVeille I missed the question mark there, I thought you were expecting @colinjlacy to answer.

Yes, it is the test suite. runEncodeTest uses yaml.Marshal, which always passes WithV3Defaults(), so the preference is QuoteLegacy, and QuoteLegacy means double quotes in the representer.

About v3 compatibility: v3 does not quote << at all, so it cannot tell us single or double here. I kept double because QuoteLegacy already gives it without a special case.

I also added two encode-opts tests for <<, one with single and one with double, so both preferences are now covered.

❯ make test-internal
go test ./internal/...    
ok      go.yaml.in/yaml/v4/internal/libyaml     (cached)
ok      go.yaml.in/yaml/v4/internal/testutil/assert     (cached)
?       go.yaml.in/yaml/v4/internal/testutil/datatest   [no test files]
ALL INTERNAL FILES PASS

[ccoVeille]

OK thank you. I feel like we should move the test suites to v4 default then.

I'm not expecting this to be done in your PR.

We need to discuss it with other maintainers