proposal: do not expose implementation details of yaml schemas

[kt3k]

Currently YAML package exposes implementation details of schemas as public APIs such as Schema class or Type class. These seem too much details to be provided as public API. I propose we should make these APIs private.

[iuioiua]

Can you please clarify which symbols you'd like to make private and which you'd like to keep public?

[kt3k]

Keep public

• stringify
• parse
• parseAll
• StyleVariant
• DumpOption
• ParseOption

Make private

• CORE_SCHEMA
• DEFAULT_SCHEMA
• EXTENDED_SCHEMA
• FAILSAFE_SCHEMA
• JSON_SCHEMA
• KindType
• RepresentFn
• Type

[iuioiua]

Nice! Related thoughts:

1. Is it worth keeping parseAll()? Is there something special about YAML that justifies having the ability to parse multiple YAML documents within a single body of text? Neither TOML or JSON(C) have this ability. If the answer to these questions is "no", I suggest removing parseAll().
2. Is DEFAULT_SCHEMA the correct default schema? According to the YAML spec, CORE_SCHEMA should be the default schema.

WDYT?

[kt3k]

Is there something special about YAML that justifies having the ability to parse multiple YAML documents within a single body of text?

YAML spec defines the special syntax for storing multiple documents in single file. parseAll supports parsing it. ref https://yaml.org/spec/1.2.0/#id2600795

Is DEFAULT_SCHEMA the correct default schema? According to the YAML spec, CORE_SCHEMA should be the default schema.

The default schema is extended from core schema. It seems adding binary, omap, pairs, set, timestamp, and merge types.

I searched the usage of these types in github search, and it seems these types are very common in actual yaml files (!!binary, !!set, << (merge) ). So I think changing the default schema to CORE_SCHEMA would cause a lot of troubles.

[iuioiua]

Ok. Valid points. Thanks for investigating further.