Poppy TeamPT
Poppy Team

English

Português

English

Português

Appearance

PT
AipoDynamic languageAbout the projectRepository ↗

Static copy

Copied from docs/en/adp/ADP-004-unicode-identifier-policy.md in https://github.com/poppy-team/aipo-lang (MIT). Pinned to revision ce47c4e070404fc41722ccd8e2b7b808b99062a8, blob cf59b1e3d61f173d4d68e453f7dd4e6f81eedeb2. The source repository remains canonical; this copy is refreshed through a sync pull request, not live.

ADP-004 — Unicode Identifier and Security Policy ​

Status: draft (open questions — characterization only, no restrictions added) Related: crates/aipo-cli/tests/unicode_security.rs (pins current behavior), docs/evidence/P01-G02-*.md, UAX #31, UTS #39 (consulted, not adopted) Authority: subordinate to the Language Reference and the no-invention policy

Verified current behavior (measured, then pinned in tests) ​

  • Identifier characters are Rust char::is_alphabetic (first) and is_alphanumeric (rest), plus _. Consequence: precomposed non-ASCII letters work (é, 中, α, Cyrillic); combining marks, zero-width characters and bidi controls are rejected with AIPO_LEX_UNEXPECTED_TOKEN.
  • The source loader does not NFC-normalize: a decomposed e + U+0301 identifier is rejected instead of composing to é. (String literals are normalized by the lexer per ADP-001 Q5; identifiers are not.)
  • Mixed scripts and confusables are unrestricted: Cyrillic сount and Latin count coexist as distinct bindings with no warning.
  • U+00A0 (no-break space) is not source whitespace: a trailing NBSP is a lexer error, not a skipped separator.
  • Emoji/ZWJ sequences are preserved in strings; len counts code points.

Open questions (all undecided — do NOT change behavior here) ​

  1. Should the loader NFC-normalize (making decomposed identifiers compose), or is rejection the specified behavior?
  2. Are confusable/mixed-script identifiers a risk worth a warning (UTS #39 highly-restrictive/confusable detection), or is unrestricted acceptance the V1 position?
  3. Is U+00A0-as-whitespace rejection intended, or should it join the skippable whitespace set?
  4. Do zero-width/bidi rejections deserve their own diagnostic code, or is AIPO_LEX_UNEXPECTED_TOKEN the right (if generic) signal?
  5. Should string contents face any restriction (they currently face none)?

Non-goals of this ADP ​

  • Adding any restriction, warning, or normalization without a follow-up design decision. These tests characterize; a policy change needs its own goal with canon sponsorship and migration of the corpus/fixtures it affects.