Version Packages

[github-actions]

This PR was opened by the Changesets release GitHub action. When you're ready to do a release, you can merge this and the packages will be published to npm automatically. If you're not ready to do a release yet, that's fine, whenever you add more changesets to main, this PR will be updated.

Releases

pascal-code-formatter@0.1.0

Minor Changes

• 00944ab: Add formatPascalSource(code, options?): an AST-driven formatter that parses the
  source and pretty-prints the parser's AST into clean, consistently indented Pascal
  text (a string). Because structure comes from the AST rather than string-matching the
  token stream, classification is exact and parentheses are precedence-aware (a + b * c,
  not (a + (b * c))). Verified to round-trip (output re-parses), be idempotent, and
  preserve semantics (the formatted source compiles to identical JS) across the golden
  fixtures. Also exports AstFormatter and the AstFormatOptions type.

  The existing token-stream formatPascalCode is unchanged. This adds a dependency on
  pascal-parser.

Patch Changes

• 3035567: Drop the objects-deep-compare dependency and simplify the indentation tracker. The
  begin/end/var nesting is now an explicit local marker stack instead of a general-purpose
  counterweight structure plus per-token deep-equality — clearer, testable without an
  external dependency, and one fewer package to install. Output is unchanged.

• f5e4ec8: Fix formatPascalCode producing output that its own parser would reject. needWhiteSpace
  had an ad-hoc allow-list with no rule to separate two adjacent word tokens, so control
  keywords glued to their neighbours (for+i, to+5, do+writeln, then+the body).
  Spacing is now driven by a general rule — a space between any two word-like tokens
  (keyword/identifier/number/boolean/string, which also covers the div/mod/and/or/not
  word operators) plus a leading space before control keywords after )/]. Trailing
  whitespace is no longer emitted on a token that ends its line. Added a round-trip test
  harness that renders the formatted lines and re-parses them, covering for/while/if-then/case
  bodies on the same line as their header.

• 1965329: Track source positions through the toolchain.

  • pascal-tokenizer: every emitted token now carries line (1-based), column
    (1-based) and offset (0-based) marking where it starts in the source. Positions
    are computed with a forward-only cursor, so tokenizing stays O(n).
  • pascal-parser: ParseError now reports where the input went wrong — its
    message is suffixed with at line L, column C and its location points at the
    offending token, instead of a bare, position-less message.
  • pascal-code-formatter: tolerates the new token fields (indentation matching now
    compares tokens by structural identity, not deep object equality).

• Updated dependencies [14fe8d3]

• Updated dependencies [a3f86d6]

• Updated dependencies [d8861ac]

• Updated dependencies [1965329]

  • pascal-parser@0.3.0
  • pascal-tokenizer@0.1.0

pascal-js-compiler@0.4.0

Minor Changes

• c043a37: Export a typed CompileError. Unsupported-Pascal errors — var (by-reference)
  parameters, bitwise and/or/not over integers, and wrong-arity builtin calls —
  now throw CompileError instead of a bare Error, so a consumer can catch and tell
  "this program uses an unsupported feature" apart from an internal compiler bug (which
  stays a plain Error). Mirrors ParseError in pascal-parser.

Patch Changes

• 9020b19: Fix silent codegen corruption when two arrays in different scopes share a name. Array
  low-bounds were tracked in a single flat map keyed only by name, so a local array[1..n]
  inside a subprogram overwrote the low-bound of a same-named outer array, making the outer
  array's 1-based index offset wrong (writes/reads landed outside the allocated array). Array
  scope is now snapshotted and restored around each subprogram body, so a homonym array in one
  scope no longer changes the meaning of the same identifier in a sibling or outer scope.
• 7855f89: Reject builtins used in the wrong position with a CompileError instead of emitting a
  call to a nonexistent JS function. A value-returning builtin used as a statement
  (abs(x);) and a statement-only builtin used in an expression (x := writeln(1)) used to
  compile to abs(x); / writeln(1), which throw ReferenceError at runtime. They now fail
  at compile time, consistent with how the compiler already rejects var parameters and
  wrong arity. A user subprogram sharing a builtin's name still shadows it in either position.
• a3f86d6: Declare "type": "commonjs" explicitly in package.json. The packages already ship
  CommonJS; stating it removes Node's module-type detection step for consumers (as flagged
  by publint) and brings these three in line with pascal-code-formatter.
• Updated dependencies [14fe8d3]
• Updated dependencies [a3f86d6]
• Updated dependencies [d8861ac]
• Updated dependencies [1965329]
  • pascal-parser@0.3.0

pascal-parser@0.3.0

Minor Changes

• 14fe8d3: Declaration is now a discriminated union (VariableDeclaration | FunctionDeclaration | ProcedureDeclaration | ConstantDeclaration) instead of one wide interface where every
  field was optional. Each variant is exported and carries exactly the fields it has, so a
  switch (decl.type) narrows to the concrete node with no cast — matching how Statement
  and Expression already work. Consumers that switch on decl.type gain precise field
  types; code that read fields like decl.parameters off a bare Declaration without
  narrowing must now narrow first.

• d8861ac: Statement AST nodes now carry their real source location (location.start = the
  line/column/offset of their first token), stamped in parseStatement. Previously
  every node location was zeroed. This enables consumers to map a statement back to
  its source line — the playground step-debugger being the first such consumer.
  Expression and declaration nodes remain zeroed until they too have a position
  consumer.

• 1965329: Track source positions through the toolchain.

  • pascal-tokenizer: every emitted token now carries line (1-based), column
    (1-based) and offset (0-based) marking where it starts in the source. Positions
    are computed with a forward-only cursor, so tokenizing stays O(n).
  • pascal-parser: ParseError now reports where the input went wrong — its
    message is suffixed with at line L, column C and its location points at the
    offending token, instead of a bare, position-less message.
  • pascal-code-formatter: tolerates the new token fields (indentation matching now
    compares tokens by structural identity, not deep object equality).

Patch Changes

• a3f86d6: Declare "type": "commonjs" explicitly in package.json. The packages already ship
  CommonJS; stating it removes Node's module-type detection step for consumers (as flagged
  by publint) and brings these three in line with pascal-code-formatter.
• Updated dependencies [a3f86d6]
• Updated dependencies [1965329]
  • pascal-tokenizer@0.1.0

pascal-tokenizer@0.1.0

Minor Changes

• 1965329: Track source positions through the toolchain.

  • pascal-tokenizer: every emitted token now carries line (1-based), column
    (1-based) and offset (0-based) marking where it starts in the source. Positions
    are computed with a forward-only cursor, so tokenizing stays O(n).
  • pascal-parser: ParseError now reports where the input went wrong — its
    message is suffixed with at line L, column C and its location points at the
    offending token, instead of a bare, position-less message.
  • pascal-code-formatter: tolerates the new token fields (indentation matching now
    compares tokens by structural identity, not deep object equality).

Patch Changes

• a3f86d6: Declare "type": "commonjs" explicitly in package.json. The packages already ship
  CommonJS; stating it removes Node's module-type detection step for consumers (as flagged
  by publint) and brings these three in line with pascal-code-formatter.