fix: preserve dynamic array index on dotted property paths

[killme2008]

Problem

The recent lexer/parser refactor for a.b.c property access (PR #693) made ExpressionLexer.scanVariable greedily absorb any [...] following a dotted name into the variable lexeme. That lexeme is then resolved via Reflector.fastGetProperty, whose bracket handling only supports integer-literal indices (Integer.valueOf). As a result, expressions that previously worked through the parser's array-access path regressed:

Expression	Before this fix
a.b[i]	NumberFormatException at runtime → "Could not find variable a.b[i]"
a.b[i + 1]	compile-time ExpressionSyntaxErrorException
a.b[0][1]	broken multi-dimensional index
foo.bars[0].name	✅ works (new feature)

Fix

scanVariable now only absorbs a mid-chain integer-literal index into the lexeme — a [digits] group immediately followed by another . (e.g. the [0] in foo.bars[0].name). That is the only case that cannot be expressed through the parser's array-access path.

Every other case — dynamic (a.b[i]), expression (a.b[i + 1]), trailing (a.b[0]) and multi-dimensional (a.b[0][1]) — stops before [ and is handled by the parser's getElement array access, which supports arbitrary index expressions. This restores the old behavior while keeping the new property-chain feature intact.

Position rollback uses StringCharacterIterator.setIndex.

Tests

• ExpressionLexerUnitTest: tokenization tests asserting dynamic / trailing / multi-dimensional indices are not absorbed, and mid-chain literal indices are.
• QuoteVarTest: end-to-end tests for dynamic and multi-dimensional indices on dotted paths.

Full suite passes locally (mvn test): 870 tests, 0 failures.

[Copilot]

Pull request overview

Restores pre-refactor behavior for dotted variables with array access by ensuring ExpressionLexer.scanVariable only absorbs mid-chain integer-literal indices (e.g. foo.bars[0].name) into the variable lexeme, while leaving dynamic/expression/trailing/multi-dimensional indices for the parser’s array-access path.

Changes:

• Refines ExpressionLexer.scanVariable to stop before [ unless it’s a [digits] immediately followed by ..
• Adds a focused lexer helper (tryAbsorbChainedIndex) with rollback via StringCharacterIterator.setIndex.
• Adds/extends tests covering dynamic, trailing, and multi-dimensional indices on dotted paths (lexer-level + end-to-end).

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 1 comment.

File	Description
src/main/java/com/googlecode/aviator/lexer/ExpressionLexer.java	Changes variable scanning to only absorb mid-chain [digits]. indices, preserving parser handling for dynamic/trailing/multi-dim indexing.
src/test/java/com/googlecode/aviator/lexer/ExpressionLexerUnitTest.java	Adds tokenization assertions to ensure indices on dotted vars are (or aren’t) absorbed as intended.
src/test/java/com/googlecode/aviator/test/function/QuoteVarTest.java	Adds end-to-end coverage for dynamic and multi-dimensional indices on dotted paths.

---

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