Replace module-level mutable state with immutable state managed within the Scope
class itself. This eliminates state leakage between parser invocations and makes
the code more functional and predictable.
Changes:
- Add pendingIdentifiers and isInParams as Scope constructor parameters
- Add helper methods: withPendingIdentifiers(), withIsInParams(), clearPending()
- Update hash() to include new state fields
- Convert all mutable state operations to return new Scope instances
- Remove module-level variables entirely
Benefits:
- No state leakage between tests or parser invocations
- Easier to reason about - state is explicit in the context
- More functional programming style with immutable updates
- Eliminates entire class of bugs related to stale module state
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
The scope tracker uses module-level state (pendingIdentifiers) that was not being
cleared after FunctionCall reductions, causing identifier state to leak between
tests. This caused the test 'readme.txt is Word when used in function' to break
the following test by leaving 'echo' in pendingIdentifiers.
- Add FunctionCall to the list of terms that clear pendingIdentifiers
- Un-skip the previously failing test 'readme.txt is Word when used in function'
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Add IdentifierBeforeDot token emitted when identifier immediately precedes '.'
- Move DotGet into @skip {} block using IdentifierBeforeDot
- Prevents 'basename . prop' from parsing as DotGet
- Allows 'basename.prop' to work as expected when identifier is in scope
- Fixes test: 'a word can be contained in parens'
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Fixed tokenizer input.read() to use absolute positions
- Fixed FN_KEYWORD term ID (33 after DotGet added to expression)
- Added DotGet to expression for use as function argument
- All 8 DotGet tests passing
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Aligns naming with ReefVM's DOT_GET opcode and better represents
that this syntax works for both dicts and arrays.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Added 'end' keywords to function definition tests to match
current grammar requirement. Tests now reveal pre-existing
multi-parameter function bug from commit a53db50.
Implement underscore (_) placeholder in pipe expressions to allow
piped values to be placed at specific positions in function calls.
Implementation:
- Scan FunctionCall arguments for underscore placeholder (detected as "_" Word token)
- When underscore is found, replace it with piped value at that position
- When no underscore is found, piped value becomes first arg (existing behavior)
Testing:
- Added test for underscore placeholder with single-parameter function
- Skipped test for multi-parameter function (blocked by existing Shrimp bug)
- All existing pipe tests continue to pass
The underscore detection logic is working correctly, as verified by the
single-parameter test. Full multi-parameter testing is blocked by a
known issue with multi-parameter function calls in Shrimp (see skipped
test in pipe.test.ts).
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Implements Task 7 from pipe expressions plan.
Tests:
- Simple pipe passes result as first argument (5 | double = 10)
- Pipe chain with three stages (3 | add-one | double | square = 64)
- Pipe with bare identifier (get-value | process)
- Pipe in assignment (result = 5 | add-ten)
One test skipped: pipes with multi-parameter functions reveal a broader
Shrimp bug where functions with 2+ parameters don't bind all arguments
correctly. This is not specific to pipes.
Note: Identifiers must be lowercase/kebab-case/emoji. CamelCase not supported.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Implement Task 6 from docs/plans/2025-10-12-pipe-expressions.md
- Add pipe operator (|) termination to tokenizer
- Update grammar to include expressionWithoutIdentifier in pipeOperand
- Add PipeExpr case to compiler switch statement
- Implement pipe compilation: piped value becomes first argument
- Store piped values in temporary __pipe_value variable
- Handle both FunctionCallOrIdentifier and FunctionCall operands
- Add integration tests for pipe expressions
Tests:
- Simple pipe (5 | double) works correctly
- Additional tests exist but have pre-existing issues with function parameters
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Auto-generated parser files from grammar change that added pipe token.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Add "|" token to @tokens section to support pipe operator. This is the
first step in implementing Unix-style pipe expressions for chaining
function calls.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
