In the compiler, when parsing a variable declaration, the `=` token is now optional. If present, the initializer expression is compiled as before. If absent, the variable is implicitly initialized to `null` via a call to `null(compiler, false)`. This change removes the previous `TODO` comment and the unconditional `consume` of `TOKEN_EQ`.
Two new test files verify the behavior for both global and local variables without initializers, asserting that reading such a variable yields `null`.
Refactor the VM's fiber representation from a stack-allocated `Fiber` struct embedded in `WrenVM` to a full `ObjFiber` object managed by the garbage collector. This includes adding the `OBJ_FIBER` object type enum, a `wrenNewFiber` allocation function, a `markFiber` traversal routine for GC marking, and updating all internal APIs (`DEF_FIBER_NATIVE`, `wrenDebugDumpStack`) to use `ObjFiber*` instead of `Fiber*`. The `Fiber` struct definition and its associated `CallFrame` type are moved from `wren_vm.h` into `wren_value.h`, and the `STACK_SIZE`/`MAX_CALL_FRAMES` constants are relocated accordingly. Additionally, the `vm->fiber` singleton pointer is removed, the `vm->unsupported` sentinel value is eliminated (replaced by returning `NULL_VAL` for invalid numeric operands), and the `wrenInterpret` function is temporarily removed. The `wrenNewClass` function's metaclass pinning order is also corrected to prevent premature GC collection.
The bootstrapping of Object and Class now uses a two-phase approach: first creating a raw class with no metaclass or superclass via wrenNewSingleClass, then explicitly wiring superclass relationships with wrenBindSuperclass. This eliminates the need for special-case handling of null superclasses in the old newClass function and makes the metaclass chain setup explicit in wrenInitializeCore.
Replace the global `io` singleton object with a static `IO` class, moving the `write` native from the instance to the metaclass. Update all benchmark, example, and test files to use `IO.write()` instead of `io.write()`.
Add support for closing over "this" in function closures defined inside methods by naming the receiver local variable "this" in method compilers. Fix upvalue closing to copy the actual value instead of the stack top, and reorder return handling to close upvalues before storing the result. Add test cases for closure over "this", nested closures, and nested classes.
Implement list subscript setter (`[ ]=`) native method in wren_core.c, enabling assignment to list elements via bracket syntax. Extract duplicate parameter count validation logic into a shared `validateNumParameters` function in wren_compiler.c, replacing inline checks in both `parameterList` and `methodCall`. Add comprehensive test suite covering basic assignment, return value semantics, negative indices, and error cases for out-of-range and excessive arguments.
- Add TOKEN_LINE matching after list elements in compiler to allow newlines before commas or closing brackets
- Add tests for list newline handling, EOF after comma/element, and class name inside body
- Remove resolved TODOs for default constructors, inherited fields, metaclass is checks, zero comparison, and bitwise not behavior
Replace the single `WrenReallocateFn` parameter in `wrenNewVM` with a `WrenConfiguration` struct containing `reallocateFn`, `initialHeapSize`, `minHeapSize`, and `heapGrowthPercent` fields. Update the VM to use these values for GC thresholds instead of hardcoded constants, and modify the CLI entry point to pass a configuration with a 100MB initial heap.
Add support for setter methods in the compiler by detecting '=' after a method name in namedCall, compiling the assigned value, and emitting the setter instruction. Includes new test cases for instance setters, static setters, assignment associativity, grouping, operator precedence errors, setter result values, and same-name getter/setter methods. Also updates the super call test with a TODO for super setter calls.
The default heap size for the Wren VM is raised by a factor of 10, from
1024*1024*10 bytes to 1024*1024*100 bytes. This change significantly
reduces garbage collection frequency, resulting in a substantial
performance improvement for the binary_trees benchmark.
Replace time-based statistics (best, mean, median, std_dev) with a score system using inverse time scaling (1000.0 / time). Increase NUM_TRIALS from 7 to 10 for more reliable results. Update comparison logic to use score ratios instead of time ratios, and swap color highlighting to show green for faster scores (>105%) and red for slower scores (<95%).
Add a new test file `test/constructor/infix_class_expression.wren` that verifies
the ability to use an infix operator on a freshly constructed object instance.
The test creates a `Foo` class with a `+` method and checks that `new Foo + "value"`
correctly evaluates to `"Foo value"`. Also includes a TODO comment noting other
constructor expression patterns that should be tested.
Remove the `this new` constructor declaration syntax in favor of a `new` keyword that creates instances directly. Superclass constructors are now invoked via bare `super(args)` instead of `super.new(args)`, eliminating the semantic weirdness of calling a constructor on a partially-constructed instance. Update the compiler to treat `new` as a keyword token, replace the `isMethod` flag with `methodName`/`methodLength` fields, and change the VM's `NEW` opcode to pop the class from the stack before creating the instance. Add a default `new` native method on Object that returns `this`. Update all benchmarks and tests to use the new syntax.
Add detailed TODO notes in wren_compiler.c's method() function to document two unresolved issues: the lack of validation for matching superclass constructors, and the problematic instance context during super() compilation where field access like _someField in super(_someField) expressions could cause undefined behavior since the receiver is still the class at that point.
Refactor the method_call benchmark across Lua, Python, Ruby, and Wren to use proper inheritance with super calls instead of direct state manipulation. This ensures all languages benchmark super method invocations consistently.
Fix a bug in the Wren compiler where constructors weren't being bound correctly by extracting method binding logic into a dedicated `bindMethod` function that properly handles both instance and static methods against the class object.
Optimize if-statement compilation by restructuring jump instruction emission to eliminate unnecessary CODE_JUMP instructions when no else branch exists, reducing bytecode size for conditional blocks.
Update the superclass constructor test to verify inherited field access through super constructor chains, confirming correct field initialization across multiple inheritance levels.
Add support for explicit superclass constructor invocation in subclass constructors using `super.constructorName(args)` syntax before the opening brace. The metaclass inheritance hierarchy now mirrors the regular class chain, with `Class` as the root metaclass superclass instead of `Object`. Includes test cases for multi-level constructor delegation and static method inheritance.
Add isInsideMethod helper that walks the compiler's parent chain to detect
if the current compilation scope is within a method. Use it in super_()
to emit a compile error when 'super' is used at top level or inside a
top-level function, with corresponding test cases for both scenarios.
Replace the separate heap-allocated `Upvalue** upvalues` pointer in `ObjClosure` with a C99 flexible array member `Upvalue* upvalues[]`. This eliminates an extra allocation per closure, reduces memory fragmentation, and simplifies the allocation logic in `wrenNewClosure` by computing the total size as `sizeof(ObjClosure) + sizeof(Upvalue*) * fn->numUpvalues`. The upvalue array is still zero-initialized to prevent GC issues during partial construction.
Migrate all boolean-typed parameters, return values, struct fields, and local variables from `int` (with 0/1 convention) to C99 `bool` (`true`/`false`). Update corresponding comments, macro definitions, and documentation strings to use `true`/`false` terminology instead of "non-zero"/"zero". Add `#include <stdbool.h>` to `src/main.c`, `src/wren_compiler.c`, and `src/wren_value.h`. Remove the stale TODO comment in `src/wren_common.h` that originally proposed this change.
Add four new test files covering variable scope edge cases:
- global_in_initializer: verify global variable can reference itself in initializer
- local_in_initializer: verify local variable referencing itself in initializer produces error
- shadow_and_local: verify shadowed variable correctly resolves to outer scope before inner declaration
- shadow_in_initializer: verify shadowed variable in initializer correctly references outer scope value
Remove "(bob)" author prefix from all TODO comments in benchmark/method_call.wren, include/wren.h, src/main.c, src/wren_common.h, src/wren_compiler.c, src/wren_core.c, src/wren_value.c, src/wren_value.h, and src/wren_vm.c. Update the TODO_PATTERN regex in metrics script to match the new format without author parentheses. Also refactor test file counting in metrics to use os.walk with fnmatch for recursive directory traversal instead of glob.iglob.
The old code subtracted oldSize from totalAllocated during deallocation,
which required knowing the object size at free time. This was unsafe
because it could read a freed object to determine its size. The new
approach tracks bytesAllocated by only adding new allocations and
relying on GC marking to account for freed memory, making it both
correct and faster with less code.
Convert all conditional compilation guards from bare `#ifdef NAN_TAGGING` to `#if WREN_NAN_TAGGING` across wren_common.h, wren_value.c, wren_value.h, and wren_vm.c. Introduce new `#ifndef`-based defaults for WREN_NAN_TAGGING, WREN_COMPUTED_GOTO, WREN_DEBUG_GC_STRESS, and WREN_TRACE_MEMORY in wren_common.h, replacing the old TODO comments and commented-out debug flags. Update preprocessor indentation style for consistency.
Remove the METHOD_CTOR method type and CODE_METHOD_CTOR opcode, replacing them with a new CODE_NEW instruction that creates class instances directly in the bytecode. This simplifies the compiler by emitting CODE_NEW at the start of constructor bodies instead of relying on a separate method dispatch path, and eliminates the redundant METHOD_CTOR enum value from the method lookup logic.
Add CODE_SUPER_0 through CODE_SUPER_16 bytecodes to the VM for invoking inherited methods, along with a `namedCall` compiler helper that parses method names and argument lists after a dot. The interpreter dispatches super calls by walking up the class hierarchy from the receiver's immediate class, skipping methods defined on that class. Includes three test cases covering calling a different superclass method, calling the same method name, and indirectly inherited methods. Also fixes error formatting to handle newline tokens gracefully and adds a TODO note about not deduplicating placeholder constants for super calls.
Reorder the opcode enum values, dispatch table entries, switch-case handlers, and debug dump cases for CODE_LIST and CODE_CLOSURE to appear after CODE_RETURN and before CODE_CLASS, improving interpreter performance by reducing branch prediction misses for the more common instructions that now appear earlier in the switch.
Switch binary_trees max depth from 14 to 12 across all languages and reduce fib argument from 30 to 28, dropping JS entirely since its performance is incomparable. Update expected output patterns in run_bench script to match new tab-free format and use best-time comparison instead of mean.
Reorder the `MethodType` enum values so that `METHOD_NONE` is placed last instead of first,
and correspondingly move the `METHOD_NONE` case in the interpreter's dispatch switch to the
end of the case list. This places the most frequently executed method types (`METHOD_PRIMITIVE`,
`METHOD_FOREIGN`, `METHOD_BLOCK`, `METHOD_CTOR`) earlier in both the enum and the switch,
which can improve branch prediction and instruction cache locality in the hot dispatch loop.
Add a `defaultConstructor` function in the compiler that emits bytecode for a constructor returning the receiver, invoked during class compilation. Reorder `CODE_CLASS`, `CODE_SUBCLASS`, and method opcodes to the end of the interpret loop's jump table and switch cases, moving `CODE_CALL_0` through `CODE_CALL_16` earlier to compensate for branch alignment issues caused by removing code from `CODE_CLASS` and `CODE_CALL` handlers.
Add the TOKEN_SUPER token type to the token enum and register "super" as a reserved keyword in the readName function. Include TOKEN_SUPER in the nextToken skip-newlines switch and mark it as UNUSED in the grammar rules table, preparing the compiler for future super keyword support.
Add a new `--language` / `-l` CLI argument that accepts multiple language
identifiers, and pass the list through to `run_benchmark` so it skips
languages not in the user-specified set. Also remove the short form `-b`
from `--generate-baseline` to avoid conflict with the new `-l` flag.
Add a new test file `test/constructor/default.wren` that validates the Wren language
behavior where a class with no explicit constructors automatically receives a
parameterless `new` constructor. The test creates a `Foo` class with only a
`toString` method, instantiates it via `Foo.new`, and asserts both the type check
(`foo is Foo`) and the string representation (`foo.toString`). Also includes a
TODO comment noting the need for a future test case covering classes that define
explicit constructors.
Create new doc/site/performance.markdown explaining Wren's position in the bytecode interpreter performance bucket and introducing NaN tagging for compact value storage. Also update doc/site/template.html to include a link to the new performance page in the "Other" navigation section.
When a class is defined, its superclass is not known until runtime since class definitions are imperative statements. This adds a new wrenBindMethod function that walks the bytecode of a method after it is bound to a class, adjusting LOAD_FIELD/STORE_FIELD indices to account for inherited fields and patching superclass dispatch targets. The newClass function now correctly accumulates numFields from the superclass chain, and the VM calls wrenBindMethod when binding methods to a class. Includes new test files for inherited field access and the is operator.
Add MAX_METHOD_NAME constant and validation in copyName() to prevent method names exceeding 64 characters. When a method name is too long, the compiler now reports an error and truncates the name to the maximum allowed length. Also update MAX_METHOD_SIGNATURE to be derived from MAX_METHOD_NAME and MAX_PARAMETERS. Add test cases for valid 64-character method names, names that exceed the limit, and a call to an overlong method name.
Add TOKEN_TILDE token type and lexer support for the '~' character in the compiler. Implement the bitwise NOT operation as a native method on the Num class, operating on 32-bit unsigned integers. Include a new OPERATOR macro for prefix/infix operator grammar rules and add comprehensive test cases covering positive, negative, floating-point, and boundary values.
Expand the syntax documentation with a new section explaining Wren's script file format (.wren extension) and top-to-bottom execution model. Replace the previous "Literals" section with a comprehensive list of reserved words, and add a new "Newlines" section covering line break handling in the language.
Add a new case in the `readRawToken` function within `src/wren_compiler.c` to map the semicolon character to the `TOKEN_LINE` token type, enabling semicolons to function as line separators in the Wren language parser.
Include a test file `test/semicolon.wren` that verifies the new behavior by using a semicolon to separate two statements on a single line, with the expected output "ok".
Extend the file scanning loop to also cover header and source files under the `include/` directory by using `itertools.chain` to combine both glob patterns, ensuring metrics are computed for all relevant source locations.
Add closure objects that wrap functions with captured upvalues, enabling
first-class closures that close over local variables from enclosing scopes.
Introduce new bytecode instructions (CODE_CLOSURE, CODE_LOAD_UPVALUE,
CODE_STORE_UPVALUE, CODE_CLOSE_UPVALUE, CODE_RETURN) and a CompilerUpvalue
tracking system with a MAX_UPVALUES limit of 256. Refactor the grammar to
strictly separate statements from expressions, fixing a bug where block
expressions incorrectly popped locals while temporaries remained on the
stack. Rename internal functions (wrenCallFunction, wrenDebugDumpInstruction,
wrenDebugDumpStack) and add Upvalue allocation, closure creation, and
debug dump support for the new instructions.
Change Token struct to store `const char* start` pointer and `int length` instead of integer start/end offsets. Update Lexer struct similarly, replacing `tokenStart` and `currentChar` offsets with direct pointers into the source string. Adjust Local struct's `length` field type from `size_t` to `int` for consistency. Simplify error reporting in `endCompiler` to use `printf` with `%.*s` format specifier and the new pointer/length fields, eliminating manual character-by-character printing loops.
Extract the common bytecode finalization logic into a dedicated endCompiler function that emits CODE_END and handles parent compiler function loading, replacing duplicated inline code. Update VM method definition opcodes to pop function from stack instead of reading from constants, enabling future closure support. Add TODO comments outlining planned closure capture mechanism and upvalue tracking requirements.
The `wrenNewVM` function now uses the provided `reallocateFn` for fiber allocation instead of calling `malloc` directly, and stores the allocator in `vm->reallocate` for later use. The `wrenReallocate` function delegates all memory operations (including freeing) to the stored allocator, removing the fallback to `realloc` and `free`. This ensures consistent memory management through the external allocator, enabling custom allocation strategies like memory pools or tracking.
Introduce a `.nav` sidebar with categorized links (Welcome, Language, Objects, Usage) alongside a `.content` area, replacing the old inline header list. Widen the page container from 640px to 840px, rename `.column` to `.page`, and update the header to show a subtitle instead of a small tagline. Adjust the GitHub ribbon z-index to prevent overlap with the new sidebar.
Extract identifier copying logic from subscript and call functions into a dedicated copyName helper function, reducing code duplication. Rename the method signature length constant to MAX_METHOD_SIGNATURE with updated documentation to clarify it includes arity spaces. Update subscript function to use the new constant name and remove stale TODO comment about handling >10 args.
Add a compile-time check in declareVariable() that emits an error when the number of simultaneous local variables reaches MAX_LOCALS (256), replacing the previous placeholder TODO. The limit is derived from the bytecode encoding constraint where CODE_LOAD_LOCAL and CODE_STORE_LOCAL use a single argument byte. Also move the MAX_LOCALS definition from a TODO comment to a proper #define, increasing the value from 255 to 256. Include four new test files: many_locals.wren (exactly 256 simultaneous locals), many_nonsimultaneous_locals.wren (more than 256 locals across nested scopes), too_many_locals.wren (257 simultaneous locals triggering the error), and too_many_locals_nested.wren (exceeding the limit across nested blocks).
Replace the linked-list Scope structure with a flat Local array indexed by depth, enabling proper shadowing semantics where inner declarations correctly hide outer ones. Remove the PUSH_SCOPE/POP_SCOPE macros and the SymbolTable-based locals tracking, replacing them with a MAX_LOCALS (255) fixed-size array that stores variable name, length, and depth. Add four new test cases covering local/parameter collision, nested block scoping, global shadowing, and local shadowing to validate the new behavior.
The .gitignore file now includes an entry for scratch.wren, a temporary Wren script left at the top level for quick testing purposes. This prevents accidental commits of ad-hoc experimentation files.
Add Lua, Python, Ruby, and Wren implementations of the binary_trees benchmark from the Computer Language Benchmarks Game. Update the benchmark runner to support multiple benchmarks with output validation, and fix the Wren compiler to handle the '\t' escape sequence in string literals.
The method function was incorrectly using the outer `compiler` pointer instead of the inner `methodCompiler` when emitting `CODE_LOAD_LOCAL` and its operand, causing incorrect bytecode generation for constructor receiver loading.
Move the dumpInstruction function and its associated opcode printing logic from wren_vm.c into a dedicated wren_debug.c source file with a corresponding wren_debug.h header. Add the new source file to the Xcode project build configuration.
Add test file for Foo class constructors covering zero, one, and two argument overloads by arity. Verify each overload produces correct output and that new instances return proper type and toString behavior.
Add MAX_PARAMETERS constant set to 16 and extend VM bytecode instructions CODE_CALL_11 through CODE_CALL_16 to support the new limit. Introduce compile-time error messages when parameter or argument count exceeds 16 in function definitions, method declarations, subscript calls, and regular method calls. Add new native fiber call stubs fn_call9 through fn_call16 and register them in the core initialization. Include comprehensive test coverage for 9-16 parameter functions and methods, plus error tests for exceeding the limit in all contexts.
Add a check in the `field()` function to detect when a field is referenced outside of a class context. When `compiler->fields` is NULL (indicating no enclosing class), emit a clear error message "Cannot reference a field outside of a class definition." and assign a dummy field index of 255 to allow continued parsing with minimal cascading errors.
Also includes minor cleanup: remove two stale TODO comments about EOF checking and better error messages in `nextToken()` and `consume()`, and add a new test file `test/field/outside_class.wren` that verifies the error is produced for top-level field assignment.
Replace the fixed-size `MAX_PINNED` array of `Obj*` pointers with a dynamically allocated linked list of `PinnedObj` nodes, removing the 16-object limit and eliminating the `numPinned` counter. Update `pinObj` to accept a caller-provided `PinnedObj` stack variable and link it at the head of the list, and simplify `unpinObj` to pop the head without requiring the object argument. Adjust `wrenNewVM` to initialize the new `pinned` pointer to NULL, and modify `collectGarbage` to iterate the linked list instead of the array.
Rename the `primitives.h` and `primitives.c` files to `wren_core.h` and `wren_core.c` respectively, reflecting the module's purpose of defining built-in classes and native methods. Update all includes, macro names (PRIMITIVE -> NATIVE, DEF_PRIMITIVE -> DEF_NATIVE, etc.), and function calls (wrenLoadCore -> wrenInitializeCore) across the codebase and Xcode project file to match the new naming convention.
Move `allocate`, `initObj`, `newClass`, and related static helpers from `vm.c` into `value.c` with consistent `wren` prefix, and update all call sites in `compiler.c`, `primitives.c`, and `vm.c` to use the new names (`wrenNewFunction`, `wrenNewString`, `wrenNewClass`, `wrenNewInstance`, `wrenNewList`). Remove the old declarations from `vm.h` and add comprehensive documentation comments to `value.h` explaining the NaN-tagging representation and type system.
Add DEF_PRIMITIVE(list_removeAt) to src/primitives.c that removes an element at a given index from an ObjList, shifts remaining elements up, shrinks capacity when excess is detected via LIST_GROW_FACTOR, and returns the removed Value. Register the primitive as "removeAt " on vm->listClass. Create test/list/remove_at.wren covering removal at positive indices 0,1,2, negative indices -1,-2,-3, out-of-bounds cases returning null, and verification that the removed value is returned.
Extract inline list growth logic into `ensureListCapacity` helper and add `validateIndex` function supporting negative indices. Implement `list_clear` primitive that resets count to zero without freeing memory, and `list_insert` primitive that shifts elements right and inserts at validated position. Add comprehensive test suites for both operations covering empty lists, normal/negative indices, and return values.
- Moved unit tests for user authentication into `tests/unit/auth/`
- Relocated integration tests for payment processing to `tests/integration/payments/`
- Grouped API endpoint tests under `tests/api/` with separate files for each resource
- Updated test runner configuration to discover tests in new directory structure
Add the `list_add` primitive to support appending elements to lists with automatic capacity management using `LIST_MIN_CAPACITY` and `LIST_GROW_FACTOR` constants. Rename internal `valuesEqual` and `reallocate` functions to `wrenValuesEqual` and `wrenReallocate` for consistent naming, and introduce `wrenPrintValue` to handle formatted output of all value types including lists. Add a test file `test/list_add.wren` verifying basic add operations and return value behavior.
Move equality and inequality primitives from Bool and Function to a single Object-level implementation using valuesEqual, and add wrenGetClass to return the runtime class of any value. Refactor newSingleClass to inherit methods from superclass and wire metaclass superclass to classClass. Add type-checking tests for all built-in types and class inheritance.
Add a new JavaScript benchmark file `benchmark/fib.js` that implements a recursive Fibonacci function and measures execution time using `process.hrtime()`. Register the JavaScript language entry in `benchmark/run_all` with the `node` interpreter and `.js` extension, enabling JS benchmarks to be included in the full benchmark suite alongside existing languages.
Replace separate allocation, deallocation, and reallocation logic with a single `reallocate` function that handles all memory operations (alloc, grow, shrink, free) and integrates garbage collection triggers based on total allocated memory thresholds. Add `allocate` and `deallocate` convenience wrappers that delegate to `reallocate`.
Introduce a `COMPUTED_GOTOS` preprocessor flag in `src/common.h` that, when defined, replaces the interpreter's traditional `for(;;) switch(instruction)` dispatch loop with a computed-goto dispatch table using GCC/LLVM labels-as-values extension. The new implementation defines a static `dispatchTable[]` array mapping opcodes to their corresponding `&&code_*` labels, and uses `DISPATCH()` macro to `goto` the next handler directly. When `COMPUTED_GOTOS` is not defined, the fallback retains the original `switch`-based loop. This change yields approximately a 5% performance improvement in benchmarks, though optimal GCC performance may require additional flags like `-fno-gcse`.
Add list_subscript primitive that supports integer indexing with negative indices counting from the end, returns null for out-of-bounds access, non-integer indices, and non-numeric argument types. Register the `[ ]` operator on the List class in loadCore and include comprehensive test coverage in test/list_subscript.wren.
Add support for list literals in the compiler with bracket-delimited syntax and comma-separated elements. Introduce the ObjList runtime type with a contiguous elements array, garbage collection marking, and proper memory deallocation. Expose a `List` class in the core library with a `count` method returning the number of elements. Update the grammar table to register `[` as a prefix token for list compilation instead of an infix subscript operator, and reorder opcode enum entries to place CODE_LIST before the load/store group.
Implement the `[]` subscript operator as an infix parser at `PREC_CALL` precedence, compiling to a method call on the receiver. Add a `string_subscript` primitive that handles integer indices (including negative for reverse indexing), returns `null` for out-of-bounds or non-integer arguments, and returns a one-character string. Register the `[ ]` method on the string class and include tests for basic indexing, negative indices, out-of-bounds, wrong types, and non-integer indices.
Add escape sequence handling to the string literal lexer in compiler.c, translating
common escape sequences (\", \\, \n) into their literal character equivalents during
tokenization. Introduce a fixed-size buffer (MAX_STRING 1024) in the Parser struct to
accumulate the unescaped string content, along with a helper function addStringChar()
to append characters. The readString() function now loops over input characters,
breaking on an unescaped double quote, and processing backslash-prefixed escapes via
a switch statement. Remove the previous TODO placeholder comment and the elapsed time
computation simplification in benchmark/fib.lua. Add a new test file
test/string_escapes.wren that verifies the three supported escape sequences, and
remove the now-obsolete TODO about escapes from test/string_literals.wren.
Add TOKEN_FIELD token type and '_' prefix handling in lexer to distinguish field identifiers from regular names. Extend Compiler struct with fields symbol table pointer for tracking enclosing class fields. Implement field() parse function and CODE_LOAD_FIELD/CODE_STORE_FIELD bytecodes in VM. Modify ObjClass to store numFields count and ObjInstance to use flexible array member for field storage. Update GC markInstance to trace all instance fields. Add newClass signature to accept numFields parameter. Include test cases for field default null, multiple fields, object references with GC, and use-before-set ordering.
Add full introductory content to the classes documentation page covering class definition, inheritance, and the Object superclass. Create a new usage documentation page with a TODO placeholder. Update the HTML page title format from "wren {title}" to "Wren - {title}" and add a navigation link to the new usage page.
Extract hardcoded trial count into NUM_TRIALS constant set to 7, collect all timing results into a list, and add graph_results function that prints a horizontal ASCII bar chart scaling each run's times relative to the global maximum across all benchmarks and languages.
Add a benchmark runner script (benchmark/run_all) that executes multiple
fibonacci implementations across languages (Lua, Python, Ruby, Wren) and
computes mean, median, and standard deviation from 7 runs. Update existing
fib benchmarks to loop 5 times at fib(30) with elapsed timing. Introduce
OS.clock primitive in C core and OS class to support timing in Wren.
Adjust number formatting to "%.14g" for consistent output precision.
This commit introduces a LICENSE file containing the full text of the MIT License, attributed to Robert Nystrom (2013), which governs the distribution and use of the Wren software.
Extract repeated two-character token matching logic into a dedicated
twoCharToken function, reducing code duplication in readRawToken.
Remove the unused skipWhitespace helper that only handled single spaces
and was replaced by more robust whitespace handling elsewhere.
Fix indentation in isKeyword and remove stale TODO comment in readName.
Add TOKEN_PIPEPIPE token for '||' syntax, implement or() parsing function with CODE_OR bytecode, and add VM interpreter logic that short-circuits on truthy left operand. Update .gitignore to exclude *.xccheckout files, extend test coverage for falsy behavior in and.wren and if.wren, and add comprehensive or.wren test suite verifying short-circuit semantics and truthiness rules.
The new IS_FALSE macro checks if a Value's type field equals VAL_FALSE,
complementing the existing IS_NULL and IS_NUM macros for type discrimination
without relying on NaN-boxing tag bits.
Add TOKEN_AMPAMP token to the lexer for parsing '&&' operator, introduce PREC_LOGIC precedence level for proper expression parsing, implement the `and()` compiler function that emits CODE_AND with a patchable jump offset, and add runtime support in the VM with IS_FALSE macro for truthiness checking. Include comprehensive test cases in test/and.wren verifying short-circuit behavior and return values.
Introduce Scope struct and PUSH_SCOPE/POP_SCOPE macros to manage nested variable scopes during compilation. Update declareVariable to distinguish global declarations from local ones based on scope presence rather than parent compiler. Add truncateSymbolTable helper to clean up local symbols when exiting a scope. Include new test files verifying scoped variable isolation in if, while, and block constructs.
Add TOKEN_WHILE to the token enum and keyword lookup in the compiler, enabling the parser to recognize 'while' statements. Introduce the CODE_LOOP opcode in the VM for backward jumps, and implement the patchJump helper to resolve forward jump offsets during compilation. Include comprehensive test cases for while loops covering single-expression bodies, block bodies, newline handling, and null result semantics.
Switch Value from a struct with type/num/obj fields to a NaN-tagged uint64_t/double union, enabling faster type checks and unboxed numbers. Refactor all macros and accessors (IS_OBJ, AS_OBJ, AS_NUM, etc.) to work with the new representation. Hoist CallFrame pointer, instruction pointer, and bytecode array into local variables inside the interpret loop, updating them on frame push/pop. Enable NAN_TAGGING define and DEBUG_GC_STRESS in common.h. Update Xcode project to use c99 standard, pedantic warnings, and treat warnings as errors. Add benchmark/fib.txt with timing comparisons showing fib benchmark within 25% of Lua.
Bob Nystrom