Introduces variable-sized AST nodes for method calls (NODE_CALL),
arrays (NODE_ARRAY), and hashes (NODE_HASH). This change improves
memory efficiency by storing elements directly within the AST node,
avoiding an extra layer of pointer indirection for their data.
This is achieved by adding new data structures and functions in both
the parser and the code generator to handle these new node types.
Variable-sized nodes are now enabled by default.
Co-authored-by: Claude <noreply@anthropic.com>
This commit introduces the core infrastructure for variable-sized AST
nodes, designed to improve memory efficiency. The previous fixed-size
nodes are replaced by nodes that can store data inline, such as
strings and integers, reducing pointer indirection and memory overhead.
Key changes include:
- A generic variable-sized node header (`mrb_ast_var_header`).
- A size-class-based memory allocation system for these nodes.
- Implementation of variable-sized nodes for core types: symbols,
strings, integers, and variables (lvar, gvar, ivar, cvar).
- Integration into the parser and code generator, controlled by a
feature flag.
- Centralized and improved type-casting macros for AST nodes.
Co-authored-by: Claude <noreply@anthropic.com>
This implements a memory optimization for AST nodes that stores location
information (lineno, filename_index) only in head nodes rather than in
every node, reducing memory usage for structure nodes.
Key changes:
- Split node types: mrb_ast_node (structure nodes without location),
mrb_ast_head_node (with location info). Sizes are platform-dependent:
8/12 bytes on 32-bit, 16/24 bytes on 64-bit platforms
- Separate allocation: cons() creates structure nodes, cons_head()
creates head nodes with location information
- Node recycling: all nodes are recycled when freed, but only smaller
structure nodes are reused from the free list to maintain type safety
- Updated macro: added headn() for consistent head node casting
- Removed NODE_LINENO macro: eliminated redundant location copying
since head-only optimization already provides adequate location info
- Fixed codegen to properly access location fields via head node casts
This optimization reduces AST memory usage while preserving all
debugging and location information functionality.
Co-authored-by: Claude <noreply@anthropic.com>
Add helper functions to reduce code duplication in codegen load operations:
- gen_load_op1/gen_load_op2: for simple literal load operations following
the pattern "if (!val) return; genop_X(...); push();"
- gen_load_nil: for conditional nil loading with "if (!val) return;" check
- gen_load_lit: for literal loading with push
Refactor 8 literal loading functions (codegen_self, codegen_nil, codegen_true,
codegen_false, codegen_sym, codegen_float, codegen_back_ref, codegen_nth_ref)
and multiple inline nil loading patterns throughout codegen.c.
Each refactored function reduced from 5-8 lines to 2-4 lines while maintaining
identical bytecode generation behavior. All 1730 tests pass.
Co-authored-by: Claude <noreply@anthropic.com>
Extract final complex cases (NODE_OP_ASGN, NODE_MASGN), unify while/until
loop handling, apply early return pattern to reduce indentation, and achieve
complete switch statement consistency.
The original 5000+ line monolithic function is now organized into 60+ focused
functions while preserving all functionality and performance.
Co-Authored-By: Claude <noreply@anthropic.com>
mruby does not provide `begin ... end while cond` that behave at-least-once
loop, like CRuby does. It remains in TODO.md for long time. But finally we have
implemented the behavior.
This commit introduces NODE_BEGIN as a distinct AST node type for
explicit begin...end blocks, separate from NODE_STMTS which represents
general statement sequences. This distinction will be essential for
implementing CRuby-compatible begin...end while/until constructs.
Key changes:
- Added NODE_BEGIN enum in node.h
- Added new_begin() function in parse.y using optimized cons() structure
- Modified begin...end grammar rule to generate NODE_BEGIN nodes
- Added NODE_BEGIN codegen support in codegen.c
- Added NODE_BEGIN to parser dump functionality
NODE_BEGIN uses a simpler cons() structure instead of list2() for
better memory efficiency, as it only contains a single body node.
Co-Authored-By: Claude <noreply@anthropic.com>
Rename NODE_BEGIN to NODE_STMTS to better reflect its purpose as a
container for statement sequences, not specifically begin-end blocks.
This prepares for adding a dedicated node type for explicit begin-end
constructs.
- Rename NODE_BEGIN enum to NODE_STMTS in node.h
- Update all references in parse.y and codegen.c
- Rename new_begin function to new_stmts
Co-Authored-By: Claude <noreply@anthropic.com>
ADDI/SUBI may fall back to method call that may clear block argument
place holder, which may be a live register. So we cannot directly call
ADDI/SUBI over local variables.
OP_LOADI stores an 8 bit integer to a register, so we renamed the
instruction name to describe the behavior more precisely, like
OP_LOADI16 and OP_LOADI32.
Add `OP_NOP` to distinguish `retry` and jump targets while maintaining instruction compatibility.
Ideally, it might be preferable to separate them into `OP_REDO`.
fixed#6439
Recent changes make mrb_irep_remove_lv() used no longer. Removing this
function would not make any compatibility issue, since it's an internal
function.
To avoid confusion with pools in irep, we renamed region-based memory
manager from pool to mempool.
- rename pool.c to mempool.c
- separate mempool.h
- rename all mrb_pool to mrb_mempool
So if someone is using pool.c functions (I suppose no one does though),
they need to rename all `mrb_pool` to `mrb_mempool` and include
`mruby/mempool.h` header at the top.
mrb_pool_value is a structure that represents a value in the irep
literal pool and is unrelated to mrb_pool, which performs region-based
memory management. It has been renamed mrb_irep_pool to avoid confusion.
To mark `MRB_PROC_ORPHAN` we need to keep track of passed block, even
after the assignment to the block argument. And `yield` should use the
original block; #5786, #5791, #6369
Yukihiro "Matz" Matsumoto