We face `[BUG] unexpected rb_parser_ary_data_type (0) for script lines`
on master branch recently.
This commit changes `enum rb_parser_ary_data_type` to start with `1`
and `0` to be invalid then it makes clear `rb_parser_ary_data_type (0)`
is not intentional.
This patch optimizes forwarding callers and callees. It only optimizes methods that only take `...` as their parameter, and then pass `...` to other calls.
Calls it optimizes look like this:
```ruby
def bar(a) = a
def foo(...) = bar(...) # optimized
foo(123)
```
```ruby
def bar(a) = a
def foo(...) = bar(1, 2, ...) # optimized
foo(123)
```
```ruby
def bar(*a) = a
def foo(...)
list = [1, 2]
bar(*list, ...) # optimized
end
foo(123)
```
All variants of the above but using `super` are also optimized, including a bare super like this:
```ruby
def foo(...)
super
end
```
This patch eliminates intermediate allocations made when calling methods that accept `...`.
We can observe allocation elimination like this:
```ruby
def m
x = GC.stat(:total_allocated_objects)
yield
GC.stat(:total_allocated_objects) - x
end
def bar(a) = a
def foo(...) = bar(...)
def test
m { foo(123) }
end
test
p test # allocates 1 object on master, but 0 objects with this patch
```
```ruby
def bar(a, b:) = a + b
def foo(...) = bar(...)
def test
m { foo(1, b: 2) }
end
test
p test # allocates 2 objects on master, but 0 objects with this patch
```
How does it work?
-----------------
This patch works by using a dynamic stack size when passing forwarded parameters to callees.
The caller's info object (known as the "CI") contains the stack size of the
parameters, so we pass the CI object itself as a parameter to the callee.
When forwarding parameters, the forwarding ISeq uses the caller's CI to determine how much stack to copy, then copies the caller's stack before calling the callee.
The CI at the forwarded call site is adjusted using information from the caller's CI.
I think this description is kind of confusing, so let's walk through an example with code.
```ruby
def delegatee(a, b) = a + b
def delegator(...)
delegatee(...) # CI2 (FORWARDING)
end
def caller
delegator(1, 2) # CI1 (argc: 2)
end
```
Before we call the delegator method, the stack looks like this:
```
Executing Line | Code | Stack
---------------+---------------------------------------+--------
1| def delegatee(a, b) = a + b | self
2| | 1
3| def delegator(...) | 2
4| # |
5| delegatee(...) # CI2 (FORWARDING) |
6| end |
7| |
8| def caller |
-> 9| delegator(1, 2) # CI1 (argc: 2) |
10| end |
```
The ISeq for `delegator` is tagged as "forwardable", so when `caller` calls in
to `delegator`, it writes `CI1` on to the stack as a local variable for the
`delegator` method. The `delegator` method has a special local called `...`
that holds the caller's CI object.
Here is the ISeq disasm fo `delegator`:
```
== disasm: #<ISeq:delegator@-e:1 (1,0)-(1,39)>
local table (size: 1, argc: 0 [opts: 0, rest: -1, post: 0, block: -1, kw: -1@-1, kwrest: -1])
[ 1] "..."@0
0000 putself ( 1)[LiCa]
0001 getlocal_WC_0 "..."@0
0003 send <calldata!mid:delegatee, argc:0, FCALL|FORWARDING>, nil
0006 leave [Re]
```
The local called `...` will contain the caller's CI: CI1.
Here is the stack when we enter `delegator`:
```
Executing Line | Code | Stack
---------------+---------------------------------------+--------
1| def delegatee(a, b) = a + b | self
2| | 1
3| def delegator(...) | 2
-> 4| # | CI1 (argc: 2)
5| delegatee(...) # CI2 (FORWARDING) | cref_or_me
6| end | specval
7| | type
8| def caller |
9| delegator(1, 2) # CI1 (argc: 2) |
10| end |
```
The CI at `delegatee` on line 5 is tagged as "FORWARDING", so it knows to
memcopy the caller's stack before calling `delegatee`. In this case, it will
memcopy self, 1, and 2 to the stack before calling `delegatee`. It knows how much
memory to copy from the caller because `CI1` contains stack size information
(argc: 2).
Before executing the `send` instruction, we push `...` on the stack. The
`send` instruction pops `...`, and because it is tagged with `FORWARDING`, it
knows to memcopy (using the information in the CI it just popped):
```
== disasm: #<ISeq:delegator@-e:1 (1,0)-(1,39)>
local table (size: 1, argc: 0 [opts: 0, rest: -1, post: 0, block: -1, kw: -1@-1, kwrest: -1])
[ 1] "..."@0
0000 putself ( 1)[LiCa]
0001 getlocal_WC_0 "..."@0
0003 send <calldata!mid:delegatee, argc:0, FCALL|FORWARDING>, nil
0006 leave [Re]
```
Instruction 001 puts the caller's CI on the stack. `send` is tagged with
FORWARDING, so it reads the CI and _copies_ the callers stack to this stack:
```
Executing Line | Code | Stack
---------------+---------------------------------------+--------
1| def delegatee(a, b) = a + b | self
2| | 1
3| def delegator(...) | 2
4| # | CI1 (argc: 2)
-> 5| delegatee(...) # CI2 (FORWARDING) | cref_or_me
6| end | specval
7| | type
8| def caller | self
9| delegator(1, 2) # CI1 (argc: 2) | 1
10| end | 2
```
The "FORWARDING" call site combines information from CI1 with CI2 in order
to support passing other values in addition to the `...` value, as well as
perfectly forward splat args, kwargs, etc.
Since we're able to copy the stack from `caller` in to `delegator`'s stack, we
can avoid allocating objects.
I want to do this to eliminate object allocations for delegate methods.
My long term goal is to implement `Class#new` in Ruby and it uses `...`.
I was able to implement `Class#new` in Ruby
[here](https://github.com/ruby/ruby/pull/9289).
If we adopt the technique in this patch, then we can optimize allocating
objects that take keyword parameters for `initialize`.
For example, this code will allocate 2 objects: one for `SomeObject`, and one
for the kwargs:
```ruby
SomeObject.new(foo: 1)
```
If we combine this technique, plus implement `Class#new` in Ruby, then we can
reduce allocations for this common operation.
Co-Authored-By: John Hawthorn <john@hawthorn.email>
Co-Authored-By: Alan Wu <XrXr@users.noreply.github.com>
Recently, `TestRubyLiteral#test_float` fails randomly.
```
1) Error:
TestRubyLiteral#test_float:
ArgumentError: SyntaxError#path changed: "(eval at /home/chkbuild/chkbuild/tmp/build/20240527T050036Z/ruby/test/ruby/test_literal.rb:642)"->"(eval at /home/chkbuild/chkbuild/tmp/build/20240527T050036Z/ruby/test/ruby/test_literal.rb:642)"
```
https://rubyci.s3.amazonaws.com/s390x/ruby-master/log/20240527T050036Z.fail.html.gz
According to Launchable, the first failure was on Apr 30.
This is just when 528c4501f4 was
committed. I don't know if the change is really the cause, but I want to
revert it once to see if the random failure disappears.
This patch moves `ast->node_buffer->config` to `ast->config` aiming to improve readability and maintainability of the source.
## Background
We could not add the `config` field to the `rb_ast_t *` due to the five-word restriction of the IMEMO object.
But it is now doable by merging https://github.com/ruby/ruby/pull/10618
## About assigning `&rb_global_parser_config` to `ast->config` in `ast_alloc()`
The approach of not setting `ast->config` in `ast_alloc()` means that the client, CRuby in this scenario, that directly calls `ast_alloc()` will be responsible for releasing it if a resource that is passed to AST needs to be released.
However, we have put on hold whether we can guarantee the above so far, thus, this patch looks like that.
```
// ruby_parser.c
static VALUE
ast_alloc(void)
{
rb_ast_t *ast;
VALUE vast = TypedData_Make_Struct(0, rb_ast_t, &ast_data_type, ast);
#ifdef UNIVERSAL_PARSER
ast = (rb_ast_t *)DATA_PTR(vast);
ast->config = &rb_global_parser_config;
#endif
return vast;
}
```
This patch adds `int line_count` field to `rb_ast_body_t` structure.
Instead, we no longer cast `script_lines` to Fixnum.
## Background
Ref https://github.com/ruby/ruby/pull/10618
In the PR above, we have decoupled IMEMO from `rb_ast_t`.
This means we could lift the five-words-restriction of the structure
that forced us to unionize `rb_ast_t *` and `FIXNUM` in one field.
## Relating refactor
- Remove the second parameter of `rb_ruby_ast_new()` function
## Attention
I will remove a code that assigns -1 to line_count, in `rb_binding_add_dynavars()`
of vm.c, because I don't think it is necessary.
But I will make another PR for this so that we can atomically revert
in case I was wrong (See the comment on the code)
This patch removes the `VALUE flags` member from the `rb_ast_t` structure making `rb_ast_t` no longer an IMEMO object.
## Background
We are trying to make the Ruby parser generated from parse.y a universal parser that can be used by other implementations such as mruby.
To achieve this, it is necessary to exclude VALUE and IMEMO from parse.y, AST, and NODE.
## Summary (file by file)
- `rubyparser.h`
- Remove the `VALUE flags` member from `rb_ast_t`
- `ruby_parser.c` and `internal/ruby_parser.h`
- Use TypedData_Make_Struct VALUE which wraps `rb_ast_t` `in ast_alloc()` so that GC can manage it
- You can retrieve `rb_ast_t` from the VALUE by `rb_ruby_ast_data_get()`
- Change the return type of `rb_parser_compile_XXXX()` functions from `rb_ast_t *` to `VALUE`
- rb_ruby_ast_new() which internally `calls ast_alloc()` is to create VALUE vast outside ruby_parser.c
- `iseq.c` and `vm_core.h`
- Amend the first parameter of `rb_iseq_new_XXXX()` functions from `rb_ast_body_t *` to `VALUE`
- This keeps the VALUE of AST on the machine stack to prevent being removed by GC
- `ast.c`
- Almost all change is replacement `rb_ast_t *ast` with `VALUE vast` (sorry for the big diff)
- Fix `node_memsize()`
- Now it includes `rb_ast_local_table_link`, `tokens` and script_lines
- `compile.c`, `load.c`, `node.c`, `parse.y`, `proc.c`, `ruby.c`, `template/prelude.c.tmpl`, `vm.c` and `vm_eval.c`
- Follow-up due to the above changes
- `imemo.{c|h}`
- If an object with `imemo_ast` appears, considers it a bug
Co-authored-by: Nobuyoshi Nakada <nobu@ruby-lang.org>
Refactor parser compile functions to reduce the dependence
on ruby functions.
This commit includes these changes
1. Refactor `gets`, `input` and `gets_` of `parser_params`
Parser needs two different data structure to get next line, function (`gets`) and input data (`input`).
However `gets_` is used for both function (`call`) and input data (`ptr`).
`call` is used for managing general callback function when `rb_ruby_parser_compile_generic` is used.
`ptr` is used for managing the current pointer on String when `parser_compile_string` is used.
This commit changes parser to used only `gets` and `input` then removes `gets_`.
2. Move parser_compile functions and `gets` functions from parse.y to ruby_parser.c
This change reduces the dependence on ruby functions from parser.
3. Change ruby_parser and ripper to take care of `VALUE input` GC mark
Move the responsibility of calling `rb_gc_mark` for `VALUE input` from parser to ruby_parser and ripper.
`input` is arbitrary data pointer from the viewpoint of parser.
4. Introduce rb_parser_compile_array function
Caller of `rb_parser_compile_generic` needs to take care about GC because ruby_parser doesn’t know
about the detail of `lex_gets` and `input`.
Introduce `rb_parser_compile_array` to reduce the complexity of ast.c.
This patch is part of universal parser work.
## Summary
- Decouple VALUE from members below:
- `(struct parser_params *)->debug_lines`
- `(rb_ast_t *)->body.script_lines`
- Instead, they are now `rb_parser_ary_t *`
- They can also be a `(VALUE)FIXNUM` as before to hold line count
- `ISEQ_BODY(iseq)->variable.script_lines` remains VALUE
- In order to do this,
- Add `VALUE script_lines` param to `rb_iseq_new_with_opt()`
- Introduce `rb_parser_build_script_lines_from()` to convert `rb_parser_ary_t *` into `VALUE`
## Other details
- Extend `rb_parser_ary_t *`. It previously could only store `rb_parser_ast_token *`, now can store script_lines, too
- Change tactics of building the top-level `SCRIPT_LINES__` in `yycompile0()`
- Before: While parsing, each line of the script is added to `SCRIPT_LINES__[path]`
- After: After `yyparse(p)`, `SCRIPT_LINES__[path]` will be built from `p->debug_lines`
- Remove the second parameter of `rb_parser_set_script_lines()` to make it simple
- Introduce `script_lines_free()` to be called from `rb_ast_free()` because the GC no longer takes care of the script_lines
- Introduce `rb_parser_string_deep_copy()` in parse.y to maintain script_lines when `rb_ruby_parser_free()` called
- With regard to this, please see *Future tasks* below
## Future tasks
- Decouple IMEMO from `rb_ast_t *`
- This lifts the five-members-restriction of Ruby object,
- So we will be able to move the ownership of the `lex.string_buffer` from parser to AST
- Then we remove `rb_parser_string_deep_copy()` to make the whole thing simple
T_TYPES was needed once Ripper jumbled NODEs and other type
objects. However such hack was already removed.
Therefore don't need to set T_TYPES of NODE.
This commit simplifies warnings for hash keys duplication and when clause duplication,
based on the discussion of https://bugs.ruby-lang.org/issues/20331.
Warnings are reported only when strings are same to ohters.
Under a configuration including `cppflags=-DUNIVERSAL_PARSER`, warnings listed below show in build time:
```
node.c:396:30: warning: initialization left-hand side might be a candidate for a format attribute [-Wsuggest-attribute=format]
396 | bug_report_func rb_bug = ast->node_buffer->config->bug;
| ^~~
```
```
ruby_parser.c:655:21: warning: initialization left-hand side might be a candidate for a format attribute [-Wsuggest-attribute=format]
655 | .compile_warn = rb_compile_warn,
| ^~~~~~~~~~~~~~~
ruby_parser.c:656:24: warning: initialization left-hand side might be a candidate for a format attribute [-Wsuggest-attribute=format]
656 | .compile_warning = rb_compile_warning,
| ^~~~~~~~~~~~~~~~~~
ruby_parser.c:657:12: warning: initialization left-hand side might be a candidate for a format attribute [-Wsuggest-attribute=format]
657 | .bug = rb_bug,
| ^~~~~~
ruby_parser.c:658:14: warning: initialization left-hand side might be a candidate for a format attribute [-Wsuggest-attribute=format]
658 | .fatal = rb_fatal,
| ^~~~~~~~
```
To fix, this patch suggests adding `__attribute__((format(printf, n, m)))` to those function declarations.
- Introduce `rb_parser_ary_t` structure to partly eliminate RArray from parse.y
- In this patch, `parser_params->tokens` and `parser_params->ast->node_buffer->tokens` are now `rb_parser_ary_t *`
- Instead, `ast_node_all_tokens()` internally creates a Ruby Array object from the `rb_parser_ary_t`
- Also, delete `rb_ast_tokens()` and `rb_ast_set_tokens()` in node.c
- Implement `rb_parser_str_escape()`
- This is a port of the `rb_str_escape()` function in string.c
- `rb_parser_str_escape()` does not depend on `VALUE` (RString)
- Instead, it uses `rb_parser_stirng_t *`
- This function works when --dump=y option passed
- Because WIP of the universal parser, similar functions like `rb_parser_tokens_free()` exist in both node.c and parse.y. Refactoring them may be needed in some way in the future
- Although we considered redesigning the structure: `ast->node_buffer->tokens` into `ast->tokens`, we leave it as it is because `rb_ast_t` is an imemo. (We will address it in the future)
Nobuyoshi Nakada
yui-knk
Aaron Patterson
S-H-GAMELINKS
Yusuke Endoh
HASUMI Hitoshi
Peter Zhu