The compiler passes are now split into three stages:
* check -- passes that check for various error conditions
* transform -- passes that transform the AST (e.g. to perform
optimizations)
* generate -- passes that are related to code generation
Splitting the passes into stages is important for plugins. For example,
if a plugin wants to add a new optimization pass, it can add it at the
end of the "transform" stage without any knowledge about other passes it
contains. Similarly, if it wants to generate something else than the
default code generator does from the AST, it can just replace all passes
in the "generate" stage by its own one(s).
More generally, the stages make it possible to write plugins that do not
depend on names and actions of specific passes (which I consider
internal and subject of change), just on the definition of stages (which
I consider a public API with to which semver rules apply).
Implements part of GH-106.
The |passes| parameter will allow to pass the list of passes from
|PEG.buildParser|. This will be used by plugins. The old way via setting
the |appliedPassNames| property is removed.
Implements part of GH-106.
This is a complete rewrite of the PEG.js code generator. Its goals are:
1. Allow optimizing the generated parser code for code size as well as
for parsing speed.
2. Prepare ground for future optimizations and big features (like
incremental parsing).
2. Replace the old template-based code-generation system with
something more lightweight and flexible.
4. General code cleanup (structure, style, variable names, ...).
New Architecture
----------------
The new code generator consists of two steps:
* Bytecode generator -- produces bytecode for an abstract virtual
machine
* JavaScript generator -- produces JavaScript code based on the
bytecode
The abstract virtual machine is stack-based. Originally I wanted to make
it register-based, but it turned out that all the code related to it
would be more complex and the bytecode itself would be longer (because
of explicit register specifications in instructions). The only downsides
of the stack-based approach seem to be few small inefficiencies (see
e.g. the |NIP| instruction), which seem to be insignificant.
The new generator allows optimizing for parsing speed or code size (you
can choose using the |optimize| option of the |PEG.buildParser| method
or the --optimize/-o option on the command-line).
When optimizing for size, the JavaScript generator emits the bytecode
together with its constant table and a generic bytecode interpreter.
Because the interpreter is small and the bytecode and constant table
grow only slowly with size of the grammar, the resulting parser is also
small.
When optimizing for speed, the JavaScript generator just compiles the
bytecode into JavaScript. The generated code is relatively efficient, so
the resulting parser is fast.
Internal Identifiers
--------------------
As a small bonus, all internal identifiers visible to user code in the
initializer, actions and predicates are prefixed by |peg$|. This lowers
the chance that identifiers in user code will conflict with the ones
from PEG.js. It also makes using any internals in user code ugly, which
is a good thing. This solves GH-92.
Performance
-----------
The new code generator improved parsing speed and parser code size
significantly. The generated parsers are now:
* 39% faster when optimizing for speed
* 69% smaller when optimizing for size (without minification)
* 31% smaller when optimizing for size (with minification)
(Parsing speed was measured using the |benchmark/run| script. Code size
was measured by generating parsers for examples in the |examples|
directory and adding up the file sizes. Minification was done by |uglify
--ascii| in version 1.3.4.)
Final Note
----------
This is just a beginning! The new code generator lays a foundation upon
which many optimizations and improvements can (and will) be made.
Stay tuned :-)
Before this commit, |PEG.buildParser| always returned a parser object.
The only way to get its source code was to call the |toSource| method on
it. While this method worked for parsers produced by |PEG.buildParser|
directly, it didn't work for parsers instantiated by executing their
source code. In other words, it was unreliable.
This commit remvoes the |toSource| method on generated parsers and
introduces a new |output| option to |PEG.buildParser|. It allows callers
to specify whether they want to get back the parser object
(|options.output === "parser"|) or its source code (|options.output ===
"source"|). This is much better and more reliable API.
Includes:
* Moving the source code from /src to /lib.
* Adding an explicit file list to package.json
* Updating the Makefile.
* Updating the spec and benchmark suites and their READMEs.
Part of a fix for GH-32.
The source code is now in the src directory. The library needs to be
built using "rake", which creates the lib/peg.js file by combining the
source files.
1. |PEG.Compiler| -> |PEG.compiler|
2. |PEG.grammarParser| -> |PEG.parser|
This brings us closer to the desired structure of the PEG object, which
is:
+-PEG
|- parser
+- compiler
|- checks
|- passes
+- emitter
These are the only things (together with the |PEG.buildParser| function
and exceptions) that I want to be publicly accessible -- as extension
points and also for easy testing of PEG.js's components.
Before this change, the start rule was the one named "start" and there
was an option to override that. This is now impossible.
The goal of this change is to contain all information for the parser
generation in the grammar itself.
In the future, some override directive for the start rule (like Bison's
"%start") may be added to the grammar.
Little change in the source grammar now does not change variables in all
the generated code. This is helpful especially when one has the
generated grammar stored in a VCS (this is true e.g. for our
metagrammar).
We want to have the rule parsing functions inside the |parse| method
because we want them to share a common environment. In the future,
initializers will be executed in this enviromnent and thus functions and
variables defined by them will be accessible to the rule parsing
functions.
Moving various private properties from the parser object into the
|parse| method was not strictly necessary, but it was a natural step
after moving the functions.
Labeled expressions lead to more maintainable code and also will allow
certain optimizations (we can ignore results of expressions not passed
to the actions).
This does not speed up the benchmark suite execution statistically
significantly on V8.
Detailed results (benchmark suite totals):
---------------------------------
Test # Before After
---------------------------------
1 28.43 kB/s 28.46 kB/s
2 28.38 kB/s 28.56 kB/s
3 28.22 kB/s 28.58 kB/s
4 28.76 kB/s 28.55 kB/s
5 28.57 kB/s 28.48 kB/s
---------------------------------
Average 28.47 kB/s 28.53 kB/s
---------------------------------
Mozilla/5.0 (X11; U; Linux i686; en-US) AppleWebKit/533.4 (KHTML, like Gecko) Chrome/5.0.375.55 Safari/533.4
This shouldn't have measurable effect on the benchmarks as there are no
proxy rules in the grammars the benchamrk uses. However the effect on
generated parsers' speed should be positive generally.
In most cases, code pattern
x === undefined
was transformed to
typeof(x) === "undefined"
and similarly with |!==|.
In the generated code, the condition was simply made less strict to
avoid performance penalty of string comparison (I don't think JavaScript
VMs optimize this specific pattern to avoid it).
This speeds up the benchmark suite execution by 0.18%, which may just be a
measurement error. (Standrad statistic tests would tell more, but I don't want
to mess with them now.) The code is little bit nicer this way though.
Going further and avoiding |apply| seems to slow thigs down a bit, possibly
because multiple array accesses. I may try improved version without array
accesses (where Action passes the Sequence variable names to save the results
into) sometime later.
Detailed results (benchmark suite totals):
---------------------------------
Test # Before After
---------------------------------
1 29.08 kB/s 28.91 kB/s
2 28.72 kB/s 28.75 kB/s
3 28.78 kB/s 28.88 kB/s
4 28.57 kB/s 28.90 kB/s
5 28.84 kB/s 28.81 kB/s
---------------------------------
Average 28.80 kB/s 28.85 kB/s
---------------------------------
Mozilla/5.0 (X11; U; Linux i686; en-US) AppleWebKit/533.2 (KHTML, like Gecko) Chrome/5.0.342.9 Safari/533.2
The action now computes the number of passed parameters during the code
generation and the parameters are declared directly as $1, $2, etc. in the
generated function.
This does not speed up the benchmark suite execution statistically significantly
on V8.
Detailed results (benchmark suite totals):
---------------------------------
Test # Before After
---------------------------------
1 28.68 kB/s 29.08 kB/s
2 28.77 kB/s 28.72 kB/s
3 28.89 kB/s 28.78 kB/s
4 28.84 kB/s 28.57 kB/s
5 28.86 kB/s 28.84 kB/s
---------------------------------
Average 28.81 kB/s 28.80 kB/s
---------------------------------
Mozilla/5.0 (X11; U; Linux i686; en-US) AppleWebKit/533.2 (KHTML, like Gecko) Chrome/5.0.342.9 Safari/533.2
This does not speed up the benchmark suite execution statistically significantly
on V8.
Detailed results (benchmark suite totals):
---------------------------------
Test # Before After
---------------------------------
1 28.84 kB/s 28.75 kB/s
2 28.76 kB/s 28.69 kB/s
3 28.72 kB/s 28.69 kB/s
4 28.84 kB/s 28.93 kB/s
5 28.82 kB/s 28.70 kB/s
---------------------------------
Average 28.80 kB/s 28.75 kB/s
---------------------------------
Mozilla/5.0 (X11; U; Linux i686; en-US) AppleWebKit/533.2 (KHTML, like Gecko)
Chrome/5.0.342.9 Safari/533.2
This does not speed up the benchmark suite execution statistically significantly
on V8.
Detailed results (benchmark suite totals):
---------------------------------
Test # Before After
---------------------------------
1 28.72 kB/s 28.84 kB/s
2 28.84 kB/s 28.76 kB/s
3 28.83 kB/s 28.72 kB/s
4 28.81 kB/s 28.84 kB/s
5 28.76 kB/s 28.82 kB/s
---------------------------------
Average 28.79 kB/s 28.80 kB/s
---------------------------------
Mozilla/5.0 (X11; U; Linux i686; en-US) AppleWebKit/533.2 (KHTML, like Gecko) Chrome/5.0.342.9 Safari/533.2
David Majda