iSharkFly-Open
/
python-peps
mirror of https://github.com/python/peps.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

Swap in Python/compile.txt text; it's more comprehensive. Changed the name of 

the PEP to reflect this.

Will delete Python/compile.txt and use this as the definitive copy.
This commit is contained in:
Brett Cannon 2006-03-02 22:04:09 +00:00
parent 794b1f8f11
commit 03ecd8842d
2 changed files with 497 additions and 100 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
pep-0000.txt
View File

@ -59,7 +59,7 @@ Index by Category
I 291 Backward Compatibility for Standard Library Norwitz I 291 Backward Compatibility for Standard Library Norwitz
I 306 How to Change Python's Grammar Hudson I 306 How to Change Python's Grammar Hudson
I 333 Python Web Server Gateway Interface v1.0 Eby I 333 Python Web Server Gateway Interface v1.0 Eby
I 339 How to Change CPython's Bytecode Cannon I 339 Design of CPython's Compiler Cannon
I 356 Python 2.5 Release Schedule Norwitz, et al I 356 Python 2.5 Release Schedule Norwitz, et al
I 3000 Python 3.0 Plans Kuchling, Cannon I 3000 Python 3.0 Plans Kuchling, Cannon
@ -392,7 +392,7 @@ Numerical Index
SR 336 Make None Callable McClelland SR 336 Make None Callable McClelland
S 337 Logging Usage in the Standard Library Dubner S 337 Logging Usage in the Standard Library Dubner
S 338 Executing modules inside packages with '-m' Coghlan S 338 Executing modules inside packages with '-m' Coghlan
I 339 How to Change CPython's Bytecode Cannon I 339 Design of CPython's Compiler Cannon
SR 340 Anonymous Block Statements GvR SR 340 Anonymous Block Statements GvR
SF 341 Unifying try-except and try-finally Brandl SF 341 Unifying try-except and try-finally Brandl
SF 342 Coroutines via Enhanced Generators GvR, Eby SF 342 Coroutines via Enhanced Generators GvR, Eby

593
pep-0339.txt
View File

@ -1,5 +1,5 @@
PEP: 339 PEP: 339
Title: How to Change CPython's Bytecode Title: Design of the CPython Compiler
Version: $Revision$ Version: $Revision$
Last-Modified: $Date$ Last-Modified: $Date$
Author: Brett Cannon <brett@python.org> Author: Brett Cannon <brett@python.org>
@ -7,104 +7,513 @@ Status: Active
Type: Informational Type: Informational
Content-Type: text/x-rst Content-Type: text/x-rst
Created: 02-Feb-2005 Created: 02-Feb-2005
Post-History: 02-Feb-2005 Post-History:
Abstract Abstract
======== --------
Python source code is compiled down to something called bytecode. This Historically (through 2.4), compilation from source code to bytecode
bytecode must implement enough semantics to perform the actions required by the involved two steps:
Language Reference [#lang_ref]_. As such, knowing how to add, remove, or change
the bytecode is important to do properly when changing the abilities of the
Python language.
This PEP covers how to accomplish this in the CPython implementation of the
language (referred to as simply "Python" for the rest of this PEP).
.. warning:: 1. Parse the source code into a parse tree (Parser/pgen.c)
The guidelines outlined in this PEP apply to Python 2.4 and earlier. 2. Emit bytecode based on the parse tree (Python/compile.c)
Current plans for Python 2.5 will lead to a significant change in how
Python's bytecode is handled. Historically, this is not how a standard compiler works. The usual
This PEP will be updated once these planned changes are committed into steps for compilation are:
CVS.
1. Parse source code into a parse tree (Parser/pgen.c)
2. Transform parse tree into an Abstract Syntax Tree (Python/ast.c)
3. Transform AST into a Control Flow Graph (Python/compile.c)
4. Emit bytecode based on the Control Flow Graph (Python/compile.c)
Starting with Python 2.5, the above steps are now used. This change
was done to simplify compilation by breaking it into three steps.
The purpose of this document is to outline how the lattter three steps
of the process works.
This document does not touch on how parsing works beyond what is needed
to explain what is needed for compilation. It is also not exhaustive
in terms of the how the entire system works. You will most likely need
to read some source to have an exact understanding of all details.
Rationale Parse Trees
========= -----------
While changing Python's bytecode is not a frequent occurence, it still happens. Python's parser is an LL(1) parser mostly based off of the
Having the required steps documented in a single location should make implementation laid out in the Dragon Book [Aho86]_.
experimentation with the bytecode easier since it is not necessarily obvious
what the steps are to change the bytecode.
This PEP, paired with PEP 306 [#PEP-306]_, should provide enough basic The grammar file for Python can be found in Grammar/Grammar with the
guidelines for handling any changes performed to the Python language itself in numeric value of grammar rules are stored in Include/graminit.h. The
terms of syntactic changes that introduce new semantics. numeric values for types of tokens (literal tokens, such as ``:``,
numbers, etc.) are kept in Include/token.h). The parse tree made up of
``node *`` structs (as defined in Include/node.h).
Querying data from the node structs can be done with the following
macros (which are all defined in Include/token.h):
- ``CHILD(node *, int)``
Returns the nth child of the node using zero-offset indexing
- ``RCHILD(node *, int)``
Returns the nth child of the node from the right side; use
negative numbers!
- ``NCH(node *)``
Number of children the node has
- ``STR(node *)``
String representation of the node; e.g., will return ``:`` for a
COLON token
- ``TYPE(node *)``
The type of node as specified in ``Include/graminit.h``
- ``REQ(node *, TYPE)``
Assert that the node is the type that is expected
- ``LINENO(node *)``
retrieve the line number of the source code that led to the
creation of the parse rule; defined in Python/ast.c
To tie all of this example, consider the rule for 'while'::
while_stmt: 'while' test ':' suite ['else' ':' suite]
The node representing this will have ``TYPE(node) == while_stmt`` and
the number of children can be 4 or 7 depending on if there is an 'else'
statement. To access what should be the first ':' and require it be an
actual ':' token, `(REQ(CHILD(node, 2), COLON)``.
Checklist Abstract Syntax Trees (AST)
========= ---------------------------
This is a rough checklist of what files need to change and how they are The abstract syntax tree (AST) is a high-level representation of the
involved with the bytecode. All paths are given from the viewpoint of program structure without the necessity of containing the source code;
``/cvsroot/python/dist/src`` from CVS). This list should not be considered it can be thought of a abstract representation of the source code. The
exhaustive nor to cover all possible situations. specification of the AST nodes is specified using the Zephyr Abstract
Syntax Definition Language (ASDL) [Wang97]_.
- ``Include/opcode.h`` The definition of the AST nodes for Python is found in the file
This include file lists all known opcodes and associates each opcode Parser/Python.asdl .
name with
a unique number. When adding a new opcode it is important to take note
of the ``HAVE_ARGUMENT`` value. This ``#define``'s value specifies the
value at which all opcodes greater than ``HAVE_ARGUMENT`` are expected
to take an argument to the opcode.
- ``Lib/opcode.py`` Each AST node (representing statements, expressions, and several
Lists all of the opcodes and their associated value. Used by the dis specialized types, like list comprehensions and exception handlers) is
module [#dis]_ to map bytecode values to their names. defined by the ASDL. Most definitions in the AST correspond to a
particular source construct, such as an 'if' statement or an attribute
lookup. The definition is independent of its realization in any
particular programming language.
- ``Python/ceval.c`` The following fragment of the Python ASDL construct demonstrates the
Contains the main interpreter loop. Code to handle the evalution of an approach and syntax::
opcode goes here.
- ``Python/compile.c`` module Python
To make sure an opcode is actually used, this file must be altered. {
The emitting of all bytecode occurs here. stmt = FunctionDef(identifier name, arguments args, stmt* body,
expr* decorators)
| Return(expr? value) | Yield(expr value)
attributes (int lineno)
}
- ``Lib/compiler/pyassem.py``, ``Lib/compiler/pycodegen.py`` The preceding example describes three different kinds of statements;
The 'compiler' package [#compiler]_ needs to be altered to also reflect function definitions, return statements, and yield statements. All
any changes to the bytecode. three kinds are considered of type stmt as shown by '|' separating the
various kinds. They all take arguments of various kinds and amounts.
- ``Doc/lib/libdis.tex`` Modifiers on the argument type specify the number of values needed; '?'
The documentation [#opcode_list]_ for the dis module contains a complete means it is optional, '*' means 0 or more, no modifier means only one
list of all the opcodes. value for the argument and it is required. FunctionDef, for instance,
takes an identifier for the name, 'arguments' for args, zero or more
stmt arguments for 'body', and zero or more expr arguments for
'decorators'.
- ``Python/import.c`` Do notice that something like 'arguments', which is a node type, is
Defines the magic word (named ``MAGIC``) used in .pyc files to detect if represented as a single AST node and not as a sequence of nodes as with
the bytecode used matches the one used by the version of Python running. stmt as one might expect.
This number needs to be changed to make sure that the running
interpreter does not try to execute bytecode that it does not know All three kinds also have an 'attributes' argument; this is shown by the
about. fact that 'attributes' lacks a '|' before it.
The statement definitions above generate the following C structure type::
typedef struct _stmt *stmt_ty;
struct _stmt {
enum { FunctionDef_kind=1, Return_kind=2, Yield_kind=3 } kind;
union {
struct {
identifier name;
arguments_ty args;
asdl_seq *body;
} FunctionDef;
struct {
expr_ty value;
} Return;
struct {
expr_ty value;
} Yield;
} v;
int lineno;
}
Also generated are a series of constructor functions that allocate (in
this case) a stmt_ty struct with the appropriate initialization. The
'kind' field specifies which component of the union is initialized. The
FunctionDef() constructor function sets 'kind' to FunctionDef_kind and
initializes the 'name', 'args', 'body', and 'attributes' fields.
Suggestions for bytecode development Memory Management
==================================== -----------------
A few things can be done to make sure that development goes smoothly when Before discussing the actual implementation of the compiler, a discussion of
experimenting with Python's bytecode. One is to delete all .py(c|o) files how memory is handled is in order. To make memory management simple, an arena
after each semantic change to Python/compile.c . That way all files will use is used. This means that a memory is pooled in a single location for easy
any bytecode changes. allocation and removal. What this gives us is the removal of explicit memory
deallocation. Because memory allocation for all needed memory in the compiler
registers that memory with the arena, a single call to free the arena is all
that is needed to completely free all memory used by the compiler.
Make sure to run the entire testing suite [#test-suite]_. Since the In general, unless you are working on the critical core of the compiler, memory
``regrtest.py`` driver recompiles all source code before a test is run it acts management can be completely ignored. But if you are working at either the
a good test to make sure that no existing semantics are broken. very beginning of the compiler or the end, you need to care about how the arena
works. All code relating to the arena is in either Include/pyarena.h or
Python/pyarena.c .
Running parrotbench [#parrotbench]_ is also a good way to make sure existing PyArena_New() will create a new arena. The returned PyArena structure will
semantics are not broken; this benchmark is practically a compliance test. store pointers to all memory given to it. This does the bookkeeping of what
memory needs to be freed when the compiler is finished with the memory it used.
That freeing is done with PyArena_Free(). This needs to only be called in
strategic areas where the compiler exits.
As stated above, in general you should not have to worry about memory
management when working on the compiler. The technical details have been
designed to be hidden from you for most cases.
Previous experiments Parse Tree to AST
==================== -----------------
This section lists known bytecode experiments that have not gone into Python.
The AST is generated from the parse tree (see Python/ast.c) using the
function ``PyAST_FromNode()``.
The function begins a tree walk of the parse tree, creating various AST
nodes as it goes along. It does this by allocating all new nodes it
needs, calling the proper AST node creation functions for any required
supporting functions, and connecting them as needed.
Do realize that there is no automated nor symbolic connection between
the grammar specification and the nodes in the parse tree. No help is
directly provided by the parse tree as in yacc.
For instance, one must keep track of
which node in the parse tree one is working with (e.g., if you are
working with an 'if' statement you need to watch out for the ':' token
to find the end of the conditional). No help is directly provided by
the parse tree as in yacc.
The functions called to generate AST nodes from the parse tree all have
the name ast_for_xx where xx is what the grammar rule that the function
handles (alias_for_import_name is the exception to this). These in turn
call the constructor functions as defined by the ASDL grammar and
contained in Python/Python-ast.c (which was generated by
Parser/asdl_c.py) to create the nodes of the AST. This all leads to a
sequence of AST nodes stored in asdl_seq structs.
Function and macros for creating and using ``asdl_seq *`` types as found
in Python/asdl.c and Include/asdl.h:
- ``asdl_seq_new()``
Allocate memory for an asdl_seq for the specified length
- ``asdl_seq_GET()``
Get item held at a specific position in an asdl_seq
- ``asdl_seq_SET()``
Set a specific index in an asdl_seq to the specified value
- ``asdl_seq_APPEND()``
Append a value to the end of an asdl_seq
- ``asdl_seq_LEN(asdl_seq *)``
Return the length of an asdl_seq
If you are working with statements, you must also worry about keeping
track of what line number generated the statement. Currently the line
number is passed as the last parameter to each stmt_ty function.
Control Flow Graphs
-------------------
A control flow graph (often referenced by its acronym, CFG) is a
directed graph that models the flow of a program using basic blocks that
contain the intermediate representation (abbreviated "IR", and in this
case is Python bytecode) within the blocks. Basic blocks themselves are
a block of IR that has a single entry point but possibly multiple exit
points. The single entry point is the key to basic blocks; it all has
to do with jumps. An entry point is the target of something that
changes control flow (such as a function call or a jump) while exit
points are instructions that would change the flow of the program (such
as jumps and 'return' statements). What this means is that a basic
block is a chunk of code that starts at the entry point and runs to an
exit point or the end of the block.
As an example, consider an 'if' statement with an 'else' block. The
guard on the 'if' is a basic block which is pointed to by the basic
block containing the code leading to the 'if' statement. The 'if'
statement block contains jumps (which are exit points) to the true body
of the 'if' and the 'else' body (which may be NULL), each of which are
their own basic blocks. Both of those blocks in turn point to the
basic block representing the code following the entire 'if' statement.
CFGs are usually one step away from final code output. Code is directly
generated from the basic blocks (with jump targets adjusted based on the
output order) by doing a post-order depth-first search on the CFG
following the edges.
AST to CFG to Bytecode
----------------------
With the AST created, the next step is to create the CFG. The first step
is to convert the AST to Python bytecode without having jump targets
resolved to specific offsets (this is calculated when the CFG goes to
final bytecode). Essentially, this transforms the AST into Python
bytecode with control flow represented by the edges of the CFG.
Conversion is done in two passes. The first creates the namespace
(variables can be classified as local, free/cell for closures, or
global). With that done, the second pass essentially flattens the CFG
into a list and calculates jump offsets for final output of bytecode.
The conversion process is initiated by a call to the function
``PyAST_Compile()`` in Python/compile.c . This function does both the
conversion of the AST to a CFG and
outputting final bytecode from the CFG. The AST to CFG step is handled
mostly by two functions called by PyAST_Compile(); PySymtable_Build() and
compiler_mod() . The former is in Python/symtable.c while the latter is in
Python/compile.c .
PySymtable_Build() begins by entering the starting code block for the
AST (passed-in) and then calling the proper symtable_visit_xx function
(with xx being the AST node type). Next, the AST tree is walked with
the various code blocks that delineate the reach of a local variable
as blocks are entered and exited using symtable_enter_block() and
symtable_exit_block(), respectively.
Once the symbol table is created, it is time for CFG creation, whose
code is in Python/compile.c . This is handled by several functions
that break the task down by various AST node types. The functions are
all named compiler_visit_xx where xx is the name of the node type (such
as stmt, expr, etc.). Each function receives a ``struct compiler *``
and xx_ty where xx is the AST node type. Typically these functions
consist of a large 'switch' statement, branching based on the kind of
node type passed to it. Simple things are handled inline in the
'switch' statement with more complex transformations farmed out to other
functions named compiler_xx with xx being a descriptive name of what is
being handled.
When transforming an arbitrary AST node, use the VISIT() macro.
The appropriate compiler_visit_xx function is called, based on the value
passed in for <node type> (so ``VISIT(c, expr, node)`` calls
``compiler_visit_expr(c, node)``). The VISIT_SEQ macro is very similar,
but is called on AST node sequences (those values that were created as
arguments to a node that used the '*' modifier). There is also
VISIT_SLICE() just for handling slices.
Emission of bytecode is handled by the following macros:
- ``ADDOP()``
add a specified opcode.
- ``ADDOP_I()``
add an opcode that takes an argument
- ``ADDOP_O(struct compiler *c, int op, PyObject *type, PyObject *obj)``
add an opcode with the proper argument based on the position of the
specified PyObject in PyObject sequence object, but with no handling of
mangled names; used for when you
need to do named lookups of objects such as globals, consts, or
parameters where name mangling is not possible and the scope of the
name is known
- ``ADDOP_NAME()``
just like ADDOP_O, but name mangling is also handled; used for
attribute loading or importing based on name
- ``ADDOP_JABS()``
create an absolute jump to a basic block
- ``ADDOP_JREL()``
create a relative jump to a basic block
Several helper functions that will emit bytecode and are named
compiler_xx() where xx is what the function helps with (list, boolop,
etc.). A rather useful one is compiler_nameop().
This function looks up the scope of a variable and, based on the
expression context, emits the proper opcode to load, store, or delete
the variable.
As for handling the line number on which a statement is defined, is
handled by compiler_visit_stmt() and thus is not a worry.
In addition to emitting bytecode based on the AST node, handling the
creation of basic blocks must be done. Below are the macros and
functions used for managing basic blocks:
- ``NEW_BLOCK()``
create block and set it as current
- ``NEXT_BLOCK()``
basically NEW_BLOCK() plus jump from current block
- ``compiler_new_block()``
create a block but don't use it (used for generating jumps)
Once the CFG is created, it must be flattened and then final emission of
bytecode occurs. Flattening is handled using a post-order depth-first
search. Once flattened, jump offsets are backpatched based on the
flattening and then a PyCodeObject file is created. All of this is
handled by calling assemble() .
Introducing New Bytecode
------------------------
Sometimes a new feature requires a new opcode. But adding new bytecode is
not as simple as just suddenly introducing new bytecode in the AST ->
bytecode step of the compiler. Several pieces of code throughout Python depend
on having correct information about what bytecode exists.
First, you must choose a name and a unique identifier number. The official
list of bytecode can be found in Include/opcode.h . If the opcode is to take
an argument, it must be given a unique number greater than that assigned to
``HAVE_ARGUMENT`` (as found in Include/opcode.h``).
Once the name/number pair
has been chosen and entered in Include/opcode.h, you must also enter it into
Lib/opcode.py and Doc/lib/libdis.tex .
With a new bytecode you must also change what is called the magic number for
.pyc files. The variable ``MAGIC`` in Python/import.c contains the number.
Changing this number will lead to
Finally, you need to introduce the use of the new bytecode. Altering
Python/compile.c will be the primary place for changes. But you will also need
to change the 'compiler' package. The key files to do that are
Lib/compiler/pyassem.py and Lib/compiler/pycodegen.py .
If you make a change here that can affect the output of bytecode that
is already in existence and you do not change the magic number constantly, make
sure to delete your old .py(c|o) files! Even though you will end up changing
the magic number if you change the bytecode, while you are debugging your work
you will be changing the bytecode output without constantly bumping up the
magic number. This means you end up with stale .pyc files that will not be
recreated. Running
``find . -name '*.py[co]' -exec rm -f {} ';'`` should delete all .pyc files you
have, forcing new ones to be created and thus allow you test out your new
bytecode properly.
Code Objects
------------
In the end, one ends up with a PyCodeObject which is defined in
Include/code.h . And with that you now have executable Python bytecode!
Important Files
---------------
+ Parser/
- Python.asdl
ASDL syntax file
- asdl.py
"An implementation of the Zephyr Abstract Syntax Definition
Language." Uses SPARK_ to parse the ASDL files.
- asdl_c.py
"Generate C code from an ASDL description." Generates
Python/Python-ast.c and Include/Python-ast.h .
- spark.py
SPARK_ parser generator
+ Python/
- Python-ast.c
Creates C structs corresponding to the ASDL types. Also
contains code for marshaling AST nodes (core ASDL types have
marshaling code in asdl.c). "File automatically generated by
Parser/asdl_c.py".
- asdl.c
Contains code to handle the ASDL sequence type. Also has code
to handle marshalling the core ASDL types, such as number and
identifier. used by Python-ast.c for marshaling AST nodes.
- ast.c
Converts Python's parse tree into the abstract syntax tree.
- compile.c
Emits bytecode based on the AST.
- symtable.c
Generates a symbol table from AST.
- pyarena.c
Implementation of the arena memory manager.
- import.c
Home of the magic number (named ``MAGIC``) for bytecode versioning
+ Include/
- Python-ast.h
Contains the actual definitions of the C structs as generated by
Python/Python-ast.c .
"Automatically generated by Parser/asdl_c.py".
- asdl.h
Header for the corresponding Python/ast.c .
- ast.h
Declares PyAST_FromNode() external (from Python/ast.c).
- code.h
Header file for Objects/codeobject.c; contains definition of
PyCodeObject.
- symtable.h
Header for Python/symtable.c . struct symtable and
PySTEntryObject are defined here.
- pyarena.h
Header file for the corresponding Python/pyarena.c .
- opcode.h
Master list of bytecode; if this file is modified you must modify
several other files accordingly (see "`Introducing New Bytecode`_")
+ Objects/
- codeobject.c
Contains PyCodeObject-related code (originally in
Python/compile.c).
+ Lib/
- opcode.py
One of the files that must be modified if Include/opcode.h is.
- compiler/
* pyassem.py
One of the files that must be modified if Include/opcode.h is
changed.
* pycodegen.py
One of the files that muc be modified if Include/opcode.h is
changed.
Known Compiler-related Experiments
----------------------------------
This section lists known experiments involving the compiler (including
bytecode).
Skip Montanaro presented a paper at a Python workshop on a peephole optimizer Skip Montanaro presented a paper at a Python workshop on a peephole optimizer
[#skip-peephole]_. [#skip-peephole]_.
@ -119,30 +528,23 @@ for new-style classes rough benchmarking showed an actual slowdown thanks to
having to support both classic and new-style classes. having to support both classic and new-style classes.
References References
========== ----------
.. [#lang_ref] Python Language Reference, van Rossum & Drake .. [Aho86] Alfred V. Aho, Ravi Sethi, Jeffrey D. Ullman.
(http://docs.python.org/ref/ref.html) `Compilers: Principles, Techniques, and Tools`,
http://www.amazon.com/exec/obidos/tg/detail/-/0201100886/104-0162389-6419108
.. [#PEP-306] PEP 306, How to Change Python's Grammar, Hudson .. [Wang97] Daniel C. Wang, Andrew W. Appel, Jeff L. Korn, and Chris
(http://www.python.org/peps/pep-0306.html) S. Serra. `The Zephyr Abstract Syntax Description Language.`_
In Proceedings of the Conference on Domain-Specific Languages, pp.
213--227, 1997.
.. [#dis] dis Module .. _The Zephyr Abstract Syntax Description Language.:
(http://docs.python.org/lib/module-dis.html) http://www.cs.princeton.edu/~danwang/Papers/dsl97/dsl97.html
.. [#compiler] 'compiler' Package .. _SPARK: http://pages.cpsc.ucalgary.ca/~aycock/spark/
(http://docs.python.org/lib/module-compiler.html)
.. [#test-suite] 'test' Package
(http://docs.python.org/lib/module-test.html)
.. [#opcode_list] Python Byte Code Instructions
(http://docs.python.org/lib/bytecodes.html)
.. [#parrotbench] Parrotbench
(ftp://ftp.python.org/pub/python/parrotbench/parrotbench.tgz,
http://mail.python.org/pipermail/python-dev/2003-December/041527.html)
.. [#skip-peephole] Skip Montanaro's Peephole Optimizer Paper .. [#skip-peephole] Skip Montanaro's Peephole Optimizer Paper
(http://www.foretec.com/python/workshops/1998-11/proceedings/papers/montanaro/montanaro.html) (http://www.foretec.com/python/workshops/1998-11/proceedings/papers/montanaro/montanaro.html)
@ -154,11 +556,6 @@ References
(http://www.python.org/sf/709744) (http://www.python.org/sf/709744)
Copyright
=========
This document has been placed in the public domain.
.. ..
@ -166,5 +563,5 @@ This document has been placed in the public domain.
mode: indented-text mode: indented-text
indent-tabs-mode: nil indent-tabs-mode: nil
sentence-end-double-space: t sentence-end-double-space: t
fill-column: 70 fill-column: 80
End: End: