Skip to content

Commit

Permalink
pythongh-119786: add code object doc, inline locations.md into it
Browse files Browse the repository at this point in the history
  • Loading branch information
iritkatriel committed Nov 14, 2024
1 parent 91f4908 commit 9817511
Show file tree
Hide file tree
Showing 7 changed files with 187 additions and 82 deletions.
4 changes: 1 addition & 3 deletions InternalDocs/README.md
Original file line number Diff line number Diff line change
Expand Up @@ -24,9 +24,7 @@ Compiling Python Source Code
Runtime Objects
---

- [Code Objects (coming soon)](code_objects.md)

- [The Source Code Locations Table](locations.md)
- [Code Objects](code_objects.md)

- [Generators (coming soon)](generators.md)

Expand Down
43 changes: 43 additions & 0 deletions InternalDocs/_code_objects.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,43 @@

Code objects
============

The interpreter uses a code object (``frame->f_code``) as its starting point.
Code objects contain many fields used by the interpreter, as well as some for use by debuggers and other tools.
In 3.11, the final field of a code object is an array of indeterminate length containing the bytecode, ``code->co_code_adaptive``.
(In previous versions the code object was a :class:`bytes` object, ``code->co_code``; it was changed to save an allocation and to allow it to be mutated.)

Code objects are typically produced by the bytecode :ref:`compiler <compiler>`, although they are often written to disk by one process and read back in by another.
The disk version of a code object is serialized using the :mod:`marshal` protocol.
Some code objects are pre-loaded into the interpreter using ``Tools/scripts/deepfreeze.py``, which writes ``Python/deepfreeze/deepfreeze.c``.

Code objects are nominally immutable.
Some fields (including ``co_code_adaptive``) are mutable, but mutable fields are not included when code objects are hashed or compared.

The locations table
-------------------

Whenever an exception is raised, we add a traceback entry to the exception.
The ``tb_lineno`` field of a traceback entry is (lazily) set to the line number of the instruction that raised it.
This field is computed from the locations table, ``co_linetable`` (this name is an understatement), using :c:func:`PyCode_Addr2Line`.
This table has an entry for every instruction rather than for every ``try`` block, so a compact format is very important.

The full design of the 3.11 locations table is written up in :cpy-file:`InternalDocs/locations.md`.
While there are rumors that this file is slightly out of date, it is still the best reference we have.
Don't be confused by :cpy-file:`Objects/lnotab_notes.txt`, which describes the 3.10 format.
For backwards compatibility this format is still supported by the ``co_lnotab`` property.

The 3.11 location table format is different because it stores not just the starting line number for each instruction, but also the end line number, *and* the start and end column numbers.
Note that traceback objects don't store all this information -- they store the start line number, for backward compatibility, and the "last instruction" value.
The rest can be computed from the last instruction (``tb_lasti``) with the help of the locations table.
For Python code, a convenient method exists, :meth:`~codeobject.co_positions`, which returns an iterator of :samp:`({line}, {endline}, {column}, {endcolumn})` tuples, one per instruction.
There is also ``co_lines()`` which returns an iterator of :samp:`({start}, {end}, {line})` tuples, where :samp:`{start}` and :samp:`{end}` are bytecode offsets.
The latter is described by :pep:`626`; it is more compact, but doesn't return end line numbers or column offsets.
From C code, you have to call :c:func:`PyCode_Addr2Location`.

Fortunately, the locations table is only consulted by exception handling (to set ``tb_lineno``) and by tracing (to pass the line number to the tracing function).
In order to reduce the overhead during tracing, the mapping from instruction offset to line number is cached in the ``_co_linearray`` field.


TODO:
- co_consts, co_names, co_varnames, and their ilk
141 changes: 138 additions & 3 deletions InternalDocs/code_objects.md
Original file line number Diff line number Diff line change
@@ -1,5 +1,140 @@

Code objects
============
# Code objects

Coming soon.
A `CodeObject` is a builtin Python type that represents a compiled executable,
such as a compiled function or class.
It contains a sequence of bytecode instructions along with its associated
metadata: data which is necessary to execute the bytecode instructions (such
as the values of the constants they access) or context information such as
the source code location, which is useful for debuggers and other tools.

Since 3.11, the final field of the `PyCodeObject` C struct is an array
of indeterminate length containing the bytecode, `code->co_code_adaptive`.
(In older versions the code object was a
[`bytes`](https://docs.python.org/dev/library/stdtypes.html#bytes)
object, `code->co_code`; this was changed to save an allocation and to
allow it to be mutated.)

Code objects are typically produced by the bytecode [compiler](compiler.md),
although they are often written to disk by one process and read back in by another.
The disk version of a code object is serialized using the
[marshal](https://docs.python.org/dev/library/marshal.html) protocol.
Some code objects are pre-loaded into the interpreter using
[`Tools/build/deepfreeze.py`](../Tools/build/deepfreeze.py),
which writes
[`Python/deepfreeze/deepfreeze.c`](../Python/deepfreeze/deepfreeze.c).

Code objects are nominally immutable.
Some fields (including `co_code_adaptive` and fields for runtime
information such as `_co_monitoring`) are mutable, but mutable fields are
not included when code objects are hashed or compared.

## Source code locations

Whenever an exception occurs, the interpreter adds a traceback entry to
the exception for the current frame, as well as each frame on the stack that
it unwinds.
The `tb_lineno` field of a traceback entry is (lazily) set to the line
number of the instruction that was executing in the frame at the time of
the exception.
This field is computed from the locations table, `co_linetable`, by the function
[`PyCode_Addr2Line`](https://docs.python.org/dev/c-api/code.html#c.PyCode_Addr2Line).
Despite its name, `co_linetable` includes more than line numbers; it represents
a 4-number source location for every instruction, indicating the precise line
and column at which it begins and ends. This is a significant amount of data,
so a compact format is very important.

Note that traceback objects don't store all this information -- they store the start line
number, for backward compatibility, and the "last instruction" value.
The rest can be computed from the last instruction (`tb_lasti`) with the help of the
locations table. For Python code, there is a convenience method
(`codeobject.co_positions`)[https://docs.python.org/dev/reference/datamodel.html#codeobject.co_positions]
which returns an iterator of `({line}, {endline}, {column}, {endcolumn})` tuples,
one per instruction.
There is also `co_lines()` which returns an iterator of `({start}, {end}, {line})` tuples,
where `{start}` and `{end}` are bytecode offsets.
The latter is described by [`PEP 626`](https://peps.python.org/pep-0626/); it is more
compact, but doesn't return end line numbers or column offsets.
From C code, you need to call
[`PyCode_Addr2Location`](https://docs.python.org/dev/c-api/code.html#c.PyCode_Addr2Location).

As the locations table is only consulted by exception handling (to set ``tb_lineno``)
and by tracing (to pass the line number to the tracing function), lookup is not
performance critical.
In order to reduce the overhead during tracing, the mapping from instruction offset to
line number is cached in the ``_co_linearray`` field.

### Format of the locations table

The `co_linetable` bytes object of code objects contains a compact
representation of the source code positions of instructions, which are
returned by the `co_positions()` iterator.

> [!NOTE]
> Not to be confused by [`Objects/lnotab_notes.txt`](Objects/lnotab_notes.txt),
> which describes the 3.10 format, that stores only that start line for each instruction.
> For backwards compatibility this format is still supported by the `co_lnotab` property.
`co_linetable` consists of a sequence of location entries.
Each entry starts with a byte with the most significant bit set, followed by zero or more bytes with most significant bit unset.

Each entry contains the following information:
* The number of code units covered by this entry (length)
* The start line
* The end line
* The start column
* The end column

The first byte has the following format:

Bit 7 | Bits 3-6 | Bits 0-2
---- | ---- | ----
1 | Code | Length (in code units) - 1

The codes are enumerated in the `_PyCodeLocationInfoKind` enum.

## Variable length integer encodings

Integers are often encoded using a variable length integer encoding

### Unsigned integers (varint)

Unsigned integers are encoded in 6 bit chunks, least significant first.
Each chunk but the last has bit 6 set.
For example:

* 63 is encoded as `0x3f`
* 200 is encoded as `0x48`, `0x03`

### Signed integers (svarint)

Signed integers are encoded by converting them to unsigned integers, using the following function:
```Python
def convert(s):
if s < 0:
return ((-s)<<1) | 1
else:
return (s<<1)
```

*Location entries*

The meaning of the codes and the following bytes are as follows:

Code | Meaning | Start line | End line | Start column | End column
---- | ---- | ---- | ---- | ---- | ----
0-9 | Short form | Δ 0 | Δ 0 | See below | See below
10-12 | One line form | Δ (code - 10) | Δ 0 | unsigned byte | unsigned byte
13 | No column info | Δ svarint | Δ 0 | None | None
14 | Long form | Δ svarint | Δ varint | varint | varint
15 | No location | None | None | None | None

The Δ means the value is encoded as a delta from another value:
* Start line: Delta from the previous start line, or `co_firstlineno` for the first entry.
* End line: Delta from the start line

*The short forms*

Codes 0-9 are the short forms. The short form consists of two bytes, the second byte holding additional column information. The code is the start column divided by 8 (and rounded down).
* Start column: `(code*8) + ((second_byte>>4)&7)`
* End column: `start_column + (second_byte&15)`
8 changes: 3 additions & 5 deletions InternalDocs/compiler.md
Original file line number Diff line number Diff line change
Expand Up @@ -443,14 +443,12 @@ reference to the source code (filename, etc). All of this is implemented by
Code objects
============

The result of `PyAST_CompileObject()` is a `PyCodeObject` which is defined in
The result of `_PyAST_Compile()` is a `PyCodeObject` which is defined in
[Include/cpython/code.h](../Include/cpython/code.h).
And with that you now have executable Python bytecode!

The code objects (byte code) are executed in [Python/ceval.c](../Python/ceval.c).
This file will also need a new case statement for the new opcode in the big switch
statement in `_PyEval_EvalFrameDefault()`.

The code objects (byte code) are executed in `_PyEval_EvalFrameDefault()`
in [Python/ceval.c](../Python/ceval.c).

Important files
===============
Expand Down
2 changes: 1 addition & 1 deletion InternalDocs/interpreter.md
Original file line number Diff line number Diff line change
Expand Up @@ -16,7 +16,7 @@ from the instruction definitions in [Python/bytecodes.c](../Python/bytecodes.c)
which are written in [a DSL](../Tools/cases_generator/interpreter_definition.md)
developed for this purpose.

Recall that the [Python Compiler](compiler.md) produces a [`CodeObject`](code_object.md),
Recall that the [Python Compiler](compiler.md) produces a [`CodeObject`](code_objects.md),
which contains the bytecode instructions along with static data that is required to execute them,
such as the consts list, variable names,
[exception table](exception_handling.md#format-of-the-exception-table), and so on.
Expand Down
69 changes: 0 additions & 69 deletions InternalDocs/locations.md

This file was deleted.

2 changes: 1 addition & 1 deletion Objects/lnotab_notes.txt
Original file line number Diff line number Diff line change
@@ -1,7 +1,7 @@
Description of the internal format of the line number table in Python 3.10
and earlier.

(For 3.11 onwards, see Objects/locations.md)
(For 3.11 onwards, see InternalDocs/locations.md)

Conceptually, the line number table consists of a sequence of triples:
start-offset (inclusive), end-offset (exclusive), line-number.
Expand Down

0 comments on commit 9817511

Please sign in to comment.