«Отладка в Python 3.6: Быстрее, Выше, Сильнее» Елизавета Шашкова, JetBrains

Debugging in Python 3.6:
Better, Faster, Stronger
Elizaveta Shashkova
JetBrains

Bio
• Software developer of PyCharm IDE at JetBrains
• Debugger
• Saint Petersburg
2

Debugging
• Adding print statements
• Logging
3

Debugging
• Adding print statements
• Logging
• Special tools: debuggers
4

Debugger’s Performance
5
Run
Debug
Debuggers are 30 times slower

Contents
• Tracing debugger
• Python 3.6
• Frame evaluation debugger
• Results
6

Contents
• Python 3.6
• Results
7

8
def tracefunc(frame, event, arg):
print(frame.f_lineno, event)
return tracefunc
sys.settrace(tracefunc)
1
2
3
4
5
6
sys.settrace(tracefunc) - set the
system tracing function
Tracing Function

def foo():
friends = ["Bob", "Tom"]
for f in friends:
print("Hi %s!” % f)
return len(friends)
foo()
1
2
3
4
5
6
7
8
9
Tracing Function
9

1
2
3
4
5
6
7
8
9
1 calldef foo():
for f in friends:
return len(friends)
foo()
Tracing Function
10

1
2
3
4
5
6
7
8
9
1 call
2 line
def foo():
for f in friends:
return len(friends)
foo()
Tracing Function
11

1
2
3
4
5
6
7
8
9
1 call
2 line
3 line
4 line
Hi Bob!
def foo():
for f in friends:
return len(friends)
foo()
Tracing Function
12

1
2
3
4
5
6
7
8
9
1 call
2 line
3 line
4 line
Hi Bob!
3 line
4 line
Hi Tom!
def foo():
for f in friends:
return len(friends)
foo()
Tracing Function
13

1
2
3
4
5
6
7
8
9
1 call
2 line
3 line
4 line
Hi Bob!
3 line
4 line
Hi Tom!
5 line
5 return
def foo():
for f in friends:
return len(friends)
foo()
Tracing Function
14

Build Python Debugger
• Breakpoints
• Stepping
15

Tracing Debugger
• Suspend program if breakpoint’s
line equals frame.f_lineno
• Handle events for stepping
16

Performance
def foo():
for f in friends:
return len(friends)
foo()
1
2
3
4
5
6
7
8
9
1 call
2 line
3 line
4 line
Hi Bob!
3 line
4 line
Hi Tom!
5 line
5 return
17

Example 1
def calculate():
sum = 0
for i in range(10 ** 7):
sum += i
return sum
1
2
3
4
5
6
7
8
9
18

Example 1
def calculate():
sum = 0
sum += i
return sum
def tracefunc(frame, event, arg):
return tracefunc
1
2
3
4
5
6
7
8
9
19

Performance
20
Run
Tracing
Breakpoints
0,80 sec
6,85 sec
19,81 sec

Performance
21
Run
Tracing
Breakpoints
0,80 sec
6,85 sec
19,81 sec

Performance
22
Run
Tracing
Breakpoints
0,80 sec
6,85 sec
19,81 sec

Performance
23
Run
Tracing
Breakpoints
0,80 sec
6,85 sec
19,81 sec
~ 25 times slower!

Problem
• Tracing call on every line
24

Optimization 1
• Skip scopes without breakpoints
• frame.f_code.co_name - name
of the function
26

Optimization 2
• Rewrite tracing function in C
27

Optimization 2
• Rewrite tracing function in C
• Works only for CPython
• ~ 4 times faster
28

Problem
29

Contents
• Python 3.6
• Results
30

Contents
• Python 3.6
• Results
31

Python 3.6
• New frame evaluation API
• PEP 523
33

PEP 523
• Handle evaluation of frames
• Add a new ﬁeld to code objects
• C API
34

Frame Evaluation
cdef frame_eval(PyFrameObject *frame, int exc):
func_name = <object> frame.f_code.co_name
line_number = <object> frame.f_lineno
print(line_number, func_name)
return _PyEval_EvalFrameDefault(frame, exc)
1
2
3
4
5
6
7
8
9
35

func_name = <object> frame.f_code.co_name f
1
2
3
4
5
6
7
8
9
Frame Evaluation
36

1
2
3
4
5
6
7
8
9
Frame Evaluation
37

1
2
3
4
5
6
7
8
9
Frame Evaluation
38

1
2
3
4
5
6
7
8
9
Frame Evaluation
39

def set_frame_eval():
cdef PyThreadState *state =
PyThreadState_Get()
state.interp.eval_frame = frame_eval
1
2
3
4
5
6
7
8
9
Frame Evaluation
40

Example
1
2
3
4
5
6
7
8
9
10
11
def first():
second()
def second():
third()
def third():
pass
set_frame_eval()
first()
41

Example
def first():
second()
def second():
third()
def third():
pass
set_frame_eval()
first()
1
2
3
4
5
6
7
8
9
10
11
1 first
4 second
7 third
42

Custom Frame Evaluation
• It works
• Executed while entering a frame
• Access to frame and code object
43

Contents
• Python 3.6
• Results
44

Problem
45

Problem
• Remove the tracing function
46

Replace tracing
function with frame
evaluation function
47

Contents
• Python 3.6
• Results
48

• Breakpoints
• Stepping
49

Breakpoints
• Access to the whole code object
50

Breakpoints
• Access to the whole code object
• Insert breakpoint’s code into
frame’s code
51

Breakpoints
def maximum(a, b):
if a > b:
return a
else:
return b
1
2
3
4
5
6
7
8
52

Breakpoints
def maximum(a, b):
if a > b:
return a # breakpoint
else:
return b
1
2
3
4
5
6
7
8
53

Breakpoints
def maximum(a, b):
if a > b:
else:
return b
1
2
3
4
5
6
7
8
breakpoint()
54

Breakpoints
def maximum(a, b):
if a > b:
breakpoint()
else:
return b
1
2
3
4
5
6
7
8
55

Python Bytecode
def maximum(a, b):
if a > b:
return a
else:
return b
import dis
dis.dis(maximum)
1
2
3
4
5
6
7
8
56

Python Bytecode
def maximum(a, b):
if a > b:
return a
else:
return b
import dis
dis.dis(maximum)
1
2
3
4
5
6
7
8
2 0 LOAD_FAST 0 (a)
2 LOAD_FAST 1 (b)
4 COMPARE_OP 4 (>)
6 POP_JUMP_IF_FALSE 12
3 8 LOAD_FAST 0 (a)
10 RETURN_VALUE
5 >> 12 LOAD_FAST 1 (b)
14 RETURN_VALUE
16 LOAD_CONST 0 (None)
18 RETURN_VALUE
57

Python Bytecode
def maximum(a, b):
if a > b:
return a
else:
return b
import dis
dis.dis(maximum)
1
2
3
4
5
6
7
8
2 0 LOAD_FAST 0 (a)
2 LOAD_FAST 1 (b)
4 COMPARE_OP 4 (>)
3 8 LOAD_FAST 0 (a)
10 RETURN_VALUE
5 >> 12 LOAD_FAST 1 (b)
14 RETURN_VALUE
18 RETURN_VALUE
58

Python Bytecode
def maximum(a, b):
if a > b:
return a
else:
return b
import dis
dis.dis(maximum)
1
2
3
4
5
6
7
8
2 0 LOAD_FAST 0 (a)
2 LOAD_FAST 1 (b)
4 COMPARE_OP 4 (>)
3 8 LOAD_FAST 0 (a)
10 RETURN_VALUE
5 >> 12 LOAD_FAST 1 (b)
14 RETURN_VALUE
18 RETURN_VALUE
59

Python Bytecode
def maximum(a, b):
if a > b:
return a
else:
return b
import dis
dis.dis(maximum)
1
2
3
4
5
6
7
8
2 0 LOAD_FAST 0 (a)
2 LOAD_FAST 1 (b)
4 COMPARE_OP 4 (>)
3 8 LOAD_FAST 0 (a)
10 RETURN_VALUE
5 >> 12 LOAD_FAST 1 (b)
14 RETURN_VALUE
18 RETURN_VALUE
60

Python Bytecode
def maximum(a, b):
if a > b:
return a
else:
return b
import dis
dis.dis(maximum)
1
2
3
4
5
6
7
8
2 0 LOAD_FAST 0 (a)
2 LOAD_FAST 1 (b)
4 COMPARE_OP 4 (>)
3 8 LOAD_FAST 0 (a)
10 RETURN_VALUE
5 >> 12 LOAD_FAST 1 (b)
14 RETURN_VALUE
18 RETURN_VALUE
61

Python Bytecode
2 0 LOAD_FAST 0 (a)
2 LOAD_FAST 1 (b)
4 COMPARE_OP 4 (>)
3 8 LOAD_FAST 0 (a)
10 RETURN_VALUE
5 >> 12 LOAD_FAST 1 (b)
14 RETURN_VALUE
18 RETURN_VALUE
62

2 0 LOAD_FAST 0 (a)
2 LOAD_FAST 1 (b)
4 COMPARE_OP 4 (>)
3 8 LOAD_FAST 0 (a)
10 RETURN_VALUE
5 >> 12 LOAD_FAST 1 (b)
14 RETURN_VALUE
18 RETURN_VALUE
Python Bytecode
63

2 0 LOAD_FAST 0 (a)
2 LOAD_FAST 1 (b)
4 COMPARE_OP 4 (>)
3 8 LOAD_FAST 0 (a)
10 RETURN_VALUE
5 >> 12 LOAD_FAST 1 (b)
14 RETURN_VALUE
18 RETURN_VALUE
Python Bytecode
64

2 0 LOAD_FAST 0 (a)
2 LOAD_FAST 1 (b)
4 COMPARE_OP 4 (>)
3 8 LOAD_FAST 0 (a)
10 RETURN_VALUE
5 >> 12 LOAD_FAST 1 (b)
14 RETURN_VALUE
18 RETURN_VALUE
Python Bytecode
65

Bytecode Modification
2 0 LOAD_FAST 0 (a)
2 LOAD_FAST 1 (b)
4 COMPARE_OP 4 (>)
3 8 LOAD_FAST 0 (a)
10 RETURN_VALUE
5 >> 12 LOAD_FAST 1 (b)
14 RETURN_VALUE
18 RETURN_VALUE
breakpoint()
66

• Insert breakpoint’s code
• Update arguments and offsets
67

• Insert breakpoint’s code
• Update arguments and offsets
• 200 lines in Python
68

2 0 LOAD_FAST 0 (a)
2 LOAD_FAST 1 (b)
4 COMPARE_OP 4 (>)
3 8 LOAD_FAST 0 (a)
10 RETURN_VALUE
5 >> 12 LOAD_FAST 1 (b)
14 RETURN_VALUE
18 RETURN_VALUE
breakpoint()
?!
69

Breakpoint Bytecode
def _stop_at_break():
# a lot of code here
def breakpoint():
_stop_at_break()
1
2
3
4
5
6
7
8
0 LOAD_GLOBAL 0
2 CALL_FUNCTION 0
4 POP_TOP
6 LOAD_CONST 0
8 RETURN_VALUE
70

• Breakpoints
• Stepping
71

Stepping
• Inserting temporary breakpoint
on every line
• Use old tracing function
72

Frame evaluation
debugger is ready
73

Example 1
def calculate():
sum = 0
sum += i
return sum
1
2
3
4
5
74

Example 1
75
Run
Tracing
Frame  
evaluation
0,80 sec
19,81 sec
0,81 sec

Example 1
76
Run
Tracing
Frame  
evaluation
0,80 sec
19,81 sec
0,81 sec

Example 2
def foo():
pass
def calculate():
sum = 0
foo()
sum += i
return sum
1
2
3
4
5
6
7
8
9
77

Example 2
78
Run
Tracing
Frame  
evaluation
1,73 sec
43,58 sec
37,41 sec

PEP 523
79

PEP 523
• Expand PyCodeObject struct
• co_extra - “scratch space” for
the code object
• Mark frames without breakpoints
80

Mark Frames Cython
1
2
3
4
5
6
7
8
9
81
flag = _PyCode_GetExtra(frame.f_code, index)
if flag == NO_BREAKS_IN_FRAME:
# check for breakpoints
...

Example 2
82
Run
Tracing
Frame  
evaluation
1,73 sec
43,58 sec
1,91 sec

PEP 523
83

Contents
• Python 3.6
• Results
84

Contents
• Python 3.6
• Results
85

Real Life Example
• Included into PyCharm 2017.1
• Works in production
87

PyCharm
88
Tracing 
w/o Cython
Tracing 
with Cython
Frame  
evaluation
11,59 sec
5,66 sec
0,28 sec

Disadvantages
• More complicated
• Only with CPython
• Only with Python 3.6
89

Frame Evaluation
• Let’s move to Python 3.6!
90

Frame Evaluation
• Let’s ﬁnd another use cases
91

Use cases
def maximum(a, b):
if a > b:
else:
return b
1
2
3
4
5
6
7
8
breakpoint()
92

PEP 523
• Pyjion project
• JIT for Python
93

Frame Evaluation
• Let’s ﬁnd another use cases
94

Links
• Prototype: https://github.com/
Elizaveta239/frame-eval
• PyCharm Community Edition source
code
• bytesinsert on PyPi
95

• Prototype: https://github.com/
Elizaveta239/frame-eval
• PyCharm Community Edition source
code
• bytesinsert on PyPi
Questions?
96
@lisa_shashkova

«Отладка в Python 3.6: Быстрее, Выше, Сильнее» Елизавета Шашкова, JetBrains

More Related Content

«Отладка в Python 3.6: Быстрее, Выше, Сильнее» Елизавета Шашкова, JetBrains