将 function_graph 的树转换为字典

Question

function_graph 跟踪器以类似于 C 代码和/或 JSON 的格式输出看似时间相关的树。是否有内置的python3可以将其转换为dict？

txt示例（每行末尾有一个\n）：

__fdget() {
  __fget_light() {
    __fget();
  }
}

Answer 1

因此，在没有找到任何内置 Python3 之后，我编写了一个将其转换为 Python3 字典对象的代码（忽略 /* cmets */）。

代码如下

import os, sys



def parse(total, i, nest, result, j, unique_elems):
    if len(total) == i: return
    # Path 1) it's a comment and some closing
    if '/*' in total[i] and '*/' in total[i]:
        if '}' in total[i]:
            nest -= 1
        if len(j)>1:
            j = j[:-1]
        parse(total, i+1, nest, result, j, unique_elems)
    # Path 2) it's only a closing thing
    elif '}' in total[i]:
        nest -= 1
        if nest == 0:
            result += [{}]
            j = [j[0]+1]
        else:
            if len(j)>1:
                j = j[:-1]  
        parse(total, i+1, nest, result, j, unique_elems)

    # Path 3) it's an opening
    elif '{' in total[i]:
        nest += 1
        localdata = total[i].split('()')[0].split(' ')[-1]
        unique_elems.add(localdata)
        update_item_and_expand(j[1:], result[j[0]], localdata)
        j.append(localdata)
        parse(total, i+1, nest, result, j, unique_elems)    

    # Path 4) it's a call that  ends  with ; without affecting nesting
    elif ';' in total[i]:
        localdata = total[i].split('()')[0].split(' ')[-1]
        unique_elems.add(localdata)
        update_item_and_expand(j[1:], result[j[0]], localdata)
        if nest == 0 and len(j)==1:
            j = [j[0]+1]
            result += [{}]
        parse(total, i+1, nest, result, j, unique_elems)    

并使用以下效用函数

def update_item_and_expand(j, result, localdata):
    if j==[]:
        result[localdata]={}
    elif len(j)==1:
        result[j[0]][localdata] = {}
    elif len(j)>1:
        update_item_and_expand(j[1:], result[j[0]], localdata)  
    else:
        print("[ERROR] [update_item_and_expand] Fatal unknown (bad design?)")
        sys.exit(1)

运行它并显示它，例如，

unique_elems = set()
result = [{}]
j=[0]
nest = 0
i=0
with open('kerneltrace.txt','r') as f:
    total = f.readlines()


parse(total, i, nest, result, j, unique_elems)

delim="\n-------------\n"
print(f"{delim}RAW{delim}")
for t in total:
    print(t)
print(f "{delim}PROCESSED{delim}")
for x in result:
    for k in x.keys():
        print(k)
        for k2 in x[k].keys():
            print(f'\t{k2}')
            print(f'\t\t{x[k][k2]}')

产量：

-------------
RAW
-------------

        /* __fdget */

        eventfd_poll() {

          __pollwait() {

            add_wait_queue() {

              _raw_spin_lock_irqsave();

              _raw_spin_unlock_irqrestore();

            }

          }

        }

        fput() {

          fput_many();

        }

        __fdget() {

          __fget_light() {

            __fget();

          }

        }

        eventfd_poll() {

          __pollwait() {

            add_wait_queue() {

              _raw_spin_lock_irqsave();

              _raw_spin_unlock_irqrestore();

            }

          }

        }


-------------
PROCESSED
-------------

eventfd_poll
    __pollwait
        {'add_wait_queue': {'_raw_spin_lock_irqsave': {}, '_raw_spin_unlock_irqrestore': {}}}
fput
    fput_many
        {}
__fdget
    __fget_light
        {'__fget': {}}
eventfd_poll
    __pollwait
        {'add_wait_queue': {'_raw_spin_lock_irqsave': {}, '_raw_spin_unlock_irqrestore': {}}}

这对于通过将空字典解释为嵌套结束来构建时间相关图/树很有用。