如何使用正则表达式从列表中填充字典？

Question

我有一个列表（“输出”）。我想从中提取值并将它们放入字典中。 到目前为止，我可以使用正则表达式提取一些单词。但是我不知道怎么填字典。

这是我的尝试

output = ['labels: imagenet_labels.txt \n', '\n', 'Model: efficientnet-edgetpu-S_quant_edgetpu.tflite \n', '\n', 'Image: img0000.jpg \n', '\n', '----INFERENCE TIME----\n', 'Note: The first inference on Edge TPU is slow because it includes loading the model into Edge TPU memory.\n', 'time: 6.0ms\n', '-------RESULTS--------\n','results: wall clock\n', 'score: 0.25781\n', '##################################### \n', ' \n', '\n']

mydict = {}

regex1 = re.compile(fr'(\w+:)\s(.*)')
match_regex1 = list(filter(regex1.match, output))
match = [line.rstrip('\n') for line in match_regex1]

字典必须如下所示：

{
'Model': "efficientnet-edgetpu-S_quant_edgetpu.tflite",
'Image': "img0000.jpg",
'time': "6.0",
'results': "wall_clock",
'score': :0.25781"
}

列表如下所示：

编辑

我已经做了这个循环。尽管它不能正常工作：

for i in output:
    reg1 = re.search(r'(\w+:)\s(.*)', i)
    if "Model" in i:
        mydict.setdefault("Model", {reg1.group()})
        print(mydict)

Answer 1

你可以试试这个，根据match列表：

import re
output = ['labels: imagenet_labels.txt \n', '\n', 'Model: efficientnet-edgetpu-S_quant_edgetpu.tflite \n', '\n', 'Image: img0000.jpg \n', '\n', '----INFERENCE TIME----\n', 'Note: The first inference on Edge TPU is slow because it includes loading the model into Edge TPU memory.\n', 'time: 6.0ms\n', '-------RESULTS--------\n','results: wall clock\n', 'score: 0.25781\n', '##################################### \n', ' \n', '\n']

mydict = {}

regex1 = re.compile(fr'(\w+:)\s(.*)')
match_regex1 = list(filter(regex1.match, output))
match = [line.rstrip('\n') for line in match_regex1]

features_wanted='ModelImagetimeresultsscore'

dct={i.replace(' ','').split(':')[0]:i.replace(' ','').split(':')[1] for i in match if i.replace(' ','').split(':')[0] in features_wanted}
mydict=dct
print(dct)

输出：

{'Model': 'efficientnet-edgetpu-S_quant_edgetpu.tflite', 'Image': 'img0000.jpg', 'time': '6.0ms', 'results': 'wallclock', 'score': '0.25781'}

dct 的解释：这是一个Dictionary Comprehension 并迭代列表匹配，所以这里是一个使用'Model: efficientnet-edgetpu-S_quant_edgetpu.tflite' 的迭代示例：

#First check if it is a feature wanted:
i='Model: efficientnet-edgetpu-S_quant_edgetpu.tflite'
i.replace(' ','')
>>>'Model:efficientnet-edgetpu-S_quant_edgetpu.tflite'
i.replace(' ','').split(':')
>>>['Model','efficientnet-edgetpu-S_quant_edgetpu.tflite']
i.replace(' ','').split(':')[0] in features_wanted  #'Model' in 'ModelImagetimeresultsscore'
>>>True
#If it is in features_wanted, an item like this is append to the dictionary:
i.replace(' ','').split(':')[0]:i.replace(' ','').split(':')[1]
>>>'Model':'efficientnet-edgetpu-S_quant_edgetpu.tflite'

Answer 2

要填写字典，您可以使用此脚本：

for item in match:
    key , value = item.split(":")
    mydict[key] = value

所以结果是这样的：

{'labels': ' imagenet_labels.txt ', 'Model': ' efficientnet-edgetpu-S_quant_edgetpu.tflite ', 'Image': ' img0000.jpg ', 'Note': ' The first inference on Edge TPU is slow because it includes loading the model into Edge TPU memory.', 'time': ' 6.0ms', 'results': ' wall clock', 'score': ' 0.25781'}

Answer 3

output = ['labels: imagenet_labels.txt \n', '\n', 'Model: efficientnet-edgetpu-S_quant_edgetpu.tflite \n', '\n', 'Image: img0000.jpg \n', '\n', '----INFERENCE TIME----\n', 'Note: The first inference on Edge TPU is slow because it includes loading the model into Edge TPU memory.\n', 'time: 6.0ms\n', '-------RESULTS--------\n','results: wall clock\n', 'score: 0.25781\n', '##################################### \n', ' \n', '\n']

d = dict( re.findall(r'(\w+):\s*([^\n]+?)\s*$', ' '.join(output), flags=re.M) )

from pprint import pprint
pprint(d)

打印：

{'Image': 'img0000.jpg',
 'Model': 'efficientnet-edgetpu-S_quant_edgetpu.tflite',
 'Note': 'The first inference on Edge TPU is slow because it includes loading '
         'the model into Edge TPU memory.',
 'labels': 'imagenet_labels.txt',
 'results': 'wall clock',
 'score': '0.25781',
 'time': '6.0ms'}

Answer 4

由于字段的分隔符始终为:，因此您可以使用str.split 方法代替正则表达式以提高效率：

dict(s.split(': ', 1) for s in map(str.rstrip, output) if ': ' in s)

演示：https://repl.it/@blhsing/SnoopyBoringComputationalscience