我正在努力解决一个相当棘手的问题。 我想计算两个列表之间的某种差异度量。一个是验证正确的数据,另一个是由程序生成的。 我想测试这个程序的准确程度,但为了达到这个目的,我需要以某种方式合并这两个列表。
示例数据可以在下面找到。
预期输出:字典列表,每个条目指定是否有两个匹配的条目。如果它们不匹配,则应该有错误类型,指明错误。
输出列表必须是两个列表中最大列表的长度,此列表中的每个条目应为以下之一:
{"matching": True}
{"matching": False, "error_type": "deleted"}
{"matching": False, "error_type": "inserted"}
{"matching": False, "error_type": "updated"}
我在Python 3中工作,所以如果有人能够提供令人惊叹的Python代码,但是一些准确描述这种算法的伪代码会有很多帮助! 如果您找到一种更简单的方法来表示输出数据,那也没关系。
到目前为止我所做的并不多,我真的无法理解如何开始这个,但是到目前为止我只有少量的代码:
compare_data = [{} for i in range(max(len(correct_data), len(program_data)))]
for i in range(len(compare_data)):
if len(program_data) <= i:
compare_data[i]['matching'] = False
compare_data[i]['error_type'] = 'deleted'
elif len(correct_data) <= i:
compare_data[i]['matching'] = False
compare_data[i]['error_type'] = 'inserted'
elif correct_data[i]['type'] != program_data[i]['type']:
compare_data[i]['matching'] = False
compare_data[i]['matching'] = 'updated'
else:
compare_data[i]['matching'] = True
# I really don't have a clue what to do...
我最初正在寻找一些通用的&#34;合并两个具有类似结构的不平等列表&#34;可以应用于此的算法,但我无法找到实现此目的的算法。
正确的数据:
[
{"type": "a", "data": ["some", "random"]},
{"type": "b", "data": ["data", "for", "people", "to"]},
{"type": "b", "data": ["mess", "with", "on", "stack"]},
{"type": "b", "data": ["over", "flow", "so", "they"]},
{"type": "c", "data": ["can", "try", "and"]},
{"type": "d", "data": ["help", "me"]}
]
该程序可以生成各种错误数据,这里有几个例子:
缺少参赛作品:
[
{"type": "a", "data":["some", "random"]},
{"type": "b", "data":["data", "for", "people", "to"]},
{"type": "b", "data":["mess", "with", "on", "stack"]},
{"type": "c", "data":["can", "try", "and"]},
{"type": "d", "data":["help", "me"]}
]
预期产出:
[
{"matching": True},
{"matching": True},
{"matching": True},
{"matching": False, "error_type": "delete"},
{"matching": True},
{"matching": True},
]
交换数据:
[
{"type": "a", "data":["some", "random"]},
{"type": "b", "data":["mess", "for", "people", "stack"]},
{"type": "b", "data":["data", "with", "on", "to"]},
{"type": "b", "data":["over", "flow", "so", "they"]},
{"type": "c", "data":["can", "try", "and"]},
{"type": "d", "data":["help", "me"]}
]
预期产出:
[
{"matching": True},
{"matching": False, "error_type": "updated"},
{"matching": False, "error_type": "updated"},
{"matching": True},
{"matching": True},
{"matching": True},
]
多个不同类型的缺失条目:
[
{"type": "b", "data":["data", "for", "people", "to"]},
{"type": "b", "data":["mess", "with", "on", "stack"]},
{"type": "c", "data":["can", "try", "and"]},
{"type": "d", "data":["help", "me"]}
]
预期产出:
[
{"matching": False, "error_type": "deleted"},
{"matching": True},
{"matching": True},
{"matching": False, "error_type": "deleted"},
{"matching": True},
{"matching": True},
]
具有相同类型的多个缺失条目:
[
{"type": "a", "data":["some", "random"]},
{"type": "b", "data":["mess", "with", "on", "stack"]},
{"type": "c", "data":["can", "try", "and"]},
{"type": "d", "data":["help", "me"]}
]
预期产出:
[
{"matching": True},
{"matching": False, "error_type": "deleted"},
{"matching": True},
{"matching": False, "error_type": "deleted"},
{"matching": True},
{"matching": True},
]
添加了条目(当然也可以添加多个条目):
[
{"type": "a", "data":["some", "random"]},
{"type": "b", "data":["data", "for", "people", "to"]},
{"type": "b", "data":["oops", "with", "got", "mangled"]},
{"type": "b", "data":["mess", "these", "on", "stack"]},
{"type": "b", "data":["over", "flow", "so", "they"]},
{"type": "c", "data":["can", "try", "and"]},
{"type": "d", "data":["help", "me"]}
]
预期产出:
[
{"matching": True},
{"matching": True},
{"matching": False, "error_type": "updated"},
{"matching": False, "error_type": "inserted"},
{"matching": True},
{"matching": True},
{"matching": True}
]
答案 0 :(得分:1)
你似乎在分歧。差异化并不容易。
Python有difflib,其中包含一些内置的diff算法。如果你对结果没问题,那么你可以使用下面概述的这类东西:
from difflib import SequenceMatcher
correct = [
{"type": "a", "data": ["some", "random"]},
{"type": "b", "data": ["data", "for", "people", "to"]},
{"type": "b", "data": ["mess", "with", "on", "stack"]},
{"type": "b", "data": ["over", "flow", "so", "they"]},
{"type": "c", "data": ["can", "try", "and"]},
{"type": "d", "data": ["help", "me"]}
]
compares = [
('Missing Entry',
[
{"type": "a", "data":["some", "random"]},
{"type": "b", "data":["data", "for", "people", "to"]},
{"type": "b", "data":["mess", "with", "on", "stack"]},
{"type": "c", "data":["can", "try", "and"]},
{"type": "d", "data":["help", "me"]}
]),
('Swapped data',
[
{"type": "a", "data":["some", "random"]},
{"type": "b", "data":["mess", "for", "people", "stack"]},
{"type": "b", "data":["data", "with", "on", "to"]},
{"type": "b", "data":["over", "flow", "so", "they"]},
{"type": "c", "data":["can", "try", "and"]},
{"type": "d", "data":["help", "me"]}
]),
# ...
]
def data_as_textlines(data):
'Turns a list of dataitems into a list of each items repr string'
return [repr(item) for item in data]
correct_text = C = data_as_textlines(correct)
for (title, prog_data) in compares:
print('\n' + title)
print('=' * len(title))
prog_text = P = data_as_textlines(prog_data)
s = SequenceMatcher(None, correct_text, prog_text)
for tag, i1, i2, j1, j2 in s.get_opcodes():
print('{:7} C[{}:{}] --> P[{}:{}] {!r:>8} --> {!r}'.format(
tag, i1, i2, j1, j2, correct_text[i1:i2], prog_text[j1:j2]))
其输出为:
Missing Entry
=============
equal C[0:3] --> P[0:3] ["{'type': 'a', 'data': ['some', 'random']}", "{'type': 'b', 'data': ['data', 'for', 'people', 'to']}", "{'type': 'b', 'data': ['mess', 'with', 'on', 'stack']}"] --> ["{'type': 'a', 'data': ['some', 'random']}", "{'type': 'b', 'data': ['data', 'for', 'people', 'to']}", "{'type': 'b', 'data': ['mess', 'with', 'on', 'stack']}"]
delete C[3:4] --> P[3:3] ["{'type': 'b', 'data': ['over', 'flow', 'so', 'they']}"] --> []
equal C[4:6] --> P[3:5] ["{'type': 'c', 'data': ['can', 'try', 'and']}", "{'type': 'd', 'data': ['help', 'me']}"] --> ["{'type': 'c', 'data': ['can', 'try', 'and']}", "{'type': 'd', 'data': ['help', 'me']}"]
Swapped data
============
equal C[0:1] --> P[0:1] ["{'type': 'a', 'data': ['some', 'random']}"] --> ["{'type': 'a', 'data': ['some', 'random']}"]
replace C[1:3] --> P[1:3] ["{'type': 'b', 'data': ['data', 'for', 'people', 'to']}", "{'type': 'b', 'data': ['mess', 'with', 'on', 'stack']}"] --> ["{'type': 'b', 'data': ['mess', 'for', 'people', 'stack']}", "{'type': 'b', 'data': ['data', 'with', 'on', 'to']}"]
equal C[3:6] --> P[3:6] ["{'type': 'b', 'data': ['over', 'flow', 'so', 'they']}", "{'type': 'c', 'data': ['can', 'try', 'and']}", "{'type': 'd', 'data': ['help', 'me']}"] --> ["{'type': 'b', 'data': ['over', 'flow', 'so', 'they']}", "{'type': 'c', 'data': ['can', 'try', 'and']}", "{'type': 'd', 'data': ['help', 'me']}"]
查看difflib.SequenceMatcher.get_opcodes的文档您将在他们的示例中找到某些相似之处: - )