使用正则表达式匹配组的优先规则
考虑以下 .NET 正则表达式：
^(REF)?(.{1,10})-(\d{12})-(\d+)$
它定义了我感兴趣的四个组，我将分别分析它们。
现在，考虑此正则表达式的输入字符串：
REFmisc03-123456789012-213
可以像这样匹配它：
(REF)(misc03)-(123456789012)-(213)
也可以像这样匹配它：
()(REFmisc03)-(123456789012)-(213)
是否记录了正则表达式引擎首选的方式，或者它是随机的?
Consider the following .NET regular expression:
^(REF)?(.{1,10})-(\d{12})-(\d+)$
It defines four groups, in which I'm interested and which I will analyse separately.
Now, consider an input string for this regexp:
REFmisc03-123456789012-213
It is possible to match it like this:
(REF)(misc03)-(123456789012)-(213)
And it is also possible to match it like this:
()(REFmisc03)-(123456789012)-(213)
Is it documented what way will be preferred by the regexp engine, or is it random?
如果你对这篇内容有疑问，欢迎到本站社区发帖提问 参与讨论，获取更多帮助，或者扫码二维码加入 Web 技术交流群。
绑定邮箱获取回复消息
由于您还没有绑定你的真实邮箱，如果其他用户或者作者回复了您的评论，将不能在第一时间通知您！
发布评论
评论（1）
这不是随机的。这归结为正则表达式引擎如何解释量词以及潜在的回溯。我所说的量词是指
(REF)?中的?。 根据 MSDN：换句话说，
?是贪婪的，而??是惰性的。两者都匹配零次或一次，但它们会对匹配的执行方式产生影响。关于回溯，MSDN 提到了：
可以在此处找到了解有关回溯的更多信息的另一个有用资源：所有格量词。
为了直接回答您的问题，我们可以比较这两种方法。
贪婪方法
原始输入：
REFmisc03-123456789012-213使用
(REF)?会将您的文本与4组匹配（不包括第一组）整场比赛）并且所有组都将成功匹配：这与您的第一个可能的比赛场景（松散定义）匹配：
只要
"misc..."部分的长度为 1-10 个字符，匹配结果将相同，全部为 1-第二组出现的10个角色。REF部分将始终在第一组中匹配。新输入：
REF-123456789012-213“misc...”部分缺失。由于(REF)?是可选的，而(.{1,10})不是可选的，因此正则表达式引擎将使用"REF"输入以满足模式的后一个（必需）部分并忽略前一个（可选）部分。这将产生以下组值：""（空字符串，Successproperty =false）惰性方法
原始输入：
REFmisc03-123456789012-213通过使用
(REF)??，并保留其余部分如果您的模式相同，则量词变得惰性，这会返回 4 个具有以下值的组：""（空字符串，Successproperty =false)这与您的第二种可能的匹配场景匹配：
由于第一组对于惰性量词是可选的，因此正则表达式引擎能够忽略它。由于
"REFmisc03"的长度为 9 个字符，因此引擎会将"REF"与"misc03"混在一起，因为它们适合>(.{1,10})组。新输入：
REF-123456789012-213这与贪婪模式的行为类似，并且适用相同的推理。
另一个新输入：
REFmisc0345-123456789012-213在此示例中，
“misc0345”部分的长度为 8 个字符。尽管该模式使用惰性量词，但它无法将"REFmisc0345"放入第二组，因为它超出了 10 个字符的限制。正则表达式引擎将回溯并匹配第一组中的"REF"和第二组中的"misc0345"：It is not random. This boils down to how quantifiers are interpreted by the regex engine and potential backtracking. By quantifier I am referring to the
?in(REF)?. According to MSDN:In other words,
?is greedy, and??is lazy. Both match zero or one time, but they will have an effect on how the matching is performed.With regards to backtracking, MSDN mentions:
Another useful resource to learn more about backtracking can be found here: Possessive Quantifiers.
To answer your question directly, we can compare both approaches.
Greedy approach
Original input:
REFmisc03-123456789012-213Usage of
(REF)?will match your text with 4 groups (excluding the first group with the entire match) and all groups will be successfully matched:This matches your first possible match scenario (loosely defined):
As long as the
"misc..."portion is 1-10 characters long, the match will be the same, with all 1-10 characters appearing in the second group. TheREFportion will always be matched in the first group.New input:
REF-123456789012-213The
"misc..."portion is absent. Since(REF)?is optional, and(.{1,10})isn't, the regex engine will use the"REF"input to satisfy the latter (required) portion of the pattern and disregard the former (optional) portion. This will yield the following group values:""(empty string,Successproperty =false)Lazy approach
Original input:
REFmisc03-123456789012-213By using
(REF)??, and keeping the rest of your pattern the same, the quantifier becomes lazy and this returns 4 groups with these values:""(empty string,Successproperty =false)This matches your second possible match scenario:
Since the first group is optional with a lazy quantifier, the regex engine is able to disregard it. Since
"REFmisc03"is 9 characters long, the engine proceeds to lump"REF"in with"misc03"because they fit into the(.{1,10})group.New input:
REF-123456789012-213This behaves similarly to the greedy pattern and the same reasoning applies.
Another new input:
REFmisc0345-123456789012-213In this example the
"misc0345"portion is 8 characters long. Although the pattern uses a lazy quantifier it can't fit"REFmisc0345"into the second group because it exceeds the 10 character limit. The regex engine will backtrack and match"REF"in the first group, and"misc0345"in the second group: