如果解析逻辑中存在拆分,则使用正则表达式解析 "hierarchical" URL

Parsing a "hierarchical" URL with regexes if there are splits in parsing logic

有什么方法可以将剩余的正则表达式模式调整为已经匹配的模式吗?一个粗略的草图来说明这个想法:

          pattern
       /     |     \
      /      |      \
prefix1   prefix2   prefix3
   |         |         |
postfix1  postfix2  postfix3

这是一个比较理论化的问题;以下实际应用仅供说明之用。

我试图在大文本中找到第一个 URL 流行的代码托管平台,如 github、gitlab 等。问题是,所有平台都有不同的 URL 模式:

github.com/<user>/<repo>
gitlab.com/<group1>/<group2>/.../<repo>
sourceforge.net/projects/<repo>

我可以使用 lookbehind 表达式,但是表达式变得非常可怕 (Python re):

pattern = re.compile(
    r"(github\.com|bitbucket\.org|gitlab\.com|sourceforge\.net)/"
    # middle part - empty for all except sourceforge
    r"(?:(?<=github\.com/)|(?<=bitbucket\.org/)|(?<=gitlab\.com/)|"
    r"(?<=sourceforge\.net/)projects/)("
    # final part, the repository pattern
    r"(?<=github\.com/)[a-zA-Z0-9_.-]+/[a-zA-Z0-9_.-]+|"
    r"(?<=bitbucket\.org/)[a-zA-Z0-9_.-]+/[a-zA-Z0-9_.-]+|"
    r"(?<=gitlab\.com/)[a-zA-Z0-9_.-]+(?:/[a-zA-Z0-9_.-]+)+|"
    r"(?<=sourceforge\.net/projects/)[a-zA-Z0-9_.-]+"
    r")")

有没有更优雅的方法来做这样的事情?

最好的方法可能是使用自定义解析器并以状态机方式进行解析:首先确定站点,然后采用特定于站点的路径:

patterns={
    'github.com': r'/(?P<user>[^/]+)/(?P<project>[^/#]+)(?:[/#]|$)',
    'sourceforge.net': r'/projects/(?P<project>)[^/]+/',
    <etc etc etc>
}

import urllib.parse
pr = urllib.parse.urlparse(url)

site = pr.hostname    # in case port is specified
parts = re.match(patterns[site], pr.path).groupdict()

路径也可以用状态机解析,而不是正则表达式,如果前面有进一步的拆分,这可能更易于管理:
(they recommend a enum instead of magic strings for states;我使用魔术字符串只是为了简化示例代码)

def parse_github(path):
    r = argparse.Namespace()
    pp = path.split('/')
    p = pp.pop(0)
    assert(p == '')
    state='user'
    for p in pp:    # we dont need to backtrack in this case,
                    # so `for' is a fitting mechanism to iterate
                    # over the parts.
                    # if we needed to backtrack, we'd have to use
                    # an index variable or a stack or something
        if state=='user':
            r.user=p
            state='project'
        else if state=='project':
            r.project==p
            state='kind'
        else if state=='kind':
            if p in {'pull','commit','blob'}:
                state=p
            else: break  #end parsing, ignore anything that's left
        else if state=='pull':
            r.pr=p
            state='pr_tab'
        <etc etc>
    return r

原则上,这里没有递归结构,所以这个可以单独用正则表达式来完成,但这很尴尬:

site_patterns = [
    r"(github\.com/)[a-zA-Z0-9_.-]+/[a-zA-Z0-9_.-]+",
    r"(bitbucket\.org/)[a-zA-Z0-9_.-]+/[a-zA-Z0-9_.-]+",
    r"(gitlab\.com/)[a-zA-Z0-9_.-]+(?:/[a-zA-Z0-9_.-]+)+",
    r"(sourceforge\.net/projects/)[a-zA-Z0-9_.-]+",
    <etc etc etc>
    ]
r_all = re.compile("("+"|".join(site_patterns)+")")   #good luck debugging this monster