正则表达式找不到xml标签内的字符串？

Question

Trying to find regex for this question :

PHP parsing xml file error

Trying to match "137b" in the following string , using negative lookahead and lookbehind assertions :

<Rate Symbol="EURTRY">
    <Bid>2.29443</Bid>
    <Ask>2.29562</Ask>
    <High>2.29841</High>
    <Low>2.28999</Low>

 137b

 <Direction>1</Direction>
    <Last>23:29:11</Last>
</Rate>

Can anyone please point out why this regex is not working :

(?<!(<\w+>))[a-zA-Z0-9_\.:]+(?!(</\w+>))

Intention : A string containing "a-zA-Z0-9_.:" not preceded and followed by an XML tag, so it should have matched "137b" , but it does not.

Here is a link to the regex : http://regexr.com?32rk4

Whereas the same regex (<\w+>)[a-zA-Z0-9_\.:]+(</\w+>) without negative assertions correctly matches all the strings WITHIN xml tags.

http://regexr.com?32rk7

Answer 1

PHP won't let you use a lookbehind for this, but lookbehind wouldn't be the best tool for the job anyway. (It almost never is.) You should be able to solve the problem with just a lookahead. It will be a lot easier if you can make certain assumptions about the document structure. For example, can you be sure the enclosing node is always named Rate, and that its child nodes will never have children of their own (attributes or elements)? In other words, you'll never see something like this:

<Rate Symbol="EURUSD">
    <Bid>1.27554</Bid>
        <foo>bar</foo>
    <Ask foo="bar">1.27578</Ask>
</Rate>

If so, you can use a positive lookahead to match any number of complete child nodes followed by a closing </Rate> tag:

[a-zA-Z0-9_.:]++(?=\s*(?><(\w+)>[^<]*</\1>\s*)*+</Rate>)

To explain:

[a-zA-Z0-9_.:]++
(?=
  \s*
  (?>
    <(\w+)>       # match an opening tag and capture its name
    [^<]*         # consume the content
    </\1>         # match the closing tag
    \s*
  )*+           # do this zero or more times
  </Rate>       # confirm we're inside a <Rate> element
)

This could even be expanded to deal with the other junk you mentioned in your original question, but the regex gets so ugly, I don't think it's worth it.