php - 根据括号递归地按逗号展开

Question

想象这样一个字符串:

field1,field2(subfield1),field3(subfield2,subfield3),field4(),field5(subfield4(subsubfield,subsubfield2))

我想要一个这样的数组:

array(
    field1 => array(),
    field2 => array(subfield1),
    field3 => array(
        subfield2,
        subfield3
    ),
    field4 => array(),
    field5 => array(
        subfield4 => array(
            subsubfield => array(),
            subsubfield => array()
        )
    )
)

我有这个正则表达式 [a-zA-Z0-9]*\([^()]*(?:(?R)[^()]*)*\) 其中做一些输出的工作:

array(
    field1,
    field2(subfield1),
    field3(subfield2,subfield3),
    field4(),
    field5(subfield4(subsubfield,subsubfield2))
)

虽然这不是我想要的。我现在有点卡住了，但到目前为止我想到的选项是:

1. 用 preg_replace_callback 做点什么
2. 为这些层次化的逗号分隔字符串编写某种自定义解析器
3. 退出，​​喝醉，明天破案(好吧，没办法，现在就得这么做)

不管怎样，我必须遍历字段和子字段。我有一些代码使用给定的正则表达式，并在稍后需要时对其值运行相同的匹配。我想一次解析整个字符串，包括它的嵌套子字符串。

有谁知道我是如何着手做这件事的？哪个选项是最好(或更好)的方法？ (可读性 vs 资源使用 vs 复杂性 vs 等)

Answer 1

简介

你描述的问题不能用正则语言来表示，因为正则语言不能平衡括号。但是，多年来，大多数正则表达式实现都添加了一些功能，这些功能允许解析比常规语言更复杂的语言。特别是，这个问题可以通过 .NET 的平衡匹配或 PCRE's recursive expressions 来解决。 (感谢@Gumbo 在评论中指出这一点)。

但是，仅仅因为您可以做某事并不意味着您应该。将正则表达式用于此类任务的问题在于，随着您扩展元语言，修改正则表达式的难度将成倍增加。而解析器往往更具可塑性和易于扩展。

因此，您或许可以构建一系列正则表达式来涵盖输入的非病态情况，但既然可以编写解析器，为什么还要尝试呢？它们易于维护、速度极快(比正则表达式快)、易于扩展并且启动起来很有趣。

本来想念这道题是找PHP解法，所以就用JavaScript写了。我将其翻译成 PHP，并在帖子末尾留下了原始的 JavaScript 解决方案。

PHP 解决方案

function parse( $s ) {
    // we will always have a "current context".  the current context is the array we're
    // currently operating in.  when we start, this is simply an empty array.  as new
    // arrays are created, this context will change.
    $context = array();
    // since we have to keep track of how deep our context is, we keep a context stack
    $contextStack = array(&$context);
    // this accumulates the name of the current array
    $name = '';
    for( $i=0; $i<strlen($s); $i++ ) {
        switch( $s[$i] ) {
            case ',':
                // if the last array hasn't been added to the current context
                // (as will be the case for arrays lacking parens), we add it now
                if( $name!='' && !array_key_exists( $name, $context ) )
                    $context[$name] = array();
                // reset name accumulator
                $name = '';
                break;
            case '(':
                // we are entering a subcontext

                // save a new array in the current context; this will become our new context
                $context[$name] = array();
                // switch context and add to context stack
                $context = &$context[$name];
                $contextStack[] = &$context;
                // reset name accumulator
                $name = '';
                break;
            case ')':
                // we are exiting a context

                // if we haven't saved this array in the current context, do so now
                if( $name!='' && !array_key_exists( $name, $context ) )
                    $context[$name] = array();
                // we can't just assign $context the return value of array_pop because
                // that does not return a reference
                array_pop($contextStack);
                if( count($contextStack) == 0 ) throw new Exception( 'Unmatched parenthesis' );
                $context = &$contextStack[count($contextStack)-1];
                // reset name accumulator
                $name = '';
                break;
            default:
                // this is part of the field name
                $name .= $s[$i];
        }
    }
    // add any trailing arrays to the context (this will cover the case
    // where our input ends in an array without parents)
    if( $name!='' && !array_key_exists( $name, $context ) )
        $context[$name] = array();
    if( count( $contextStack ) != 1 ) throw new Exception( 'Unmatched parenthesis' );
    return array_pop( $contextStack );
}

原始 JavaScript 解决方案

function parse(s) {
    var root = { parent: null, children: [] };
    var field = { parent: root, name: '', start_idx: 0, children: [] };
    root.children.push( field );
    for( var i=0; i<s.length; i++ ) {
        switch( s[i] ) {
            case ',':
                // if this field didn't have any children, we have to set its text
                if( !field.children.length )
                    field.text = s.substr( field.start_idx, i - field.start_idx + 1 );
                // start a new field; create new field and change context
                var newfield = { parent: field.parent, name: '', start_idx: i, children:[] };
                field.parent.children.push(newfield);
                field = newfield;
                break;
            case '(':
                // start of a subfield; create subfield and change context
                var subfield = { parent: field, name: '', start_idx: i, children:[] };
                field.children.push(subfield);
                field = subfield;
                break;
            case ')':
                // end of a subfield; fill out subfield details and change context
                if( !field.parent ) throw new Error( 'Unmatched parenthesis!' );
                field.text = s.substr( field.start_idx, i - field.start_idx + 1 );
                if( field.text==='()' ) {
                    // empty subfield; pop this subfield so it doesn't clutter the parent
                    field.parent.children.pop();
                }
                field = field.parent;
                break;
            default:
                // this is part of the field name
                field.name += s[i];
                field.name = field.name.trim();
        }
    }
    return root;
}

现在我们有了您的语言的解析树，我们可以很容易地创建一些递归代码来发出您的 PHP:

function formatphp_namedarray(arr,indent,lastchild) {
    var php = indent + arr.name + ' => array(';
    if( arr.children.length ) {
        if( arr.children.length===1 && arr.children[0].length===0 ) {
            php += arr.children[0].name;
        } else {
            php += '\n';
            indent += '\t';
            for( var i=0; i<arr.children.length; i++ )
                php += formatphp_namedarray(arr.children[i],indent,i===arr.children.length-1);
            indent = indent.replace(/\t$/,'');
            php += indent;
        }
    }
    php += (lastchild?')':'),') + '\n';
    return php;
}

function formatphp(t) {
    var php = 'array(\n';
    for( var i=0; i<t.children.length; i++ )
        php += formatphp_namedarray( t.children[i], '\t', i===t.children.length-1 );
    php += ')'
    return php;
}       

在这里查看所有工作:http://jsfiddle.net/6bguY/1/