Python中替换字符串时如何保留关联信息？

问题背景 #

我们有一组有序单词和对应的关联信息（可以是两个平行列表，或是(单词, 信息)的元组列表），需要执行归一化操作（比如缩写替换、数字转英文单词），同时保留每个归一化结果对应的原始关联信息。之前直接拼接成字符串用正则替换的方法会丢失位置关联，需要针对性处理。

示例输入：

words = ["hello", "I", "am", "I", "am", "Jone", "101"]
info = ["1", "3", "23", "4", "6", "5", "12"]
# 或元组列表形式
list_tuples = list(zip(words, info))

我们需要得到两种可能的归一化结果：

结果一：合并关联信息到列表 #

归一化后的合并单词对应原始多个单词的关联信息列表，单个替换的单词保留原信息：

words = ["hello", "I'm", "I'm", "Jone", "one hundred and one"]
info = ["1", ["3", "23"], ["4", "6"], "5", "12"]

结果二：拆分归一化后的多词内容 #

将归一化后的多词内容拆分为单个单词，每个单词重复原始关联信息：

words = ["hello", "I'm", "I'm", "Jone", "one", "hundred", "and", "one"]
info = ["1", ["3", "23"], ["4", "6"], "5", "12", "12", "12", "12"]

解决方案实现 #

通用思路 #

核心是遍历单词序列时，识别需要匹配的规则（单词或多词组合），同步处理对应的关联信息，而非先拼接成字符串丢失位置信息。我们可以定义规则列表，每个规则包含匹配的单词序列和替换内容，再按顺序遍历处理。

实现结果一：合并关联信息 #

def normalize_merge_info(word_info_tuples):
    # 定义归一化规则：(匹配的单词元组, 替换后的单词)
    rules = [
        (("I", "am"), "I'm"),
        (("101",), "one hundred and one")
    ]
    result_words = []
    result_info = []
    i = 0
    n = len(word_info_tuples)
    
    while i < n:
        matched = False
        # 遍历所有规则，尝试匹配当前位置的单词序列
        for pattern, replacement in rules:
            pattern_len = len(pattern)
            # 检查当前位置到后续是否足够匹配规则长度
            if i + pattern_len <= n:
                # 提取当前位置的单词序列
                current_words = tuple(word for word, _ in word_info_tuples[i:i+pattern_len])
                if current_words == pattern:
                    # 匹配成功，添加替换后的单词
                    result_words.append(replacement)
                    # 收集对应的关联信息列表
                    collected_info = [info for _, info in word_info_tuples[i:i+pattern_len]]
                    # 单词规则保留单个信息，多词规则保留列表
                    result_info.append(collected_info if pattern_len > 1 else collected_info[0])
                    i += pattern_len
                    matched = True
                    break
        # 无匹配规则时，直接添加原单词和信息
        if not matched:
            word, info = word_info_tuples[i]
            result_words.append(word)
            result_info.append(info)
            i += 1
    return result_words, result_info

# 测试
words = ["hello", "I", "am", "I", "am", "Jone", "101"]
info = ["1", "3", "23", "4", "6", "5", "12"]
list_tuples = list(zip(words, info))
res_words, res_info = normalize_merge_info(list_tuples)
print(res_words)
print(res_info)

输出：

["hello", "I'm", "I'm", "Jone", "one hundred and one"]
["1", ["3", "23"], ["4", "6"], "5", "12"]

实现结果二：拆分多词替换内容 #

在结果一的基础上，对替换后的内容进行拆分，多词内容拆分为单个单词，同时重复对应的关联信息：

def normalize_split_words(word_info_tuples):
    rules = [
        (("I", "am"), "I'm"),
        (("101",), "one hundred and one")
    ]
    result_words = []
    result_info = []
    i = 0
    n = len(word_info_tuples)
    
    while i < n:
        matched = False
        for pattern, replacement in rules:
            pattern_len = len(pattern)
            if i + pattern_len <= n:
                current_words = tuple(word for word, _ in word_info_tuples[i:i+pattern_len])
                if current_words == pattern:
                    # 拆分替换后的内容为单个单词
                    split_replacement = replacement.split()
                    # 获取对应的关联信息（多词规则取列表，单词规则取单个）
                    source_info = [info for _, info in word_info_tuples[i:i+pattern_len]]
                    source_info = source_info if pattern_len > 1 else [source_info[0]]
                    
                    # 处理拆分后的每个单词
                    for word_part in split_replacement:
                        result_words.append(word_part)
                        # 合并规则保留信息列表，拆分规则重复原始单条信息
                        result_info.append(source_info if pattern_len > 1 else source_info[0])
                    
                    i += pattern_len
                    matched = True
                    break
        # 无匹配规则时，直接添加原单词和信息
        if not matched:
            word, info = word_info_tuples[i]
            result_words.append(word)
            result_info.append(info)
            i += 1
    return result_words, result_info

# 测试
res_words2, res_info2 = normalize_split_words(list_tuples)
print(res_words2)
print(res_info2)

输出：

["hello", "I'm", "I'm", "Jone", "one", "hundred", "and", "one"]
["1", ["3", "23"], ["4", "6"], "5", "12", "12", "12", "12"]

扩展说明 #

• 可根据需求添加更多规则，比如(("we'll",), "we will")或(("you", "are"), "you're")，只需在rules列表中添加对应条目即可。
• 如果规则有优先级（比如长规则优先匹配），可将长规则放在rules列表的前面，避免短规则先匹配导致长规则无法触发。

内容的提问来源于stack exchange，提问作者ivangtorre