如何在PHP中拆分HTML为多段并保留标记（适配Google TTS配额）

Great question! That existing str_truncate function works perfectly for single truncations, but to split your HTML content into all valid, quota-compliant chunks (while preserving tag structure, whole words, and staying under your character limit), here's a practical, step-by-step approach:

1. Modify str_truncate to Return Critical Metadata #

The original function only returns the truncated HTML. To iterate through chunks, we need it to return extra context to maintain tag validity and track progress:

• The list of open tags that were closed at the end of the truncation (so we can re-open them in the next chunk)
• The actual plain-text character count used in the truncated chunk (to stay within your quota)
• The unprocessed portion of the original HTML (stripped of the truncated content)

Here's a adjusted version of the function that returns this metadata:

function str_truncate_with_metadata($text, $length = 100, $ending = '', $exact = true, $considerHtml = false) {
    // Keep all original function logic here...
    
    // Instead of just returning $truncate, return an array with context
    return [
        'chunk' => $truncate,
        'closed_tags' => $open_tags, // Tags that were closed to end the chunk
        'used_plain_length' => $total_length - strlen($ending),
        'remaining_html' => extract_remaining_html($text, $truncate, $open_tags) // Helper to get unprocessed content
    ];
}

// Helper to extract the unprocessed HTML after truncation
function extract_remaining_html($original, $truncated, $closed_tags) {
    // Reverse-engineer where the truncation stopped, accounting for closed tags
    $clean_truncated = preg_replace('/<\/' . implode('>|<\/', $closed_tags) . '>$/', '', $truncated);
    $remaining = substr($original, strpos($original, $clean_truncated) + strlen($clean_truncated));
    return trim($remaining);
}

2. Build an Iterative Chunking Wrapper #

Create a main function that loops through the content, using the modified truncate function to generate each chunk, and re-establishes tag context for subsequent chunks.

Example wrapper:

function split_html_into_quota_chunks($html, $max_plain_length = 5000) {
    $chunks = [];
    $current_html = $html;
    $tag_prefix = ''; // Tracks tags to re-open for the next chunk

    while (true) {
        // Calculate plain-text length of remaining content
        $remaining_plain_length = strlen(preg_replace('/<.*?>/', '', $current_html));
        
        if ($remaining_plain_length <= $max_plain_length) {
            // Add the final chunk with any necessary tag prefix
            $chunks[] = $tag_prefix . $current_html;
            break;
        }

        // Generate the next chunk
        $result = str_truncate_with_metadata(
            $tag_prefix . $current_html,
            $max_plain_length,
            '', // No ending needed since we're splitting, not truncating
            true,
            true
        );

        $chunks[] = $result['chunk'];

        // Prepare tag prefix for next chunk: re-open tags that were closed
        $tag_prefix = '';
        foreach (array_reverse($result['closed_tags']) as $tag) {
            $tag_prefix .= "<$tag>";
        }

        // Update to unprocessed content
        $current_html = $result['remaining_html'];
    }

    return $chunks;
}

3. Key Edge Cases to Handle #

• Nested Tags: Ensure that if a chunk ends mid-nested tag (like <div><span>...), the next chunk starts with <span> to maintain valid HTML.
• Entity Characters: Keep the original function's logic for counting entities as single characters to avoid over-quota issues.
• Self-Closing Tags: Ignore tags like <br/> or <img> when tracking open/closed tags, since they don't need to be re-opened.
• Word Boundaries: Maintain the $exact = true setting to avoid splitting words across chunks.

4. Test with Your Example #

Using your sample HTML and a 70-character plain-text limit, this setup would generate the 4 valid chunks you provided—each with complete HTML tags, whole words, and no broken structure.

内容的提问来源于stack exchange，提问作者alexander.khmelnitskiy