study4/study4_scraper_extension/content.js

(function() {
    function extractData(doc) {
        let items = [];
        const blocks = doc.querySelectorAll('.termlist-item.contentblock');
        
        blocks.forEach(block => {
            let word = "";
            let type = "";
            let pronunciation = "";
            
            // Extract from h2
            const h2 = block.querySelector('h2.h3');
            if (h2) {
                // Word is typically the first text node
                word = Array.from(h2.childNodes)
                    .filter(node => node.nodeType === Node.TEXT_NODE)
                    .map(node => node.textContent.trim())
                    .join(' ')
                    .trim();
                
                const spans = h2.querySelectorAll('span');
                if (spans.length >= 2) {
                    type = spans[0].textContent.trim();
                    pronunciation = spans[1].textContent.trim();
                } else if (spans.length === 1) {
                    type = spans[0].textContent.trim();
                }
            }
            
            // Definition
            let definition = "";
            const defDiv = block.querySelector('.prewrap.mb-2');
            if (defDiv) {
                definition = defDiv.textContent.trim();
            }
            
            // Examples
            let examples = [];
            const exListItems = block.querySelectorAll('.termlist-item-examples li');
            exListItems.forEach(li => {
                // Ignore the audio button text (usually inside span elements)
                const text = Array.from(li.childNodes)
                    .filter(node => node.nodeType === Node.TEXT_NODE)
                    .map(node => node.textContent.trim())
                    .join(' ')
                    .trim();
                if (text) examples.push(text);
            });
            
            // Image
            let imageUrl = "";
            const img = block.querySelector('.termlist-item-images img');
            if (img) {
                imageUrl = img.getAttribute('data-src') || img.src;
                if (imageUrl && imageUrl.startsWith('/')) {
                    imageUrl = window.location.origin + imageUrl;
                }
            }
            
            // Audio (UK and US)
            let ukAudio = "";
            let usAudio = "";
            
            // The structure is usually <audio><source src="...">
            const audioSpans = block.querySelectorAll('.jq-audio-player audio');
            audioSpans.forEach(audio => {
                const source = audio.querySelector('source');
                if (source) {
                    let src = source.getAttribute('src');
                    if (src && src.startsWith('/')) {
                        src = window.location.origin + src;
                    }
                    if (src && src.includes('lang=en-GB')) {
                        ukAudio = src;
                    } else if (src && src.includes('lang=en-US')) {
                        usAudio = src;
                    }
                }
            });
            
            if (word) {
                items.push({
                    word: word,
                    type: type,
                    pronunciation: pronunciation,
                    definition: definition,
                    examples: examples,
                    imageUrl: imageUrl,
                    audio: {
                        uk: ukAudio,
                        us: usAudio
                    }
                });
            }
        });
        
        return items;
    }

    // Try to extract from the main document first
    let items = extractData(document);
    
    // If none found, see if there's an iframe (like on the lesson page)
    if (items.length === 0) {
        try {
            const iframe = document.querySelector('iframe.learncourse-iframe');
            if (iframe && iframe.contentDocument) {
                items = extractData(iframe.contentDocument);
            }
        } catch (e) {
            // Might fail due to cross-origin if running locally without properly configured URLs,
            // but on study4.com it should work.
            console.error("Cannot access iframe content:", e);
        }
    }
    
    return items;
})();