public static function parse($html, $url) { $recipe = RecipeParser_Parser_MicrodataSchema::parse($html, $url); libxml_use_internal_errors(true); $doc = new DOMDocument(); $html = mb_convert_encoding($html, 'HTML-ENTITIES', "UTF-8"); $doc->loadHTML('<?xml encoding="UTF-8">' . $html); $xpath = new DOMXPath($doc); // Ingredients $recipe->resetIngredients(); $nodes = $xpath->query('//div[@id = "recipe-ingredients"]/*'); foreach ($nodes as $node) { if ($node->nodeName == 'p') { $value = trim($node->nodeValue); // Older recipes will have ingredients jumbled into a single <p> // rather than using 'ingredients' classes. If the node value looks // like multiple lines, treat it like a section header followed by // section ingredients. $lines = explode("\n", $value); if (count($lines) > 1) { for ($i = 0; $i < count($lines); $i++) { $line = trim($lines[$i]); if ($i == 0) { $line = RecipeParser_Text::formatSectionName($line); $recipe->addIngredientsSection($line); } else { $line = trim($line); $recipe->appendIngredient($line); } } // Otherwise, we're dealing with a normal section for hrecipe, and // ingredients for the section will follow as <ul> elements. } else { $value = RecipeParser_Text::formatSectionName($value); $recipe->addIngredientsSection($value); } } else { if ($node->nodeName == 'ul') { $subnodes = $xpath->query('./li[@class = "ingredient"]', $node); foreach ($subnodes as $subnode) { $value = trim($subnode->nodeValue); $recipe->appendIngredient($value); } } } } // Notes $nodes = $xpath->query('//*[@id="recipe-intronote"]'); if ($nodes->length) { $value = $nodes->item(0)->nodeValue; $recipe->notes = RecipeParser_Text::formatAsParagraphs($value); } // Photo URL to replace og:image $nodes = $xpath->query('//img[@itemprop="image"]'); if ($nodes->length) { $photo_url = $nodes->item(0)->getAttribute("src"); $recipe->photo_url = RecipeParser_Text::formatPhotoUrl($photo_url, $url); } return $recipe; }
public static function getBookmarkAsRecipeStruct($html, $url) { // Turn off libxml errors to prevent mismatched tag warnings. libxml_use_internal_errors(true); $html = mb_convert_encoding($html, 'HTML-ENTITIES', "UTF-8"); $doc = new DOMDocument(); $doc->loadHTML('<?xml encoding="UTF-8">' . $html); $xpath = new DOMXPath($doc); // This recipe will be stored as a bookmark $recipe = new RecipeStruct(); $recipe->url = $url; $recipe->status = "bookmark"; // Find the page title $title = ""; $title_tag = ""; $title_og_meta = ""; $nodes = $xpath->query('//title'); if ($nodes->length) { $line = $nodes->item(0)->nodeValue; $line = RecipeParser_Text::formatTitle($line); if ($line) { $title_tag = $line; } } $nodes = $xpath->query('//meta[@property="og:title"]'); if ($nodes->length) { $line = $nodes->item(0)->getAttribute("content"); $line = RecipeParser_Text::formatTitle($line); if ($line) { $title_og_meta = $line; } } // Which title string to use? if ($title_og_meta) { $title = $title_og_meta; } else { if ($title_tag) { $title = $title_tag; } else { $title = "Recipe from {$url}"; } } $recipe->title = $title; // Get image from Open Graph tag $nodes = $xpath->query('//meta[@property="og:image"]'); if ($nodes->length) { $photo_url = $nodes->item(0)->getAttribute("content"); if ($photo_url) { $recipe->photo_url = RecipeParser_Text::formatPhotoUrl($photo_url, $url); } } return $recipe; }
public static function parse($html, $url) { $recipe = RecipeParser_Parser_Microformat::parse($html, $url); libxml_use_internal_errors(true); $html = mb_convert_encoding($html, 'HTML-ENTITIES', "UTF-8"); $doc = new DOMDocument(); $doc->loadHTML('<?xml encoding="UTF-8">' . $html); $xpath = new DOMXPath($doc); $hrecipe = $xpath->query('//*[@class="hrecipe"]'); if ($hrecipe->length) { $hrecipe = $hrecipe->item(0); // Title is not marked up with class="fn" $nodes = $xpath->query('.//h1', $hrecipe); if ($nodes->length) { $value = $nodes->item(0)->nodeValue; $recipe->title = RecipeParser_Text::formatTitle($value); } // Yield -- Class names are conflated $nodes = $xpath->query('.//*[@class="info yield"]', $hrecipe); if ($nodes->length) { $line = $nodes->item(0)->nodeValue; $recipe->yield = RecipeParser_Text::formatYield($line); } // Prep Times -- Class names are conflated $nodes = $xpath->query('.//*[@class="info preptime"]', $hrecipe); if ($nodes->length) { $value = $nodes->item(0)->nodeValue; $recipe->time['prep'] = RecipeParser_Times::toMinutes($value); } // Total Time / Duration -- Class names are conflated $nodes = $xpath->query('.//*[@class="info duration"]', $hrecipe); if ($nodes->length) { $value = $nodes->item(0)->nodeValue; $recipe->time['total'] = RecipeParser_Times::toMinutes($value); } } // Photo $nodes = $xpath->query('//section[@class="content-unit"]/img'); if ($nodes->length) { $photo_url = $nodes->item(0)->getAttribute('src'); if ($photo_url) { $recipe->photo_url = RecipeParser_Text::formatPhotoUrl($photo_url, $url); } } // Remove recipe title intros -- e.g. "Sunday Dinner: Pork Ribs" changes to "Pork Ribs" if (strpos($recipe->title, ": ") !== false) { $recipe->title = preg_replace("/^[^:]+: (.+)/", "\$1", $recipe->title); } return $recipe; }
public static function parse($html, $url) { $recipe = RecipeParser_Parser_Microformat::parse($html, $url); libxml_use_internal_errors(true); $html = mb_convert_encoding($html, 'HTML-ENTITIES', "UTF-8"); $doc = new DOMDocument(); $doc->loadHTML('<?xml encoding="UTF-8">' . $html); $xpath = new DOMXPath($doc); // --- Items not properly definied in Recipe.com's microformat markup. // Title -- Fallback if "fn" is not defined. if (!$recipe->title) { $nodes = $xpath->query('//*[contains(concat(" ", normalize-space(@class), " "), " heading1 ")]'); if ($nodes->length) { $recipe->title = trim($nodes->item(0)->nodeValue); } } // Photo -- Fallback if "photo" is not defined. if (!$recipe->photo_url) { $nodes = $xpath->query('//*[contains(concat(" ", normalize-space(@class), " "), " PB10 ")]/img'); if ($nodes->length) { $url = $nodes->item(0)->getAttribute('src'); $recipe->photo_url = RecipeParser_Text::formatPhotoUrl($url, $this->url); } } // Yield $nodes = $xpath->query('//*[@class="servingsize"]'); if ($nodes->length) { $line = trim($nodes->item(0)->nodeValue); $recipe->yield = RecipeParser_Text::formatYield($line); } // Credits $nodes = $xpath->query('//*[contains(concat(" ", normalize-space(@class), " "), " partnerName ")]'); if ($nodes->length) { $line = RecipeParser_Text::FormatAsOneLine($nodes->item(0)->nodeValue); $line = preg_replace('/\\s*Recipe from\\s+(.*)$/', "\$1", $line); $recipe->credits = trim($line); } return $recipe; }
public static function parse($html, $url) { // Turn off libxml errors to prevent mismatched tag warnings. libxml_use_internal_errors(true); $doc = new DOMDocument(); $html = mb_convert_encoding($html, 'HTML-ENTITIES', "UTF-8"); $doc->loadHTML('<?xml encoding="UTF-8">' . $html); $xpath = new DOMXPath($doc); $recipe = new RecipeParser_Recipe(); // Title $nodes = $xpath->query('//div[@id="detail_content"]/h1'); if ($nodes->length) { $recipe->title = trim($nodes->item(0)->nodeValue); } // Yield and Times $nodes = $xpath->query('//p[@id="yield"]'); if ($nodes->length) { $line = trim($nodes->item(0)->nodeValue); $recipe->yield = RecipeParser_Text::formatYield($line); } // Notes, instructions, and ingredients are not very well structured. $found_ingredients = false; $found_instructions = false; $nodes = $xpath->query('//div[@id="detail_content"]/*'); foreach ($nodes as $node) { // Notes -- Weird, but this is the only <p> that doesn't have attributes // on the tag. if ($node->nodeName == 'p') { if (!$node->hasAttributes()) { $recipe->notes = trim($node->nodeValue); continue; } } // Ingredients/ingredients markers if ($node->nodeName == 'h5') { $line = strtolower(trim($node->nodeValue)); if ($line == 'ingredients') { $found_ingredients = true; continue; } else { if ($line == 'instructions') { $found_instructions = true; } } } // Ingredients if ($found_ingredients && !$found_instructions) { if ($node->nodeName == 'h6') { $line = trim($node->nodeValue); $line = RecipeParser_Text::formatSectionName($line); $recipe->addIngredientsSection($line); } else { if ($node->nodeName == 'ul') { $sub_nodes = $node->childNodes; foreach ($sub_nodes as $sub) { $line = trim($sub->nodeValue); // Add spaces between quantities and units $line = preg_replace('/(\\d+)([A-Za-z]+)/', "\$1 \$2", $line); // Remove spaces before commas (not sure why this happens in their HTML) $line = str_replace(' ,', ',', $line); // Condense multiple spaces $line = str_replace(' ', ' ', $line); $recipe->appendIngredient($line); } } } } // Instructions if ($found_instructions) { if ($node->nodeName == 'ul') { $sub_nodes = $node->childNodes; foreach ($sub_nodes as $sub) { $line = trim($sub->nodeValue); $line = RecipeParser_Text::stripLeadingNumbers($line); $recipe->appendInstruction($line); } } } } // Photo $nodes = $xpath->query('//img[@class="detail"]'); if ($nodes->length) { $photo_url = $nodes->item(0)->getAttribute('src'); $recipe->photo_url = RecipeParser_Text::formatPhotoUrl($photo_url, $url); } return $recipe; }
public static function parse($html, $url) { $recipe = new RecipeParser_Recipe(); libxml_use_internal_errors(true); $doc = new DOMDocument(); $html = mb_convert_encoding($html, 'HTML-ENTITIES', "UTF-8"); $doc->loadHTML('<?xml encoding="UTF-8">' . $html); $xpath = new DOMXPath($doc); // Title $nodes = $xpath->query('//div[@id="rightCol"]/h1'); if ($nodes->length) { $recipe->title = trim($nodes->item(0)->nodeValue); } // Yield $nodes = $xpath->query('//h4[@class="detailHeader"]'); if ($nodes->length) { $line = trim($nodes->item(0)->nodeValue); $recipe->yield = RecipeParser_Text::formatYield($line); } // Notes $nodes = $xpath->query('//div[@class="dek"]'); if ($nodes->length) { $line = trim($nodes->item(0)->nodeValue); $recipe->notes = $line; } // Ingredients $nodes = $xpath->query('//ul[@class="recipe_ingredients"]/li'); foreach ($nodes as $node) { // Section names have class="ingredientSectionTitle", // ingredients themselves have no class. if ($node->hasAttributes()) { $line = trim($node->nodeValue); $line = RecipeParser_Text::formatSectionName($line); $recipe->addIngredientsSection($line); } else { $line = trim($node->nodeValue); // Add spaces between quantities and units $line = preg_replace('/(\\d+)([A-Za-z]+)/', "\$1 \$2", $line); // Remove spaces before commas (not sure why this happens in their HTML) $line = str_replace(' ,', ',', $line); // Condense multiple spaces $line = str_replace(' ', ' ', $line); $recipe->appendIngredient($line); } } // Instructions $nodes = $xpath->query('//ol[@class="recipe_instructions"]/li'); foreach ($nodes as $node) { $line = trim($node->nodeValue); $line = RecipeParser_Text::stripLeadingNumbers($line); $recipe->appendInstruction($line); } // Photo $nodes = $xpath->query('//img[@class="recipeImg"]'); if ($nodes->length) { $photo_url = $nodes->item(0)->getAttribute('src'); $recipe->photo_url = RecipeParser_Text::formatPhotoUrl($photo_url, $url); } else { // Second option for where to find recipe image $nodes = $xpath->query('//img[@id="splashImage"]'); if ($nodes->length) { $photo_url = $nodes->item(0)->getAttribute('src'); $recipe->photo_url = RecipeParser_Text::formatPhotoUrl($photo_url, $url); } } return $recipe; }
public static function parse($html, $url) { $recipe = new RecipeParser_Recipe(); libxml_use_internal_errors(true); $doc = new DOMDocument(); $html = mb_convert_encoding($html, 'HTML-ENTITIES', "UTF-8"); $doc->loadHTML('<?xml encoding="UTF-8">' . $html); $xpath = new DOMXPath($doc); // Title $nodes = $xpath->query('//*[@property="v:name"]'); if ($nodes->length) { $recipe->title = trim($nodes->item(0)->nodeValue); } // Summary $nodes = $xpath->query('//*[@property="v:summary"]'); if ($nodes->length) { $value = trim($nodes->item(0)->nodeValue); $recipe->description = $value; } // Times $searches = array('v:prepTime' => 'prep', 'v:cookTime' => 'cook', 'v:totalTime' => 'total'); foreach ($searches as $itemprop => $time_key) { $nodes = $xpath->query('//*[@property="' . $itemprop . '"]'); if ($nodes->length) { if ($value = $nodes->item(0)->getAttribute('content')) { $value = RecipeParser_Text::iso8601ToMinutes($value); } else { $value = trim($nodes->item(0)->nodeValue); $value = RecipeParser_Times::toMinutes($value); } if ($value) { $recipe->time[$time_key] = $value; } } } // Yield $nodes = $xpath->query('//*[@property="v:yield"]'); if ($nodes->length) { $line = trim($nodes->item(0)->nodeValue); $line = preg_replace('/\\s+/', ' ', $line); $recipe->yield = RecipeParser_Text::formatYield($line); } // Ingredients $nodes = null; // (data-vocabulary) $nodes = $xpath->query('//*[@rel="v:ingredient"]'); foreach ($nodes as $node) { $value = $node->nodeValue; $value = RecipeParser_Text::formatAsOneLine($value); if (empty($value)) { continue; } if (RecipeParser_Text::matchSectionName($value)) { $value = RecipeParser_Text::formatSectionName($value); $recipe->addIngredientsSection($value); } else { $recipe->appendIngredient($value); } } // Instructions $found = false; // Some sites will use an "instruction" class for each line. if (!$found) { $nodes = $xpath->query('//*[@property="v:instructions"]//*[@property="v:instruction"]'); if ($nodes->length) { RecipeParser_Text::parseInstructionsFromNodes($nodes, $recipe); $found = true; } } // Look for markup that uses <li>, <p> or other tags for each instruction. $search_sub_nodes = array("p", "li"); while (!$found && ($tag = array_pop($search_sub_nodes))) { $nodes = $xpath->query('//*[@property="v:instructions"]//' . $tag); if ($nodes->length) { RecipeParser_Text::parseInstructionsFromNodes($nodes, $recipe); $found = true; } } // Either multiple instrutions nodes, or one node with a blob of text. if (!$found) { $nodes = $xpath->query('//*[@property="v:instructions"]'); if ($nodes->length > 1) { // Multiple nodes RecipeParser_Text::parseInstructionsFromNodes($nodes, $recipe); $found = true; } else { if ($nodes->length == 1) { // Blob $str = $nodes->item(0)->nodeValue; RecipeParser_Text::parseInstructionsFromBlob($str, $recipe); $found = true; } } } // Photo $photo_url = ""; $nodes = $xpath->query('//*[@rel="v:photo"]'); if ($nodes->length) { $photo_url = $nodes->item(0)->getAttribute('src'); } if (!$photo_url) { // for <img> as sub-node of rel="v:photo" $nodes = $xpath->query('//*[@rel="v:photo"]//img'); if ($nodes->length) { $photo_url = $nodes->item(0)->getAttribute('src'); } } if ($photo_url) { $recipe->photo_url = RecipeParser_Text::formatPhotoUrl($photo_url, $url); } // Credits $nodes = $xpath->query('//*[@property="v:author"]'); if ($nodes->length) { $line = $nodes->item(0)->nodeValue; $recipe->credits = RecipeParser_Text::formatCredits($line); } return $recipe; }
public static function parse($html, $url) { $recipe = new RecipeParser_Recipe(); libxml_use_internal_errors(true); $doc = new DOMDocument(); $html = mb_convert_encoding($html, 'HTML-ENTITIES', "UTF-8"); $doc->loadHTML('<?xml encoding="UTF-8">' . $html); $xpath = new DOMXPath($doc); $microdata = null; $nodes = $xpath->query('//*[contains(@itemtype, "//schema.org/Recipe") or contains(@itemtype, "//schema.org/recipe")]'); if ($nodes->length) { $microdata = $nodes->item(0); } // Parse elements if ($microdata) { // Title $nodes = $xpath->query('.//*[@itemprop="name"]', $microdata); if ($nodes->length) { $value = trim($nodes->item(0)->nodeValue); $recipe->title = RecipeParser_Text::formatTitle($value); } // Summary $nodes = $xpath->query('.//*[@itemprop="description"]', $microdata); if ($nodes->length) { $value = $nodes->item(0)->nodeValue; $value = RecipeParser_Text::formatAsParagraphs($value); $recipe->description = $value; } // Times $searches = array('prepTime' => 'prep', 'cookTime' => 'cook', 'totalTime' => 'total'); foreach ($searches as $itemprop => $time_key) { $nodes = $xpath->query('.//*[@itemprop="' . $itemprop . '"]', $microdata); if ($nodes->length) { if ($value = $nodes->item(0)->getAttribute('content')) { $value = RecipeParser_Text::iso8601ToMinutes($value); } else { if ($value = $nodes->item(0)->getAttribute('datetime')) { $value = RecipeParser_Text::iso8601ToMinutes($value); } else { $value = trim($nodes->item(0)->nodeValue); $value = RecipeParser_Times::toMinutes($value); } } if ($value) { $recipe->time[$time_key] = $value; } } } // Yield $nodes = $xpath->query('.//*[@itemprop="recipeYield"]', $microdata); if (!$nodes->length) { $nodes = $xpath->query('.//*[@itemprop="recipeyield"]', $microdata); } if ($nodes->length) { if ($nodes->item(0)->hasAttribute('content')) { $line = $nodes->item(0)->getAttribute('content'); } else { $line = $nodes->item(0)->nodeValue; } $recipe->yield = RecipeParser_Text::formatYield($line); } // Ingredients $nodes = $xpath->query('//*[@itemprop="ingredients"]'); foreach ($nodes as $node) { $value = $node->nodeValue; $value = RecipeParser_Text::formatAsOneLine($value); if (empty($value)) { continue; } if (strlen($value) > 150) { // probably a mistake, like a run-on of existing ingredients? continue; } if (RecipeParser_Text::matchSectionName($value)) { $value = RecipeParser_Text::formatSectionName($value); $recipe->addIngredientsSection($value); } else { $recipe->appendIngredient($value); } } // Instructions $found = false; // Look for markup that uses <li> tags for each instruction. if (!$found) { $nodes = $xpath->query('//*[@itemprop="recipeInstructions"]//li'); if ($nodes->length) { RecipeParser_Text::parseInstructionsFromNodes($nodes, $recipe); $found = true; } } // Look for instructions as direct descendents of "recipeInstructions". if (!$found) { $nodes = $xpath->query('//*[@itemprop="recipeInstructions"]/*'); if ($nodes->length) { RecipeParser_Text::parseInstructionsFromNodes($nodes, $recipe); $found = true; } } // Some sites will use an "instruction" class for each line. if (!$found) { $nodes = $xpath->query('.//*[@itemprop="recipeInstructions"]//*[contains(concat(" ", normalize-space(@class), " "), " instruction ")]'); if ($nodes->length) { RecipeParser_Text::parseInstructionsFromNodes($nodes, $recipe); $found = true; } } // Either multiple recipeInstructions nodes, or one node with a blob of text. if (!$found) { $nodes = $xpath->query('.//*[@itemprop="recipeInstructions"]'); if ($nodes->length > 1) { // Multiple nodes RecipeParser_Text::parseInstructionsFromNodes($nodes, $recipe); $found = true; } else { if ($nodes->length == 1) { // Blob $str = $nodes->item(0)->nodeValue; RecipeParser_Text::parseInstructionsFromBlob($str, $recipe); $found = true; } } } // Photo $photo_url = ""; if (!$photo_url) { // try to find open graph url $nodes = $xpath->query('//meta[@property="og:image"]'); if ($nodes->length) { $photo_url = $nodes->item(0)->getAttribute('content'); } } if (!$photo_url) { $nodes = $xpath->query('.//*[@itemprop="image"]', $microdata); if ($nodes->length) { $photo_url = $nodes->item(0)->getAttribute('src'); } } if (!$photo_url) { // for <img> as sub-node of class="photo" $nodes = $xpath->query('.//*[@itemprop="image"]//img', $microdata); if ($nodes->length) { $photo_url = $nodes->item(0)->getAttribute('src'); } } if ($photo_url) { $recipe->photo_url = RecipeParser_Text::formatPhotoUrl($photo_url, $url); } // Credits $line = ""; $nodes = $xpath->query('.//*[@itemprop="author"]', $microdata); if ($nodes->length) { $line = $nodes->item(0)->nodeValue; } $nodes = $xpath->query('.//*[@itemprop="publisher"]', $microdata); if ($nodes->length) { $line = $nodes->item(0)->nodeValue; } $recipe->credits = RecipeParser_Text::formatCredits($line); } return $recipe; }
public static function parse($html, $url) { $recipe = new RecipeParser_Recipe(); libxml_use_internal_errors(true); $doc = new DOMDocument(); $html = mb_convert_encoding($html, 'HTML-ENTITIES', "UTF-8"); $doc->loadHTML('<?xml encoding="UTF-8">' . $html); $xpath = new DOMXPath($doc); // Title $nodes = $xpath->query('//h3[@class = "title"]'); if ($nodes->length) { $value = trim($nodes->item(0)->nodeValue); $recipe->title = $value; } // Cook times $nodes = $xpath->query('//div[@class = "recipe-metadata"]/ul/li'); foreach ($nodes as $node) { $sub_nodes = $node->childNodes; $key = null; $value = null; foreach ($sub_nodes as $sub_node) { if ($sub_node->nodeName == 'h5') { $key = trim($sub_node->nodeValue); } if ($sub_node->nodeName == 'p') { $value = trim($sub_node->nodeValue); } } // Inspect keys/values we've found. if ($key == 'Total Time:') { $value = self::cleanupTime($value); $recipe->time['total'] = RecipeParser_Times::toMinutes($value); } if ($key == 'Prep Time:') { $value = self::cleanupTime($value); $recipe->time['prep'] = RecipeParser_Times::toMinutes($value); } } $node_list = $xpath->query('//dd[@class = "preptime"]'); if ($node_list->length) { $value = $node_list->item(0)->nodeValue; $recipe->time['prep'] = RecipeParser_Times::toMinutes($value); } $node_list = $xpath->query('//dd[@class = "cooktime"]'); if ($node_list->length) { $value = $node_list->item(0)->nodeValue; $recipe->time['cook'] = RecipeParser_Times::toMinutes($value); } $node_list = $xpath->query('//dd[@class = "duration totaltime special"]'); if ($node_list->length) { $value = $node_list->item(0)->nodeValue; $recipe->time['total'] = RecipeParser_Times::toMinutes($value); } // Ingredients, Yield, Description, Notes, etc. $nodes = $xpath->query('//div[@class = "recipe-body"]/*'); $section_title = null; foreach ($nodes as $node) { // Section titles if ($node->nodeName == 'h4') { $value = $node->nodeValue; $value = trim(strtolower($value)); $section_title = $value; continue; } $in_section = false; if ($node->nodeName == 'div') { // Ensure that we're in a <div class="section"> node. foreach ($node->attributes as $attr_name => $attr_node) { if ($attr_name == 'class' && $attr_node->value == 'section') { $in_section = true; } } if (!$in_section) { continue; } // Description should be first text, before any section titles. if (!$section_title) { $value = $node->nodeValue; $value = preg_replace("/^(Drink\\:|Top Chef).*\$/m", '', $value); $value = str_replace("\n\n", "\n", $value); $value = trim($value); $recipe->description = $value; // Yield } else { if ($section_title == 'yield') { $value = trim($node->nodeValue); $recipe->yield = $value; // Notes } else { if ($section_title == 'notes') { $value = trim($node->nodeValue); $value = str_replace("\n\n", "\n", $value); $recipe->notes = $value; // Ingredients } else { if ($section_title == 'ingredients') { $sub_nodes = $node->childNodes; foreach ($sub_nodes as $sub_node) { if ($sub_node->nodeName == 'h5') { $value = RecipeParser_Text::formatSectionName($sub_node->nodeValue); $recipe->addIngredientsSection($value); } else { if ($sub_node->nodeName == 'ul') { $li_nodes = $sub_node->childNodes; foreach ($li_nodes as $li_node) { $value = trim($li_node->nodeValue); $recipe->appendIngredient($value); } } } } // Instructions } else { if ($section_title == 'directions') { $sub_nodes = $node->childNodes; foreach ($sub_nodes as $sub_node) { $value = trim($sub_node->nodeValue); // Section titles appear in all-caps. if ($value && ($value == strtoupper($value) || preg_match('/:$/', $value))) { $value = RecipeParser_Text::formatSectionName($value); $recipe->addInstructionsSection($value); } else { $value = RecipeParser_Text::stripLeadingNumbers($value); $recipe->appendInstruction($value); } } } } } } } } } // Source / Chef $nodes = $xpath->query('//div[@class = "recipe-sidebar"]/div/*'); $section_title = null; $chef_name = null; $show_name = 'Bravo TV'; foreach ($nodes as $node) { if ($node->nodeName == 'h4') { $value = trim($node->nodeValue); $section_title = strtolower($value); continue; } if ($node->nodeName == 'small') { if ($section_title == 'chef' || $section_title == 'author') { $value = trim($node->nodeValue); $chef_name = $value; break; } } } $nodes = $xpath->query('//div[@class = "section"]/p[1]'); if ($nodes->length) { $value = $nodes->item(0)->nodeValue; if (strpos($value, 'Top Chef Masters') !== false) { $show_name = 'Top Chef Masters'; } else { if (strpos($value, 'Top Chef') !== false) { $show_name = 'Top Chef'; } } } $recipe->credits = $chef_name . ', ' . $show_name; $nodes = $xpath->query('//div[@class = "recipe-header clearfix"]//img'); if ($nodes->length) { $photo_url = $nodes->item(0)->getAttribute('src'); $photo_url = str_replace('/medium/', '/original/', $photo_url); $recipe->photo_url = RecipeParser_Text::formatPhotoUrl($photo_url, $url); } return $recipe; }
public static function parse($html, $url) { if (strpos($url, "www.nytimes.com/recipes/") !== false) { // // "RECIPES" SECTION // $recipe = new RecipeParser_Recipe(); libxml_use_internal_errors(true); $doc = new DOMDocument(); $html = mb_convert_encoding($html, 'HTML-ENTITIES', "UTF-8"); $doc->loadHTML('<?xml encoding="UTF-8">' . $html); $xpath = new DOMXPath($doc); // Title $nodes = $xpath->query('//h1[@class="recipe-title recipeName"]'); if ($nodes->length) { $value = $nodes->item(0)->nodeValue; $value = RecipeParser_Text::formatTitle($value); $recipe->title = $value; } // Yield $nodes = $xpath->query('//*[@itemprop="recipeYield"]'); if ($nodes->length) { $value = $nodes->item(0)->nodeValue; $value = RecipeParser_Text::formatYield($value); $recipe->yield = $value; } // Ingredients $nodes = $xpath->query('//div[@class="ingredientsGroup"]/*'); foreach ($nodes as $node) { if ($node->nodeName == "h3") { $value = trim($node->nodeValue); if (!preg_match('/^Ingredients:?$/i', $value)) { $value = RecipeParser_Text::formatSectionName($value); $recipe->addIngredientsSection($value); } } else { foreach ($node->childNodes as $child) { $value = trim($child->nodeValue); $recipe->appendIngredient($value); } } } // Instructions $nodes = $xpath->query('//*[@itemprop="recipeInstructions"]/dd'); foreach ($nodes as $node) { $value = $node->nodeValue; $value = RecipeParser_Text::formatAsOneLine($value); $recipe->appendInstruction($value); } // Notes if (!$recipe->notes) { $nodes = $xpath->query('//div[@class="yieldNotesGroup"]//*[@class="note"]'); if ($nodes->length) { $value = trim($nodes->item(0)->nodeValue); $value = preg_replace("/^Notes?:?\\s*/i", '', $value); $recipe->notes = trim($value); } } } else { // // DINING SECTION RECIPES // $recipe = new RecipeParser_Recipe(); libxml_use_internal_errors(true); $doc = new DOMDocument(); $html = mb_convert_encoding($html, 'HTML-ENTITIES', "UTF-8"); $doc->loadHTML('<?xml encoding="UTF-8">' . $html); $xpath = new DOMXPath($doc); // Title $nodes = $xpath->query('//div[@id = "article"]//h1'); if ($nodes->length) { $value = trim($nodes->item(0)->nodeValue); $recipe->title = $value; } // Time and Yield $nodes = $xpath->query('//div[@id = "article"]//p'); foreach ($nodes as $node) { $text = trim($node->nodeValue); if (preg_match('/^Yield:? (.+)/', $text, $m)) { $recipe->yield = RecipeParser_Text::formatYield($m[1]); } else { if (preg_match('/^Time:? (.+)/', $text, $m)) { $str = trim($m[1]); $str = preg_replace('/About (.+)/', '$1', $str); $str = preg_replace('/(.+) plus.*/', '$1', $str); $recipe->time['total'] = RecipeParser_Times::toMinutes($str); } } } // Ingredients $nodes = $xpath->query('//div[@class="recipeIngredientsList"]/p'); foreach ($nodes as $node) { $line = trim($node->nodeValue); // Section names if ($line && $line == strtoupper($line)) { $line = RecipeParser_Text::formatSectionName($line); $recipe->addIngredientsSection($line); continue; } $recipe->appendIngredient($line); } // Instructions and notes $nodes = $xpath->query('//div[@class="articleBody"]//p'); if (!$nodes->length) { $nodes = $xpath->query('//div[@id="articleBody"]//p'); } $notes = ''; $in_notes_section = false; foreach ($nodes as $node) { $line = trim($node->nodeValue); // Skip some of the useless lines if (preg_match('/^(Adapted from|Time|Yield)/i', $line)) { continue; } // Instructions start with line numbers if (!$in_notes_section && preg_match('/^\\d+\\./', $line)) { $line = RecipeParser_Text::stripLeadingNumbers($line); $recipe->appendInstruction($line); continue; } // Look for lines that start the notes section. $note = ''; if (preg_match('/^Notes?:?(.*)/i', $line, $m)) { $in_notes_section = true; $note = trim($m[1]); } else { if ($in_notes_section) { $note = $line; } } if ($note) { $notes .= $note . "\n\n"; } } if ($notes) { $notes = str_replace(" ", " ", $notes); // Some unnecessary spaces $notes = trim($notes); $recipe->notes = $notes; } // Photo $nodes = $xpath->query('//div[@class="image"]//img'); if ($nodes->length) { $photo_url = $nodes->item(0)->getAttribute('src'); $photo_url = str_replace('-articleInline.jpg', '-popup.jpg', $photo_url); $recipe->photo_url = RecipeParser_Text::formatPhotoUrl($photo_url, $url); } } return $recipe; }
public static function parse($html, $url) { $recipe = new RecipeParser_Recipe(); libxml_use_internal_errors(true); $html = mb_convert_encoding($html, 'HTML-ENTITIES', "UTF-8"); $doc = new DOMDocument(); $doc->loadHTML('<?xml encoding="UTF-8">' . $html); $xpath = new DOMXPath($doc); // Title $nodes = $xpath->query('//h1[@itemprop="name"]'); if ($nodes->length) { $value = trim($nodes->item(0)->nodeValue); $recipe->title = $value; } // Times and yield // <meta content="PT3H30M" itemprop="totalTime"> $nodes = $xpath->query('//meta[@itemprop="totalTime"]'); if ($nodes->length) { if ($value = $nodes->item(0)->getAttribute('content')) { $value = RecipeParser_Text::iso8601ToMinutes($value); $recipe->time['total'] = $value; } } $nodes = $xpath->query('//*[@itemprop="recipeYield"]'); if ($nodes->length) { $value = $nodes->item(0)->nodeValue; $recipe->yield = RecipeParser_Text::formatYield($value); } // Ingredients $nodes = $xpath->query('//div[@id = "ingredients"]/*'); foreach ($nodes as $node) { if ($node->nodeName == 'h2') { $value = trim($node->nodeValue); $value = RecipeParser_Text::formatSectionName($value); if ($value != "Ingredients") { $recipe->addIngredientsSection($value); } } else { if ($node->nodeName == 'ol') { $subnodes = $xpath->query('./li/span', $node); foreach ($subnodes as $subnode) { $value = trim($subnode->nodeValue); $recipe->appendIngredient($value); } } } } // Instructions $nodes = $xpath->query('//div[@id = "directions"]/ol/li'); foreach ($nodes as $node) { $value = trim($node->nodeValue); $value = RecipeParser_Text::stripLeadingNumbers($value); $parts = self::splitDirections($value); if ($parts['section']) { $parts['section'] = RecipeParser_Text::formatSectionName($parts['section']); $recipe->addInstructionsSection($parts['section']); } $recipe->appendInstruction($parts['direction']); } // Notes $nodes = $xpath->query('//div[@id = "directions"]/div[@id = "endnotes"]'); if ($nodes->length) { $value = trim($nodes->item(0)->nodeValue); $recipe->notes = $value; } // Photo $nodes = $xpath->query('//img[@itemprop="image"]'); if ($nodes && $nodes->item(0)) { $photo_url = $nodes->item(0)->getAttribute('src'); if (strpos($photo_url, 'default-recipe-image.gif') === false && strpos($photo_url, 'placeholder.gif') === false) { $recipe->photo_url = RecipeParser_Text::formatPhotoUrl($photo_url, $url); } } return $recipe; }
public static function parse($html, $url) { $recipe = new RecipeParser_Recipe(); libxml_use_internal_errors(true); $html = mb_convert_encoding($html, 'HTML-ENTITIES', "UTF-8"); $doc = new DOMDocument(); $doc->loadHTML('<?xml encoding="UTF-8">' . $html); $xpath = new DOMXPath($doc); $hrecipe = null; if (!$hrecipe) { $nodes = $xpath->query('.//*[contains(concat(" ", normalize-space(@class), " "), " hrecipe ")]'); if ($nodes->length) { $hrecipe = $nodes->item(0); } } if (!$hrecipe) { $nodes = $xpath->query('.//*[contains(concat(" ", normalize-space(@class), " "), " hRecipe ")]'); if ($nodes->length) { $hrecipe = $nodes->item(0); } } if ($hrecipe) { // Title $nodes = $xpath->query('.//*[contains(concat(" ", normalize-space(@class), " "), " fn ")]', $hrecipe); if ($nodes->length) { $line = $nodes->item(0)->nodeValue; $recipe->title = RecipeParser_Text::formatTitle($line); } // Summary $nodes = $xpath->query('.//*[@class="summary"]', $hrecipe); if ($nodes->length) { $line = $nodes->item(0)->nodeValue; $recipe->description = RecipeParser_Text::formatAsParagraphs($line); } // Credits $nodes = $xpath->query('.//*[@class="author"]', $hrecipe); if ($nodes->length) { $line = $nodes->item(0)->nodeValue; $recipe->credits = RecipeParser_Text::formatCredits($line); } // Photo $photo_url = ""; $nodes = $xpath->query('.//*[contains(concat(" ", normalize-space(@class), " "), " photo ")]', $hrecipe); if ($nodes->length) { $photo_url = $nodes->item(0)->getAttribute('src'); } if (!$photo_url) { // for <img> as sub-node of class="photo" $nodes = $xpath->query('.//*[contains(concat(" ", normalize-space(@class), " "), " photo ")]//img', $hrecipe); if ($nodes->length) { $photo_url = $nodes->item(0)->getAttribute('src'); } } if ($photo_url) { $recipe->photo_url = RecipeParser_Text::formatPhotoUrl($photo_url, $url); } // Yield $nodes = $xpath->query('.//*[@class="yield"]', $hrecipe); if ($nodes->length) { $line = $nodes->item(0)->nodeValue; $recipe->yield = RecipeParser_Text::formatYield($line); } // Prep Times $nodes = $xpath->query('.//*[@class="prepTime"]//*[@class="value-title"]', $hrecipe); if ($nodes->length) { $value = $nodes->item(0)->getAttribute('title'); $recipe->time['prep'] = RecipeParser_Text::iso8601ToMinutes($value); } else { $nodes = $xpath->query('.//*[@class="preptime"]', $hrecipe); if ($nodes->length) { $value = $nodes->item(0)->nodeValue; $recipe->time['prep'] = RecipeParser_Times::toMinutes($value); } } // Cook Times $nodes = $xpath->query('.//*[@class="cookTime"]//*[@class="value-title"]', $hrecipe); if ($nodes->length) { $value = $nodes->item(0)->getAttribute('title'); $recipe->time['cook'] = RecipeParser_Text::iso8601ToMinutes($value); } else { $nodes = $xpath->query('.//*[@class="cooktime"]', $hrecipe); if ($nodes->length) { $value = $nodes->item(0)->nodeValue; $recipe->time['cook'] = RecipeParser_Times::toMinutes($value); } } // Total Time / Duration $nodes = $xpath->query('.//*[@class="totalTime"]//*[@class="value-title"]', $hrecipe); if ($nodes->length) { $value = $nodes->item(0)->getAttribute('title'); $recipe->time['total'] = RecipeParser_Text::iso8601ToMinutes($value); } else { $nodes = $xpath->query('.//*[@class="duration"]//*[@class="value-title"]', $hrecipe); if ($nodes->length) { $value = $nodes->item(0)->getAttribute('title'); $recipe->time['total'] = RecipeParser_Text::iso8601ToMinutes($value); } else { $nodes = $xpath->query('.//*[@class="duration"]', $hrecipe); if ($nodes->length) { $value = $nodes->item(0)->nodeValue; $recipe->time['total'] = RecipeParser_Times::toMinutes($value); } } } // Ingredients $nodes = $xpath->query('//*[contains(concat(" ", normalize-space(@class), " "), " ingredient ")]'); foreach ($nodes as $node) { $line = $node->nodeValue; $line = trim($line); $line = RecipeParser_Text::formatAsOneLine($line); // Skip lines that contain no word-like characters (sometimes used as section dividers). if (!preg_match("/\\w/", $line)) { continue; } // Section name delineated with dashes. E.g. "---Cake---" if (preg_match('/^\\-+([^\\-]{1}.*[^\\-]{1})\\-+$/', $line, $m)) { $line = RecipeParser_Text::formatSectionName($m[1]); $recipe->addIngredientsSection($line); continue; } // Section name with colon. if (preg_match('/^(.+)\\:$/', $line, $m)) { $line = RecipeParser_Text::formatSectionName($m[1]); $recipe->addIngredientsSection($line); continue; } $recipe->appendIngredient($line); } // Instructions $found = false; // Look for usage of <li> to denote each step of the instructions. if (!$found) { $nodes = $xpath->query('//*[contains(concat(" ", normalize-space(@class), " "), " instructions ")]//li'); if ($nodes->length) { RecipeParser_Text::parseInstructionsFromNodes($nodes, $recipe); $found = true; } } // Look for "instruction class for each step of the instructions. if (!$found) { $query = '//*[contains(concat(" ", normalize-space(@class), " "), " instructions ")]' . '//*[contains(concat(" ", normalize-space(@class), " "), " instruction ")]'; $nodes = $xpath->query($query); if ($nodes->length) { RecipeParser_Text::parseInstructionsFromNodes($nodes, $recipe); $found = true; } } // Default. Multiple instructions nodes, or one with a blob of text. if (!$found) { $nodes = $xpath->query('//*[contains(concat(" ", normalize-space(@class), " "), " instructions ")]'); if ($nodes->length > 1) { // Multiple nodes RecipeParser_Text::parseInstructionsFromNodes($nodes, $recipe); $found = true; } else { if ($nodes->length == 1) { // Blob $str = $nodes->item(0)->nodeValue; RecipeParser_Text::parseInstructionsFromBlob($str, $recipe); $found = true; } } } } return $recipe; }