recognize and parse Schema.org Markup

src/com/dom_distiller/client/SchemaOrgParser.java

+// Copyright 2014 The Chromium Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+package com.dom_distiller.client;
+
+import java.util.ArrayList;
+import java.util.HashMap;
+import java.util.List;
+import java.util.Map;
+
+import com.google.gwt.dom.client.AnchorElement;
+import com.google.gwt.dom.client.Element;
+import com.google.gwt.dom.client.ImageElement;
+import com.google.gwt.dom.client.MetaElement;
+import com.google.gwt.dom.client.Node;
+import com.google.gwt.dom.client.NodeList;
+
+/**
+ * This class recognizes and parses schema.org markup tags, and returns the properties that matter
+ * to distilled content.
+ * Schema.org markup (http://schema.org) is based on the microdata format
+ * (http://www.whatwg.org/specs/web-apps/current-work/multipage/microdata.html).
+ * For the basic Schema.org Thing type, the basic properties are: name, url, description, image.
+ * In addition, for each type that we support, we also parse more specific properties:
+ * - Article: headline (i.e. title), publisher, copyright year, copyright holder, date published,
+ *            date modified, author, article section
+ * - ImageObject: headline (i.e. title), publisher, copyright year, copyright holder, content url,
+ *                encoding format, caption, representative of page, width, height
+ * - Person: family name, given name
+ * - Organization: legal name.
+ * The value of a Schema.Org property can be a Schema.Org type, i.e. embedded.  E.g., the author or
+ * publisher of article or publisher of image could be a Schema.Org Person or Organization type;
+ * in fact, this is the reason we support Person and Organization types.
+ */
+public class SchemaOrgParser {
+    static final String NAME_PROP = "name";
+    static final String URL_PROP = "url";
+    static final String DESCRIPTION_PROP = "description";
+    static final String IMAGE_PROP = "image";
+    static final String HEADLINE_PROP = "headline";
+    static final String PUBLISHER_PROP = "publisher";
+    static final String COPYRIGHT_HOLDER_PROP = "copyrightHolder";
+    static final String COPYRIGHT_YEAR_PROP = "copyrightYear";
+    static final String CONTENT_URL_PROP = "contentUrl";
+    static final String ENCODING_FORMAT_PROP = "encodingFormat";
+    static final String CAPTION_PROP = "caption";
+    static final String REPRESENTATIVE_PROP = "representativeOfPage";
+    static final String WIDTH_PROP = "width";
+    static final String HEIGHT_PROP = "height";
+    static final String DATE_PUBLISHED_PROP = "datePublished";
+    static final String DATE_MODIFIED_PROP = "dateModified";
+    static final String AUTHOR_PROP = "author";
+    static final String CREATOR_PROP = "creator";
+    static final String SECTION_PROP = "articleSection";
+    static final String ASSOCIATED_MEDIA_PROP = "associatedMedia";
+    static final String ENCODING_PROP = "encoding";
+    static final String FAMILY_NAME_PROP = "familyName";
+    static final String GIVEN_NAME_PROP = "givenName";
+    static final String LEGAL_NAME_PROP = "legalName";
+    static final String AUTHOR_REL = "author";
+
+    enum Type {  // All these types are extended from Thing, directly or indirectly.
+        IMAGE,
+        ARTICLE,
+        PERSON,
+        ORGANIZATION,
+        UNSUPPORTED,
+    }
+
+    static class ThingItem {
+        private final Type mType;
+        private final String[] mStringPropertyNames;

[cjhopman, 2014/04/25 20:52:34]

How this all works together is rather confusing to me. 

It think that a ThingItem instance should just have a map<String, String>
mStringProperties and map<String, ThingItem> mItemProperties. Whether or not a
certain property is supported by some ThingItem type should be a class-level
thing (and they should share those supported by the base ThingItem class).

[kuan, 2014/04/29 00:23:10]

Done.  i was using map b4, but changed to arrays when u didn't want map in the
ie parser.

+        private final String[][] mItemPropertyNames;
+        private final String[] mStringProperties;
+        private final ThingItem[] mItemProperties;
+
+        // |stringPropertyNames| and |itemPropertyNames| are names of properties that this

[cjhopman, 2014/04/25 20:52:34]

nit: Use javadoc comment format

[kuan, 2014/04/29 00:23:10]

comments r not needed now.

+        // ThingItem extracts from the page.
+        // @param stringPropertyNames is a String[] of property names whose values are String type.
+        // @param itemPropertyNames is a 2-dimensional array of String where:
+        // - 1st row: array of property names whose values are of extended-ThingItem type
+        // - 2nd row: array of property names whose values are to be retrieved from the
+        //   corresponding extended-ThingItem object.
+        ThingItem(Type type, String[] stringPropertyNames, String[][] itemPropertyNames) {
+            mType = type;
+            mStringPropertyNames = stringPropertyNames;
+            mItemPropertyNames = itemPropertyNames;
+            mStringProperties = new String[mStringPropertyNames.length];
+            mItemProperties = new ThingItem[mItemPropertyNames.length];
+        }
+
+        MarkupParser.Image getImage() {
+            return null;
+        }
+
+        MarkupParser.Article getArticle() {

[cjhopman, 2014/04/25 20:52:34]

Let's get rid of these functions that only make sense on certain types (and so
the caller should only call them on the right types). So: getArticle and
isImageRepresentativeOfPage should be removed.

[kuan, 2014/04/29 00:23:10]

Done.  i was doing it this way to avoid casting ThingItem, which i thought was
discouraged as per the review for ie parser.

+            return null;
+        }
+
+        String getStringProperty(String name) {
+            // Check if property exists in |mStringProperties|.
+            for (int i = 0; i < mStringPropertyNames.length; i++) {
+                if (name.equals(mStringPropertyNames[i])) {
+                    String value = mStringProperties[i];
+                    if (value != null && !value.isEmpty()) return value;
+                    break;
+                }
+            }
+            // Otherwise, repeat for |mItemProperties|.

[cjhopman, 2014/04/25 20:52:34]

This is weird that we go looking into the properties of other items and treating
them like string properties on this item. Where do we use this and is there a
better way of doing it?

[kuan, 2014/04/29 00:23:10]

Done.  author, publisher and copyright holder can be a simple name of string
type, or a schema.org person, or a schema.org organization.  i was trying to
shield the callers from having to deal with string and item property types.  in
any case, i've removed this and changed the impl.

+            for (int i = 0; i < mItemPropertyNames.length; i++) {
+                if (!name.equals(mItemPropertyNames[i][0])) continue;
+                if (mItemProperties[i] != null) {
+                    return mItemProperties[i].getStringProperty(mItemPropertyNames[i][1]);
+                }
+                break;
+            }
+            return "";
+        }
+
+        final ThingItem getItemProperty(String name) {
+            for (int i = 0; i < mItemPropertyNames.length; i++) {
+                if (name.equals(mItemPropertyNames[i][0])) return mItemProperties[i];
+            }
+            return null;
+        }
+
+        final Type getType() { return mType; }
+
+        final boolean isSupported() { return mType != Type.UNSUPPORTED; }
+
+        final boolean isImageRepresentativeOfPage() {
+            String value = getStringProperty(REPRESENTATIVE_PROP);
+            return value.equalsIgnoreCase("true");
+        }
+
+        // Store |value| for property with |name|.
+        // @param override: set to true to override the property's value, false to keep property's
+        // first non-empty value and ignore |value|.
+        final void putStringValue(String name, String value, boolean override) {
+            for (int i = 0; i < mStringPropertyNames.length; i++) {
+                if (name.equals(mStringPropertyNames[i])) {
+                    String existing = override ? null : mStringProperties[i];
+                    if (existing == null || existing.isEmpty()) mStringProperties[i] = value;
+                    break;
+                }
+            }
+        }
+
+        // Store |value| for property with |name|, unless the property already has a non-null value,
+        // in which case, |value| will be ignored.  This means we only keep the first value.
+        final void putItemValue(String name, ThingItem value) {
+            for (int i = 0; i < mItemPropertyNames.length; i++) {
+                if (name.equals(mItemPropertyNames[i][0])) {
+                    if (mItemProperties[i] == null) mItemProperties[i] = value;
+                    break;
+                }
+            }
+        }
+    }
+
+    private final List<ThingItem> mItemScopes = new ArrayList<ThingItem>();
+    private String mAuthorFromRel = "";
+    private static final Map<String, Type> sTypeUrls;
+    private static final Map<String, String[]> sTagAttributesMap;
+    private static final String[] sEmptyStringPropertyNames = {
+        // Intentionally empty, declared so that it's initialized statically.
+    };
+    private static final String[][] sEmptyItemPropertyNames = {
+        // Intentionally empty, declared so that it's initialized statically.
+    };
+
+    static {
+        sTypeUrls = new HashMap<String, Type>();
+        sTypeUrls.put("http://schema.org/ImageObject", Type.IMAGE);
+        sTypeUrls.put("http://schema.org/Article", Type.ARTICLE);
+        sTypeUrls.put("http://schema.org/BlogPosting", Type.ARTICLE);
+        sTypeUrls.put("http://schema.org/NewsArticle", Type.ARTICLE);
+        sTypeUrls.put("http://schema.org/ScholarlyArticle", Type.ARTICLE);
+        sTypeUrls.put("http://schema.org/TechArticle", Type.ARTICLE);
+        sTypeUrls.put("http://schema.org/Person", Type.PERSON);
+        sTypeUrls.put("http://schema.org/Organization", Type.ORGANIZATION);
+        sTypeUrls.put("http://schema.org/Corporation", Type.ORGANIZATION);
+        sTypeUrls.put("http://schema.org/EducationalOrganization", Type.ORGANIZATION);
+        sTypeUrls.put("http://schema.org/GovernmentOrganization", Type.ORGANIZATION);
+        sTypeUrls.put("http://schema.org/NGO", Type.ORGANIZATION);
+
+        // The key for |sTagAttributesMap| is the tag name, while the entry value is an array of
+        // attributes in the specified tag from which to extract information:
+        // - 0th attribute: contains the value for the property specified in itemprop
+        // - 1st attribute: if available, contains the value for the author property.
+        sTagAttributesMap = new HashMap<String, String[]>();

[cjhopman, 2014/04/25 20:52:34]

nit: move this (and the sTagAttributesMap declaration) down by the function that
uses it.

[kuan, 2014/04/29 00:23:10]

Done.

+        sTagAttributesMap.put("IMG", new String[] { "SRC" });
+        sTagAttributesMap.put("AUDIO", new String[] { "SRC" });
+        sTagAttributesMap.put("EMBED", new String[] { "SRC" });
+        sTagAttributesMap.put("IFRAME", new String[] { "SRC" });
+        sTagAttributesMap.put("SOURCE", new String[] { "SRC" });
+        sTagAttributesMap.put("TRACK", new String[] { "SRC" });
+        sTagAttributesMap.put("VIDEO", new String[] { "SRC" });
+        sTagAttributesMap.put("A", new String[] { "HREF", "REL" });

[cjhopman, 2014/04/25 20:52:34]

I can't find documentation anywhere that says to use the rel= attribute.

[kuan, 2014/04/29 00:23:10]

http://schema.org/author, or from http://schema.org/Article, look for "author"
property.

[cjhopman, 2014/04/29 17:04:19]

That's interesting. Maybe you could add a comment somehere pointing to that.

[kuan, 2014/04/29 23:26:43]

Done.  comment is in parse() where the rel attribute is extracted.

+        sTagAttributesMap.put("LINK", new String[] { "HREF", "REL" });
+        sTagAttributesMap.put("AREA", new String[] { "HREF" });
+        sTagAttributesMap.put("META", new String[] { "CONTENT" });
+        sTagAttributesMap.put("TIME", new String[] { "DATETIME" });
+        sTagAttributesMap.put("OBJECT", new String[] { "DATA" });
+        sTagAttributesMap.put("DATA", new String[] { "VALUE" });
+        sTagAttributesMap.put("METER", new String[] { "VALUE" });
+    }
+
+    /**
+     * The object that extracts and verifies Schema.org markup tags from |root|.
+     */
+    public SchemaOrgParser(Element root) {
+        // TODO(kuan): Parsing all tags is pretty expensive, should we do so only lazily?
+        // If parse lazily, all get* methods will need to check for parsed state and, if necessary,
+        // parse before returning the requested properties.
+        // Note that the <html> element can also be the start of a Schema.org item, and hence needs
+        // to be parsed.
+        parse(root, null);
+    }
+
+    // Returns the first item that has the requested property value.

[cjhopman, 2014/04/25 20:52:34]

I don't think that this is something we would ever want to do. The only time we
use it is to find the headline which is a property only of articles. In that
case we should be iterating over the articles and checking each of them for a
headline.

[kuan, 2014/04/29 00:23:10]

Done.  Image also has headline, but i assume we only care abt Article.

+    String findStringProperty(String name) {
+        if (mItemScopes.isEmpty()) return "";
+        for (int i = 0; i < mItemScopes.size(); i++) {
+            String value = mItemScopes.get(i).getStringProperty(name);
+            if (!value.isEmpty()) return value;
+        }
+        return "";
+    }
+
+    ThingItem findFirstArticle() {

[cjhopman, 2014/04/25 20:52:34]

This should return an ArticleItem

[kuan, 2014/04/29 00:23:10]

Done.

+        for (int i = 0; i < mItemScopes.size(); i++) {
+            ThingItem item = mItemScopes.get(i);
+            if (item.mType == Type.ARTICLE) return item;
+        }
+        return null;
+    }
+
+    final List<ThingItem> getItemScopes() { return mItemScopes; }
+
+    final String getAuthorFromRel() { return mAuthorFromRel; }
+
+    static String concat(String first, String second) {
+        String concat = first;
+        if (!concat.isEmpty() && !second.isEmpty()) concat += " ";
+        concat += second;
+        return concat;
+    }
+
+    private void parse(Element e, ThingItem parentItem) {
+        ThingItem newItem = null;
+        boolean isItemScope = isItemScope(e);
+        // A non-null |parentItem| means we're currently parsing the elements for a schema.org type.
+        String[] propertyNames = parentItem != null ? getItemProp(e) : new String[0];
+
+        if (isItemScope) {
+            // The "itemscope" and "itemtype" attributes of |e| indicate the start of an item.
+            // Create the corresponding extended-ThingItem, and add it to the list if:
+            // 1) its type is supported, and
+            // 2) if the parent is an unsupported type, it's not an "itemprop" attribute of the
+            //    parent, based on the rule that an item is a top-level item if its element doesn't
+            //    have an itemprop attribute.
+            newItem = createItemForElement(e);
+            if (newItem != null && newItem.isSupported() &&
+                (parentItem == null || parentItem.isSupported() || propertyNames.length == 0)) {
+                mItemScopes.add(newItem);
+            }
+        }
+
+        // If parent is a supported type, parse the element for >= 1 properties in "itemprop"
+        // attribute.
+        if (propertyNames.length > 0 && parentItem.isSupported() &&
+            (newItem == null || newItem.isSupported())) {
+            for (int i = 0; i < propertyNames.length; i++) {
+                // If a new item was created above, the property value of this "itemprop" attribute
+                // is an embedded item, so add it to the parent item.
+                if (newItem != null) {
+                    parentItem.putItemValue(propertyNames[i], newItem);
+                } else {
+                   // Otherwise, extract the property value from the tag itself, and add it to the
+                   // parent item.
+                   parentItem.putStringValue(propertyNames[i], getPropertyValue(e), false);
+                }
+            }
+        }
+
+        // If <a> or <link> tags specify rel="author", extract it.
+        if (mAuthorFromRel.isEmpty()) mAuthorFromRel = getAuthorFromRelAttribute(e);
+
+        // Now, parse each child element recursively.
+        NodeList<Node> children = e.getChildNodes();
+        for (int i = 0; i < children.getLength(); i++) {
+            Node child = children.getItem(i);
+            if (child.getNodeType() != Node.ELEMENT_NODE) continue;
+            parse(Element.as(child), newItem != null ? newItem : parentItem);
+        }
+    }
+
+    private Type getItemType(Element e) {
+        // "itemtype" attribute is case-sensitive.
+        String type = e.getAttribute("ITEMTYPE");
+        return sTypeUrls.containsKey(type) ? sTypeUrls.get(type) : Type.UNSUPPORTED;
+    }
+
+    private ThingItem createItemForElement(Element e) {
+        ThingItem newItem = null;
+        Type type = getItemType(e);
+        switch (type) {
+            case IMAGE:
+                newItem = new ImageItem();
+                break;
+            case ARTICLE:
+                newItem = new ArticleItem();
+                break;
+            case PERSON:
+                newItem = new PersonItem();
+                break;
+            case ORGANIZATION:
+                newItem = new OrganizationItem();
+                break;
+            case UNSUPPORTED:
+                newItem = new UnsupportedItem();
+                break;
+            default:
+                return null;
+        }
+        return newItem;
+    }
+
+    private static class ImageItem extends ThingItem {
+        private static final String[] sStringPropertyNames = {
+            NAME_PROP,
+            URL_PROP,
+            DESCRIPTION_PROP,
+            IMAGE_PROP,
+            HEADLINE_PROP,
+            PUBLISHER_PROP,
+            COPYRIGHT_HOLDER_PROP,
+            COPYRIGHT_YEAR_PROP,
+            CONTENT_URL_PROP,
+            ENCODING_FORMAT_PROP,
+            CAPTION_PROP,
+            REPRESENTATIVE_PROP,
+            WIDTH_PROP,
+            HEIGHT_PROP,
+        };
+
+        private static final String[][] sItemPropertyNames = {
+            new String[] { PUBLISHER_PROP, NAME_PROP },
+            new String[] { COPYRIGHT_HOLDER_PROP, NAME_PROP },
+        };
+
+        ImageItem() {
+            super(Type.IMAGE, sStringPropertyNames, sItemPropertyNames);
+        }
+
+        @Override
+        String getStringProperty(String propertyName) {
+            if (!propertyName.equals(CONTENT_URL_PROP) && !propertyName.equals(URL_PROP)) {
+                return super.getStringProperty(propertyName);
+            }
+            // Returns either the value of CONTENT_URL_PROP or URL_PROP.
+            String url = super.getStringProperty(CONTENT_URL_PROP);
+            return url.isEmpty() ? super.getStringProperty(URL_PROP) : url;
+        }
+
+        @Override
+        MarkupParser.Image getImage() {
+            MarkupParser.Image image = new MarkupParser.Image();
+            image.image = getStringProperty(CONTENT_URL_PROP);
+            image.url = image.image;
+            image.type = getStringProperty(ENCODING_FORMAT_PROP);
+            image.caption = getStringProperty(CAPTION_PROP);
+            try {
+                image.width = Integer.parseInt(getStringProperty(WIDTH_PROP), 10);
+            } catch (Exception e) {
+            }
+            try {
+                image.height = Integer.parseInt(getStringProperty(HEIGHT_PROP), 10);
+            } catch (Exception e) {
+            }
+            return image;
+        }
+    }
+
+    private static class ArticleItem extends ThingItem {
+        private static final String[] sStringPropertyNames = {
+            NAME_PROP,
+            URL_PROP,
+            DESCRIPTION_PROP,
+            IMAGE_PROP,
+            HEADLINE_PROP,
+            PUBLISHER_PROP,
+            COPYRIGHT_HOLDER_PROP,
+            COPYRIGHT_YEAR_PROP,
+            DATE_MODIFIED_PROP,
+            DATE_PUBLISHED_PROP,
+            AUTHOR_PROP,
+            CREATOR_PROP,
+            SECTION_PROP,
+        };
+
+        private static final String[][] sItemPropertyNames = {
+            new String[] { PUBLISHER_PROP, NAME_PROP },
+            new String[] { COPYRIGHT_HOLDER_PROP, NAME_PROP },
+            new String[] { AUTHOR_PROP, NAME_PROP },
+            new String[] { CREATOR_PROP, NAME_PROP },
+            new String[] { ASSOCIATED_MEDIA_PROP, CONTENT_URL_PROP },
+            new String[] { ENCODING_PROP, CONTENT_URL_PROP },
+        };
+
+        ArticleItem() {
+            super(Type.ARTICLE, sStringPropertyNames, sItemPropertyNames);
+        }
+
+        @Override
+        MarkupParser.Image getImage() {
+            // If "associatedMedia" or "encoding" property exists, set the "representativeOfPage"
+            // property of the corresponding ImageItem to "true", so that that image (which will
+            // be picked up when looping through |mItemScopes|) will be the dominant one.
+            ThingItem imageItem = getItemProperty(ASSOCIATED_MEDIA_PROP);
+            if (imageItem == null) imageItem = getItemProperty(ENCODING_PROP);
+            if (imageItem != null) {
+                imageItem.putStringValue(REPRESENTATIVE_PROP, "true", true);

[cjhopman, 2014/04/25 20:52:34]

This is strange, and only seems to work if I do getImage() on the article before
processing image types.

I'd prefer to just remove setting representativeOfPage on this item and handle
it in a separate change.

[kuan, 2014/04/29 00:23:10]

Done.  i've removed setting representativeOfPage and handle the logic in
getImages(), not sure if that's what u mean by "in a separate change".

+                return null;
+            }
+
+            // Use value of IMAGE_PROP to create a MarkupParser.Image.
+            String imageUrl = getStringProperty(IMAGE_PROP);
+            if (imageUrl.isEmpty()) return null;
+            MarkupParser.Image image = new MarkupParser.Image();
+            image.image = imageUrl;
+            image.url = imageUrl;
+            return image;
+        }
+
+        @Override
+        MarkupParser.Article getArticle() {
+            MarkupParser.Article article = new MarkupParser.Article();
+            article.publishedTime = getStringProperty(DATE_PUBLISHED_PROP);
+            article.modifiedTime = getStringProperty(DATE_MODIFIED_PROP);
+            article.section = getStringProperty(SECTION_PROP);
+            String author = getStringProperty(AUTHOR_PROP);
+            if (author.isEmpty()) author = getStringProperty(CREATOR_PROP);
+            article.authors = author.isEmpty() ? new String[0] : new String[] { author };
+            return article;
+        }
+    }
+
+    private static class PersonItem extends ThingItem {
+        private static final String[] sStringPropertyNames = {
+            NAME_PROP,
+            URL_PROP,
+            DESCRIPTION_PROP,
+            IMAGE_PROP,
+            FAMILY_NAME_PROP,
+            GIVEN_NAME_PROP,
+        };
+
+        PersonItem() {
+            super(Type.PERSON, sStringPropertyNames, sEmptyItemPropertyNames);
+        }
+
+        @Override
+        String getStringProperty(String propertyName) {

[cjhopman, 2014/04/25 20:52:34]

Overriding getStringProperty like this is a little odd to me. I think it would
be a lot easier to reason about if ThingItem's string property stuff acting just
like a map.

If it is common to want a different "name" concept than the schema.org name
property, ThingItem can have a getName function that is overridable (and the
default is just the name property).

[kuan, 2014/04/29 00:23:10]

Done.  i've made getStringProperty final, but implemented the logic differently
from ur "getName" suggestion.

+            if (!propertyName.equals(NAME_PROP)) return super.getStringProperty(propertyName);
+            // Returns either the value of NAME_PROP, or concatenated values of GIVEN_NAME_PROP and
+            // FAMILY_NAME_PROP delimited by a whitespace.
+            String fullname = super.getStringProperty(NAME_PROP);
+            if (fullname.isEmpty()) {
+                fullname = concat(super.getStringProperty(GIVEN_NAME_PROP),
+                                  super.getStringProperty(FAMILY_NAME_PROP));
+            }
+            return fullname;
+        }
+    }
+
+    private static class OrganizationItem extends ThingItem {
+        private static final String[] sStringPropertyNames = {
+            NAME_PROP,
+            URL_PROP,
+            DESCRIPTION_PROP,
+            IMAGE_PROP,
+            LEGAL_NAME_PROP,
+        };
+
+        OrganizationItem() {
+            super(Type.ORGANIZATION, sStringPropertyNames, sEmptyItemPropertyNames);
+        }
+
+        @Override
+        String getStringProperty(String propertyName) {
+            if (!propertyName.equals(NAME_PROP)) return super.getStringProperty(propertyName);
+            // Returns either the value of NAME_PROP or LEGAL_NAME_PROP.
+            String name = super.getStringProperty(NAME_PROP);
+            if (name.isEmpty()) name = super.getStringProperty(LEGAL_NAME_PROP);
+            return name;
+        }
+    }
+
+    private static class UnsupportedItem extends ThingItem {
+        UnsupportedItem(){
+            super(Type.UNSUPPORTED, sEmptyStringPropertyNames, sEmptyItemPropertyNames);
+        }
+    }
+
+    private static boolean isItemScope(Element e) {
+        return e.hasAttribute("ITEMSCOPE") && e.hasAttribute("ITEMTYPE");
+    }
+
+    private static String[] getItemProp(Element e) {
+        // "itemprop" attribute is case-sensitive, and can have multiple properties.
+        String itemprop = e.getAttribute("ITEMPROP");
+        if (itemprop.isEmpty()) return new String[0];
+        String[] splits = StringUtil.split(itemprop, "\\s+");
+        return splits.length > 0 ? splits : new String[] { itemprop };
+    }
+
+    // Extracts the property value from |e|.  For some tags, the value is a specific attribute,
+    // while for others, it's the text between the start and end tags.
+    private static String getPropertyValue(Element e) {
+        String value = "";
+        String tagName = e.getTagName();
+        if (sTagAttributesMap.containsKey(tagName)) {
+            value = e.getAttribute(sTagAttributesMap.get(tagName)[0]);
+        }
+        if (value.isEmpty()) value = e.getInnerText();
+        return value;
+    }
+
+    // Extracts the author property from |e|'s "rel=author" attribute.
+    private static String getAuthorFromRelAttribute(Element e) {
+        String author = "";
+        String tagName = e.getTagName();
+        if (sTagAttributesMap.containsKey(tagName)) {
+            String[] attrs = sTagAttributesMap.get(tagName);
+            if (attrs.length > 1 && e.getAttribute(attrs[1]).equals(AUTHOR_REL)) {
+                author = e.getInnerText();
+            }
+        }
+        return author;
+    }
+}