7.2. Article Iterator

Plugin Key

plugin_article_iterator_factory

Plugin Value Type

String

Plugin Value Type

The string is the fully-qualified name of a Java class implementing the org.lockss.plugin.ArticleIteratorFactory interface.

Sample
<entry>
  <string>plugin_article_iterator_factory</string>
  <string>edu.example.plugin.publisherx.PublisherXArticleIteratorFactory</string>
</entry>
Description

The article iterator is part of the metadata-extraction pipeline. Its function is enumerate the articles (where "article" is meant as "item" or "object") in the archival unit (AU). Each article is represented by an ArticleFiles instance.

Rather than traversing the AU's URLs manually through the ArchivalUnit interface and implementing typical inner workings of a Java Iterator, many article iterators make use of utility classes available in the LOCKSS software, such as SubTreeArticleIterator and SubTreeArticleIteratorBuilder.

7.2.1. ArticleFiles

An org.lockss.plugin.ArticleFiles instance groups the main URLs of an article (item) together and labels them. It is a mapping from roles to URLs [1] (which are internally represented as objects implementing org.lockss.plugin.CachedUrl).

Roles are arbitrary strings, but many typical role strings are defined as constants in the ArticleFiles class, for example:

  • ArticleFiles.ROLE_FULL_TEXT_HTML: a URL for the full text of a work, in HTML form.

  • ArticleFiles.ROLE_FULL_TEXT_PDF: a URL for the full text of a work, in PDF form.

  • ArticleFiles.ROLE_FULL_TEXT_EPUB: a URL for the full text of a work, in EPUB form.

  • ArticleFiles.ROLE_FULL_TEXT_XML: a URL for the full text of a work, in XML form.

  • ArticleFiles.ROLE_ABSTRACT: a URL for a work's abstract.

  • ArticleFiles.ROLE_REFERENCES: a URL for a work's list of works cited.

  • ArticleFiles.ROLE_FIGURES: a URL for a landing page of the work's figures and illustrations.

  • ArticleFiles.ROLE_TABLES: a URL for a landing page of the work's tables.

  • ArticleFiles.ROLE_SUPPLEMENTARY_MATERIALS: a URL for a landing page of the work's supplementary materials.

  • ArticleFiles.ROLE_CITATION: a URL for a landing page for the work's citation files.

  • ArticleFiles.ROLE_CITATION_BIBTEX: a URL for a BibTeX citation file for the work.

  • ArticleFiles.ROLE_CITATION_ENDNOTE: a URL for an EndNote citation file for the work.

  • ArticleFiles.ROLE_CITATION_RIS: a URL for a RIS citation file for the work.

  • ArticleFiles.ARTICLE_METADATA: a URL from which metadata for the work can be found.

In addition to the mapping from roles to URL, one URL has a special status in the ArticleFiles instance as the designated, "best" full text URL for the work. It is referred to as the full text URL or full text CU (for CachedUrl) of the article, and is set via the setFullTextCu(...) method. In plugins written by the LOCKSS Program, for articles with multiple full text representations, HTML is favored above all, then PDF, then EPUB, and lastly XML (in subjective order of richness of rendering experience in a Web browser).

7.2.2. SubTreeArticleIterator

The org.lockss.plugin.SubTreeArticleIterator class implements Iterator<ArticleFiles> and can be returned by an article iterator factory. It traverses an AU's URLs, restricting them in various ways according to a specification. These restrictions include considering only certain subdirectory trees (hence the name), applying regular expressions, selecting a media type, or applying a custom condition.

The SubTreeArticleIterator.Spec class contains the specification:

  • The specification identifies root URLs, limiting which AU URLs are enumerated to those under these root URLs only. This restriction applies to the directory structure, not the URL strings -- in effect, root URLs end with a slash. Root URLs can be specified directly with setRoot(...) or setRoots(...), or via printf templates (which expand to URLs) expressed as single Java strings with setRootTemplate(...) or setRootTemplates(...). The printf templates are expressed as single Java strings, which can be tricky to read, for example: "\"%s%s/%d/\", base_url, journal_id, year" (with the convention that the base URL value ends with a slash). If the specification specifies no root URL, all URLs in the AU are enumerated.

  • The specification can optionally have an include pattern or an exclude pattern. With an include pattern, URLs below each root URL that match the include pattern are considered (but others are ignored). With an exclude pattern, URLs below each root URL that match the exclude pattern are ignored (but others are considered). No error will happen if both are specified, but the exclude pattern will be ignored in favor of the include pattern. These patterns (regular expressions) can be specified directly with setIncludeSubTreePattern(...) or setExcludeSubTreePattern(...), or via printf templates (which expand to regular expressions) expressed as single Java strings with setIncludeSubTreePatternTemplate(...) or setExcludeSubTreePatternTemplate(...).

  • The specification can optionally have a general pattern, which is a regular expression applied to any URL still under consideration. It can be specified directly with setPattern(...), or via a printf template (which expands to a regular expression) expressed as a single Java string with setPatternTemplate(...).

  • The specification can have an optional media type (for example text/html), and only URLs under consideration that have this media type will be enumerated. There is currently no way to specify multiple media types, which can be problematic for media types with multiple common representations, like text/xml and application/xml. The class could be enhanced in the future to allow multiple media types.

The logic for which URLs are enumerated is found in the isArticleCu(...) method, which can be customized in a subclass.

By default, for each successful URL, this iterator makes one ArticleFiles instance that has its designated full text URL set, and no roles set. This behavior can be customized in visitArticleCu(...) and createArticleFiles(...) in a subclass.

7.2.3. SubTreeArticleIteratorBuilder

The org.lockss.plugin.SubTreeArticleIteratorBuilder class assists in the creation of a SubTreeArticleIterator instance under circumstances where the URLs of various aspects of an article (for example its abstract URL, its full text HTML URL, its full text PDF URL, etc.) can all be derived from one another through mutually compatible regular expressions and replacement strings.

An example of such mutual compatibility would be a journal where articles have full text HTML URLs that look like this: http://www.example.com/vol12/iss3/art45 and full text PDF URLs that look like this: http://www.example.com/pdf/article_12_3_45.pdf (assuming these URLs represent volume 12, issue 3, page 45). A regular expression for the full text HTML URLs (expressed as a Java string) could be "/vol(\\d+)/iss(\\d+)/art(\\d+)$", and the replacement string "/pdf/article_$1_$2_$3.pdf" would yield the corresponding full text PDF URL from a match; likewise a regular expressin for the full text PDF URL could be "/pdf/article_(\\d+)_(\\d+)_(\\d+)\\.pdf$", and the replacement string "/vol$1/iss$2/art$3" would yield the corresponding full text HTML URL from a match.

The SubTreeArticleIteratorBuilder class has convenience methods to:

  • Create a SubTreeArticleIterator.Spec specification. See setSpec(...), or use newSpec() to manipulate an empty Spec from scratch.

  • Define major aspects with one or more regular expressions matching the aspect's URLs, one or more replacement strings yielding the aspect's URLs from matchers for URLs of other major aspects, and one or more roles for the aspect.

  • Define minor aspects with one or more replacement strings yielding the aspect's URLs from matchers for URLs of major aspects, and one or more roles for the aspect.

The key differences between major and minor aspects are:

  • URLs enumerated by the SubTreeArticleIterator are tried against the regular expressions of the major aspects only.

  • The earliest URL for a major aspect to match for a given article is also designated as the article's full text CU.

All aspects are defined with the variants of the addAspect(...) method and associated methods.

When an ArticleFiles is complete, methods like setRoleFromOtherRoles(...) and setFullTextFromRoles(...) can be used to designate additional roles or the full text CU from the value associated with an ordered list of possibilities from other roles.

The getSubTreeArticleIterator() method can finally be used to obtain a SubTreeArticleIterator instance behaving in the specified manner.


Footnotes