ArticleSearchIndex::_updateTextIndex() clears prior indexed content

[ctgraham]

Calls to ArticleSearchDAO::insertObject() in OJS 2.4.8 and OJS 3.0.2 clear existing object keywords if they exist.

The function ArticleSearchIndex::_updateTextIndex() calls ArticleSearchDAO::insertObject() in 2.4.8 and 3.0.2.

The function ArticleSearchIndex::_updateTextIndex() is called sequentially with the same arguments in subroutine ArticleSearchIndex::articleFilesChanged() because both ArticleSearchIndex::articleFileChanged() and ArticleSearchIndex::suppFileMetadataChanged() (via _updateTextIndex()) call insertObject() on the same file id.

The result is that the fulltext object keyword relationships for supplemental files are lost.

[ctgraham]

Similarly here: https://github.com/pkp/ojs/blob/ojs-stable-2_4_8/classes/submission/form/SuppFileForm.inc.php#L255-L276

My first thought was to add a new constant ARTICLE_SEARCH_SUPPLEMENTARY_FILE_METADATA to ArticleSearch.inc.php, but this would potentially confuse the mapping of ArticleSearch::getIndexFileMap() and dependencies of the use and display of two different search bitvalues mapped to a single search "field"?

An alternate approach is to provide a parameterized flag to ensure calls can avoid clearing existing keyword relationships when adding additional keywords to an object.

[ctgraham]

ArticleSearch::_indexObjectKeywords() assumes a empty index, so parameterizing a flag to to avoid clearing existing keywords starts to get messy quickly.

[asmecher]

Thanks, @ctgraham -- merged and ported to ojs-stable-2_4_8.

[ctgraham]

@asmecher, it looks like forward porting this to 3.x is unnecessary because 3.x doesn't index the metadata for files, only metadata for articles? Am I reading that right in ArticleSearchIndex.inc.php and SubmissionSearchIndex.inc.php?

[asmecher]

OJS 3.x doesn't include a separate "supplementary files" set, so those are no longer indexed per the code that was problematic here.