10 What's new since version 1.0?

10.1 Changes in version 1.4.9

Bug fix:

• Documentation was updated.

10.2 Changes in version 1.4.8

Bug fix:

• A bug which caused a VERSION file to be created containing a git error message if the user chose to customize the test suite instead of creating their own configuration has been fixed. A random version identifier is created instead.

10.3 Changes in version 1.4.7

Bug fixes:

• A bug which caused a tokenization error because the XSLT processor was using an old Unicode dataset has been handled with a temporary workaround, adding the codepoint U+A78F to the tokenization mechanism.
• When searching for a phrase containing an ampersand, the phrase would be successfully found, but if the same search was reloaded from the browser URL, it would fail. This has now been corrected. Thanks to the TMP team for reporting this.

10.4 Changes in version 1.4.6

Bug fixes:

• A bug which caused weighting values to be lost during tokenizing has been corrected.
• The mechanism which scrolls the search page automatically to display search results has been adjusted in an attempt to avoid hiding the first result(s) behind a sticky or fixed page header.
• The build process has been modified in an attempt to ensure that the search page which is generated contains one and only one charset declaration.
• A warning is now issued when no title is found for a document being indexed. This situation means that any search results will be less useful to the end user because no document title can be displayed for the document.

10.5 Changes in version 1.4.5

Bug fixes:

• Two configuration parameters, <stopwordsFile> and <dictionaryFile>, were previously documented as optional, but in fact the build process requires them. The documentation has been updated.
• There was a mismatch between files which were included in the indexing process and files which were checked for well-formedness ahead of indexing; it was possible for files that were not intended to be indexed at all to be checked and fail the well-formedness test even though they were not part of the build. This has been fixed.

10.6 Changes in version 1.4.4

Bug fix:

• A deprecated JavaScript function, substr(), which was used in one line of code only, has been replaced by substring().

10.7 Changes in version 1.4.3

New feature:

• A new method has been added to the StaticSearch class: preProcessSearchString(), which is called from parseSearchQuery(), does nothing as it is defined, but can be overridden by end-users to enable them to do any special processing they might need to do on an input search string. This is sometimes necessary when (for example) specific diacritics are being ignored by the search indexing, but end-users might type them in anyway. The search string is also being unicode-normalized (to NFC) at the beginning of the parsing function.

10.8 Changes in version 1.4.2

Bug fix:

• A bug which could cause a divide-by-zero error in the JSON generation stage of the build when the target site consisted of fewer than ten documents was fixed via a pull request from Norman Walsh. Thanks Norman!

10.9 Changes in version 1.4.1

Bug fix:

• A minor bug whereby a rare combination of circumstances could lead to document hit scores being reported as concatenated numbers rather than summed numbers has been fixed.

10.10 Changes in version 1.4

Deprecations requiring changes to existing projects:

• Filter sort keys must be declared using the all lower-cased data attribute data-ssfiltersortkey. While the documentation in 7.1.1.1 Sort order for description filters correctly specified the attribute's name, the processing code only accepted the camel-case version, which is invalid XHTML5. In 1.4, using data-ssFilterSortKey will result in WARNINGs; in all subsequent versions, using the camel-case attribute name will result in build failures.

New features and enhancements:

• A new Search only in feature has been added. This enables you to specify regions of documents and label them using the label attribute on <context> (per issue #20). Users can then check only the regions they would like to get search results from.
• A new feature filter has been added to the collection of search facets. This provides an option for cases where the number of items in a description filter might be so large that providing a long list of individual checkboxes for all the options is not practical. Instead, the feature filter offers a method of finding and selecting items using a typeahead control.
• Search page controls are now WCAG2-compliant.
• When you navigate back or forward to a previous search, if any of the filters which were used in that search are hidden inside closed HTML details elements, those elements will be opened.
• A noscript element is now inserted into the search page to handle cases where JavaScript is turned off in the user's browser.
• A ‘Loading...’ splash screen is shown when the search page is initially configuring itself.
• All inline CSS and JavaScript has now been moved to external files, to better suit Content Security Policy constraints.
• The staticSearch report has been simplified and no longer produces a concordance of stems by default. The concordance can be built at the command line by calling the concordance target in ant:

  ant concordance -DssConfig=cfg.xml⚓

• The version attribute has been added to the root <config> element to better future-proof the alignment of configuration files and the staticSearch codebase. See 7.5.1 The config element for more details.

Bug fixes:

• A bug which caused number filters to be ignored when navigating back to a previous search has been fixed.
• A bug which caused the tokenizer to assume wordbreaks when encountering certain diacritics has been fixed.

All issues and tickets related to version 1.4 can be found on GitHub.

10.11 Changes in version 1.3

Note that version 1.2 was withdrawn in favour of version 1.3, so the list below includes changes from the original version 1.2 and the current 1.3.

Deprecations requiring changes to existing projects:

• All staticSearch classes with periods have been deprecated in favour of underscores as periods in class names conflict with the standard "." chaining selector in CSS and JavaScript. (See issue 149 for the full discussion.) This affects the majority of staticSearch meta classes, which should be changed from staticSearch. to staticSearch_; see the full list below:

  DEPRECATED VALUE	REPLACE WITH
  staticSearch.desc	staticSearch_desc
  staticSearch.bool	staticSearch_bool
  staticSearch.num	staticSearch_num
  staticSearch.date	staticSearch_date
  staticSearch.docTitle	staticSearch_docTitle
  staticSearch.docImage	staticSearch_docImage
  staticSearch.docSortKey	staticSearch_docSortKey

  For version 1.3, using the deprecated period syntax will result in WARNINGs; in all subsequent versions, using the period syntax will result in build failures.
• The original parameters config and configFile have been renamed ssConfig and ssConfigFile to minimize the chances of naming collisions with parameters in other build processes. IF YOU HAVE SCRIPTED A staticSearch BUILD AS PART OF YOUR OWN BUILD PROCESS, YOU WILL NEED TO UPDATE THESE PARAMETER NAMES.

New features and enhancements:

• The JavaScript source code has now been split into several distinct source files, and is compiled and optimized using the Google Closure Compiler at build time. See JavaScript compilation for more information.
• Support for French in captions etc. in the search page has been improved.
• A French stemmer is now available, as well as a caption set for French search pages.
• Images can now be configured for specific parts of a document, as well as for the whole document, for display alongside KWICs in results. This is part of a new extension mechanism using custom attributes.
• The CSS for the search page inserted by the indexing process is now more easily accessible in a separate file css/ssSearch.css, which is linked into the search page at build time.
• Links to target documents from keyword-in-context results now include a search string parameter that specifies the hit text, so that JavaScript running in the target page can highlight the search hit(s) and scroll to them. See Highlighting search hits on target pages for more information.
• Documentation has been significantly improved with additional explanatory remarks for many elements, and the staticSearch build of the documentation now includes hit highlighting (the feature described above).
• Only ancestor ids are indexed when <linkToFragmentId> is enabled; formerly, any preceding id value was used.
• Results can now optionally be viewed in batches by setting the new <resultsPerPage> configuration option.
• The maximum number of results that a search can return has been set to 2000 results by default and can be changed using the new <resultsLimit> element. If a search returns a set of results that exceeds this limit, staticSearch does not render the results and advises the user to try a more precise search.
• The minimum length of a word to be indexed is now configurable, so in unusual circumstances you can now enable searching for 1- or 2-letter words using the <minWordLength> parameter.
• The staticSearch report is now discussed in the documentation (see 7.9 Generated report) and the "Not in Dictionary" and "Foreign Words" reports have been improved.
• The filter creation process has been rationalized such that all filter processing happens in json.xsl, which has also improved the build performance slightly.

Bug fixes:

• The encoding structure for docImage, docSortKey, and docTitle has been constrained such that each doc* <meta> must include both a name and class value:
  <meta name="docTitle"
   class="staticSearch_docTitle" content="My custom document title"/>
• Temporary XML files from dictionaries are now removed during the clean step of the build process.
• All HTML characters are properly escaped in context snippets.

10.12 Changes in version 1.1

New features and enhancements:

• Search results can now be sorted using a user-supplied sort key. This is useful when searching only with filters (so all documents have the same relevance score) or where many results have the same relevance score.
• Using the new <linkToFragmentId> parameter, keyword-in-context results can now have individual links to nearest ancestor fragment id, so the searcher can go directly to the relevant section of a document.
• The order of parameter elements in the configuration file is no longer fixed. The schema now allows elements to appear in any order.
• We have added a new How Do I... section in the documentation.
• Phrasal searches are now case-sensitive, meaning that you can use a ‘phrasal search’ to search for proper names, by putting quotation marks around them and making the first letter upper-case.
• Indexing is faster and the size of the index is smaller because we have eliminated the upper-case index.

Bug fixes:

• URL containing empty search query string caused unwanted page scroll.
• The build process would fail to open the search report when running under very recent versions of Java.
• Punctuation is no longer stripped from phrasal searches, so searching for longer phrases including punctuation should be more successful.