[meta] Inventory search.module tests. File issues to address test holes.

• All of the node/user search tests implicitly test this.
• SearchKeywordsConditionsTest verifies that a plugin can override the isSearchExecutable() method to allow searching without keywords set, and that GET query parameters are made available to plugins during search execution.
• SearchPageOverrideTest verifies that a plugin can override the buildResults() method to control what the search results page looks like.

• SearchKeywordsConditions test class needs better one-line and getInfo() documentation of what it actually does.
• SearchPageOverrideTest does too -- it says it tests hook_search_page() which doesn't even exist any more.

These are part of #2183951: Fix up some search tests.

• SearchConfigSettingsFormTests::searchModuleSettingsPage tests that plugin-specific config can be edited in the UI, although it does not verify that this config can be used later for anything.
• SearchConfigSettingsFormTests::testSearchModuleDisabling() tests that a default can be set by visiting the correct URL, and that it works from the search page and the search block, and that if this is the only enabled search type, the other tabs are not visible. Also verifies that with other tabs enabled, all are shown after directly going to the correct URL search/node. However, it doesn't disable via the UI but via $entity->disable() method in code.
• SearchConfigSettingsFormTests::testDefaultSearchPageOrdering() tests that out of the box the order of search tabs at URL /search is correct.
• SearchConfigSettingsFormTests::testMultipleSearchPages() tests that you can add new search pages with the UI, and that they are both displayed at /search, and that the order can be set via the UI (in that the order "sticks"). It also tests that the operations links are present and that after enable, disable, and delete, the search config screen has the right pages and operations visible, but it doesn't verify this is reflected at /search.

• Core Node plugin uses this system, so Node searching tests implicitly test this.
• SearchMatchTest - tests adding information by calling search_index() with the type set to various strings, including some in Japanese, and then querying it via the SearchQuery extender. Tests cases with single keywords, multiple AND keywords, OR keywords, negative (exclude) keywords, phrases, and complex combinations. Verifies that the different search "types" are separated in the index too.

• SearchCommentTest: Relies on extensive use of the search_mark_for_reindex function, and would fail if this function didn't mark a node for reindexing.
• SearchNodeUpdateAndDeletionTest - verifies that after $node->save() API call and $node_search_plugin->updateIndex(), a modified node has been reindexed, and that after $node->delete() the node is immediately removed from the index and is unsearchable.

• SearchMatchTest kind of looks like it is testing keyword relevance ordering, but it doesn't seem to include cases where the keyword relevance is different between the results.
• SearchRankingTest does test keyword relevance ordering in a valid way. It also tests that HTML tags influence keyword relevance rankings in the right way.

• Lower-case everything
• Invoke hook_search_preprocess() with the right language passed in
• If the "index.overlap_cjk" setting is TRUE, make "words" by splitting CJK text into overlapping sequences of characters of length equal to the "index.minimum_word_size" config setting.
• Removes all punctuation inside numbers.
• Treats sequences of multiple . and/or - as word boundaries.
• Removes ., _, and -.
• Treats remaining punctuation as word boundaries.
• Splits into words
• Truncates all words to 50 characters.

Phew!

• The basic CJK overlap functionality is tested in SearchMatchTest, which searches for several subsets of some Japanese text after indexing it.
• Tokenizing all known ranges of CJK characters is tested in SearchTokenizerTest, which also verifies that English letter characters are not tokenized this way. Calls search_simplify() explicitly with the config set to tokenize at one character.
• SearchNodeAccessTest::testPhraseSearchPunctuation tests indexing and searching for a word with an apostrophe. Why it's on this test class, I have no idea. Came from D7 that way.
• SearchExcerptTest ends up testing some of the preprocessing stuff implicitly: punctuation in numbers, hyphen removal, punctuation as word boundaries, hook_search_preprocess()
• SearchNumberMatchingTest implicitly tests the number preprocessing by doing node index and search from the UI for number/punctuation strings
• SearchNumbersTest has some more test cases for numbers, including one over 50 characters long (again, makes nodes and tests in the UI
• SearchPreprocessLangcodeTest - verifies that language is considered in preprocessing
• SearchSimplifyTest - verifies that punctuation rules are followed in search_simplify(). Also verifies that non-word characters are removed, and that for word characters, the output of search_simplify() is at least as long as the input (nothing is removed) [Note that word and non-word are as defined in Unicode and every Unicode character is tested].

• SearchMultilingualEntityTest uses API to set the throttle (not the UI), and assumes that the throttle should count each translation as a separate item, and verifies that the indexing is stopped after the correct number of translations is reached.

• Only "positive" keywords should be highlighted (if you are excluding words from an advanced search, those shouldn't be highlighted even if they occur)
• In the case that the match occurs because a preprocessed keyword matches the preprocessed text, highlighting should also work.
• The excerpt length should be at least 256 characters and should have about 60 characters around keywords (this may be difficult to test because of the "about" part)
• If the text is broken up, there should be ... between chunks
• If the text is broken up, the order should match the input text order, not the keyword order.
• All matched keywords should be highlighted
• HTML tags in the input text should be removed

• SearchNodeAccessTest looks like it should be testing node access, but all it actually does is have a module turned on that includes node access, without actually exercising any of the node access stuff. ?!?

• SearchAdvancedSearchFormTest - verifies that node title can be searched for via submitting with the title in the keyword portion of the URL (matching and non-matching cases are tested).
• SearchExactTest - verifies that keyword search and phrase search work as expected -- that searching for a phrase finds only the exact phrase, not any node with the keywords. It also tests that the right number of pages of results are found for a multi-page result.
• SearchPageTextTest - verifies some error/warning messages

• SearchAdvancedSearchFormTest - verifies that the node type filter works (find node if it matches, don't find node if it doesn't). Also verifies that you can put the title in the "Or" keywords field and the node will be found
• SearchLanguagetest - verifies that language selections in Advanced form are added to the search URL GET params, and that configuring or un-configuring languages makes the choices available or not available. Doesn't actually test search filtering with languages though. Hence #2161067: Advanced search with Language filter does not work
• SearchMultilingualEntityTest - calling methods on the NodeSearch plugin directly, verifies that search results can be filtered by language.

• SearchMultilingualEntityTest - verifies that in a generic Node search you can get results in any language, and you will get a separate result for each language item that matches. This is calling the NodeSearch::setSearch() and execute() methods directly, not the UI though. When these nodes are rendered as search results you will find that they are rendered in English and not the language whose keywords matched. The output is described on #2161067: Advanced search with Language filter does not work.

• SearchCommentCountToggleTest: Verifies that with various comment status settings on a node, the search results display the comment count appropriately.
• SearchCommentTest: Verifies that comment subject and body can be searched, and that HTML tags are not rendered in the search results. Also verifies that if comments are hidden, comment body cannot be searched. Also verifies comment-related permissions with searching fairly extensively (anonymous user needs "access comments" or comments are not indexed, and searching user needs "access comments" or comment information is not shown on the search results page). Also verifies that the "Add new comment" text does not get into the search index or appear on the search page, but only for the case where anonymous users cannot comment.

• SearchPageTextTest - verifies some error/warning messages. Note however that the dreaded "bike shed" text is not even appropriate for User Search but it is still displayed.
• UserSearchTest (in User module lib) tests searching by user name works

• SearchEmbedFormTest verifies that you can display a form embedded in search results (via hook_preprocess_search_result() in the theme system) and the form will work (such as an Add to Cart form for Commerce).
• SearchSetLocaleTest verifies that a locale with commas as decimal places can be set and search will still work.