That is, don't replace non-alphanumerics with a space. This matches the
behavior of Formatter::RemoveDiacriticsAndNonalphanumerics(), which
would be used here except that function would also remove quotes. We
actually discussed not introducing spaces previously, but I made a
mistake and didn't apply the same change to the user's search query:
https://github.com/standardebooks/web/pull/470#discussion_r1929591492
This change fixes queries with compound words like:
`Haycraft-Queen`
and also fixes queries for authors with apostrophes:
`O'Neill`
No changes to existing DBs are necessary because they already have terms
like `haycraftqueen` and `oneill` stored in `IndexableCollections` and
`IndexableAuthors`.
* Replace `EbookUrl` with `EbookId` in `Artworks`
* Add a `FullUrl` member to `Ebook`
Add documentation about when to use it versus Url.
The full URL is also being used as an ID in RSS feeds, so use `FullUrl` there.
* Store an `EbookId` in `Artworks`
The `Validate()` method is correctly setting it to null, but then the
`UPDATE` SQL statement is triggering another call to
`GetIndexableText()`. Without this change, empty strings are being
written to the `IndexableText` column.
This allows the user to run a keyword search and then change the sort
order. `Default` is interpreted as `Relevance` if a query is present,
`Newest` if not.
Having it in the SELECT fields was causing warnings like this:
```
NOTICE: PHP message: PHP Deprecated: Creation of dynamic property Ebook::$RelevanceScore is deprecated in /standardebooks.org/web/lib/Traits/Accessor.php
```
The data in these fields are separate:
* `IndexableText`
* `Title`
* `IndexableAuthors`
* `IndexableCollections`
There are also on indices on each of these fields so that they can have
separate weight in the relevance scoring.
Here's what's in `IndexableText` right now:
1. Title
2. Collections
3. Authors
4. Tags
5. LocSubjects
6. TocEntries
Here is the proposed new ranking:
```
10 * Title +
8 * Authors +
3 * Collections +
IndexableText
```
New indices for existing DBs:
```
ALTER TABLE `Ebooks` ADD COLUMN `IndexableAuthors` text NOT NULL;
ALTER TABLE `Ebooks` ADD COLUMN `IndexableCollections` text NULL;
ALTER TABLE `Ebooks` ADD FULLTEXT `indexSearchTitle` (`Title`);
ALTER TABLE `Ebooks` ADD FULLTEXT `idxSearchAuthors` (`IndexableAuthors`);
ALTER TABLE `Ebooks` ADD FULLTEXT `idxSearchCollections` (`IndexableCollections`);
```
Searching by `UrlName`, now for both `Artwork` and `Artist`, is clearer.
Take care to remove the apostrophes outright, don't replace them with a
space because otherwise the letter after the apostrophe becomes its own
term to match.
Follow up to the fix for #461
`Artwork::GetAllByFilter()` is already removing diacritics from search
terms, and `UrlName` stores the artist's name with diacritics removed by
calling `Formatter::MakeUrlSafe()`.
Keeping the search by `Name` because `Formatter::MakeUrlSafe()` makes
other changes to the name, too.
Fixes#461
