v2.0.0-beta.2

This is the second beta release of the Stork 2.0.0 release branch.

The beta signifier means that the release is documented and usable, though the API is not final and not all features have been implemented yet. Until 2.0.0 is released, the source code is stored in the v2 branch of the repository.

The documentation is available here.

---

This beta release includes the following changes compared to beta 1:

• Restores HTML parsing behavior with included and excluded selectors
• Updates the scoring algorithm for accuracy
• Includes search results for stemmed alternatives for specific words
• Reports build errors while documents are read, rather than batched at the end
• Fixes a bug where certain theme styles were not being properly set
• Fixes some bugs related to correctly reporting the loading state of the WASM blob and the search index within Javascript

v2.0.0-beta.1

This is the first beta release of the Stork 2.0.0 release branch. The beta signifier means that the release is documented and usable, though the API is not final and not all features have been implemented yet.

Until 2.0.0 is released, the source code is stored in the v2 branch of the repository.

The documentation is available here.

Note
In this beta, you'll have to explicitly register the WASM blob with a custom URL:

stork.initialize("https://files.stork-search.net/releases/latest/stork.wasm")

---

2.0.0 greatly simplifies the search algorithm. As a result, the indexing binary, WASM blob, and Javascript file are all smaller than the 1.x branch. Searches are faster. The configuration API is more powerful.

The following features are still not working:

• Sharded indexes. The file format supports them, the library supports merging shards together, but the builder doesn't yet know how to split an index into multiple parts.
• Stemming
• HTML parsing, including the included and excluded selectors
• Matching search results in titles
• Result and excerpt scoring, leading to strange ordering of results

Currently, only search indexes made with the 2.0.0 indexer binary can be used with the 2.0.0 web code. This lets me greatly simplify the support matrix, so I'm inclined to leave it that way; please let me know if this would be troublesome for your workflow.

Please give me feedback if you have any!

---

The following artifacts have been uploaded to the Stork CDN:

• https://files.stork-search.net/releases/v2.0.0-beta.1/stork.js
• https://files.stork-search.net/releases/v2.0.0-beta.1/stork.wasm
• https://files.stork-search.net/releases/v2.0.0-beta.1/stork.css
• https://files.stork-search.net/releases/v2.0.0-beta.1/stork-ubuntu-20-04
• https://files.stork-search.net/releases/v2.0.0-beta.1/stork-ubuntu-22-04
• https://files.stork-search.net/releases/v2.0.0-beta.1/stork-macos-11
• https://files.stork-search.net/releases/v2.0.0-beta.1/stork-macos-12
• https://files.stork-search.net/releases/v2.0.0-beta.1/federalist.st
• https://files.stork-search.net/releases/v2.0.0-beta.1/3b1b.st

v2.0.0-alpha.3

Despite the version number, this is the first Stork 2.0.0 alpha candidate, released from the v2 branch of the repository.

2.0.0 greatly simplifies the search algorithm. As a result, the indexing binary, WASM blob, and Javascript file are all smaller than the 1.x branch. Searches are faster. The new index format includes support for sharded indexes, meaning future indexes will be able to be made up of multiple size-limited files that get loaded dynamically as your user types. Other features, like metadata filters, will also be available in the 2.x branch.

Many things are not working currently. The config API is different, but there doesn't yet exist documentation for the new format. Stemming, title matching, and highlighting are broken, and displayed excerpts are not trimmed to word boundaries. Some settings aren't wired up properly. Many parts of the release process are not working.

Currently, only search indexes made with the 2.0.0 indexer binary can be used with the 2.0.0 web code. This lets me greatly simplify the support matrix, so I'm inclined to leave it that way; please let me know if this would be troublesome for your workflow.

Please feel free to try out the new artifacts (attached to the Github release) and let me know what you think!

v1.6.0

New Features

• Timestamp formatting for SRT files is now functional, and the documentation, which previously referenced unimplemented features, is now accurate. (Thanks for the bug report, @selamhabibi!)
• The x SVG is now horizontally centered within the button. (Thanks @kkwteh!)

Other

• Internal dependency updates (Thanks @dependabot!)

Project Update

I'm still hard at work on 2.0.0! The new version's key feature will be sharded search indexes, but implementing this has meant that I rethink the data format of the serialized index to support loading single chunks at a time while keeping the search interface snappy. Maintaining Stork's speed and resource size while redesigning the data format has proven difficult.

In personal news, I got engaged to be married over the holidays, which while exciting, has meant I have less time to work on Stork than I hoped. I'm still around and looking at the repo every day (despite the slowed progress of releases) and hope to get a 2.0.0 RC into your hands before the end of March.

The RC will be published to Github for all to try, but please let me know if you are particularly interested in poking at it! I'm happy to help walk anyone through the changes.

Best,
James

v1.5.0

New Features

• Adds the Javascript registration option transformResultUrl, which, when set, transforms search result URLs from what's described in the search index to what gets output in the DOM. (Thanks, @ArsenArsen!)
• Setting output.excerpts_per_result to 0 in your configuration file will now enable index serialization optimizations that can reduce your index filesize by 20-50%. Note that this will make all excerpts disappear from the search UI.

Bug Fixes

• Fixes a bug where multiple instances of the element described in the exclude_html_selector configuration option were not being correctly excluded. (Thanks, @ezekg!)
• Fixes a crash when trying to parse noncompliant Markdown contents (Thanks, @userjy!)
• Fixes a bug when highlighting search results where highlights wouldn't surround an entire word. (Thanks, @kdheepak!)

Other

• Internal dependency updates

v1.4.2

Bug Fixes

• Fixes a regression where the test server (stork test) was inaccessible
• Fixes a regression where error messages weren't displayed if no files could successfully be indexed
• Themes now compensate for border radius when displaying progress bar (Thanks @jmooring!)

v1.4.1

Bug Fixes

• Fixes a regression that prevented the Stork CLI from reading index configuration from stdin
• Fixes an architectural decision that prevented Stork from being published to crates.io

v1.4.0

Important Changes

• Starting with 1.4.0, the JS, CSS, and WASM files at the root of the files.stork-search.net CDN will no longer be updated. From this release going forward, you will need to change your <script> tags and <link rel="stylesheet"> tags to upgrade to a new version of Stork. See this Github announcement for more details.

New Features

• Configuration files can now be in JSON format, in addition to TOML format
• The web library now takes the number of occurrences in each file into account when ordering search results.
• CLI: The --json flag for the search subcommand is now deprecated, as it did nothing in previous versions of Stork.
• CLI: A new command line option for the search subcommand, --format <VALUE>, now determines how search results will be displayed in the terminal. The default value is json, which will display the search results in the JSON format understood by the browser module. A new value, pretty, will format the search results in a readable, understandable way. In the 2.0.0 Stork release, pretty will be the default value for this flag.
• Two new CSS themes, edible and edible-dark, have been added. See https://stork-search.net/themes for examples and instructions on how to integrate them into your project.

Bug Fixes

• Removes a stray console.log from the Javascript application
• Fixes a Javascript runtime bug where registered indexes weren't always reporting as ready
• Fixes a crash when the indexed contents contained words longer than 128 characters
• The --timings flag previously did nothing. Now, when included with the build or search subcommands, timing information will be displayed at the end of the terminal output via stderr.
• Document titles that were comprised of several words separated by hyphens would crash the search interface. This has been fixed.

v1.3.0

New Features

• Indexes alt and title attributes on HTML elements
• Adds configuration keys to set an HTML selector as excluded from indexing
• Adds a configuration key, output.save_nearest_html_id, that, when set to true, will index the nearest HTML IDs for each word. The web interface will link to that ID; clicking on that search result will jump to the text's location on the page.
• Reduces JS and WASM artifact sizes by changing build system settings
• Updates CSS themes, and adds a new theme: flat

Bug Fixes

• Fixes a bug where users were able to accidentally download two instances of Stork's WASM on the page (Thanks @justinmayer!)
• Fixes a bug where the indexer was hanging in environments where stdin was not passed in as an empty stream (Thanks @Aethon!)

Other

• Updates dependencies

v1.2.1

Bug fixes:

• Fixes issue where [x] button in Basic or Dark themes wouldn't respond to the resizing of the stork-wrapper container (#176)
• Fixes issue where searching for three characters wouldn't display results on the web page (#172)
• Fixes crash when the title of a document included non-unicode characters (#173 - thanks @Erwan-le-Gall!)