Stats for numbers-type people
All file sizes are pre-compression. All files that come from files.stork-search.net are gzipped; you should compress your search index before serving it.
|Version||federalist.st size||stork.wasm size||stork.js size||Search duration*|
|1.1.0||1.13 MB||487.04 kB||106.87 kB||2.5441 ms.|
|1.0.4||1.13 MB||484.57 kB||92.24 kB||2.4904 ms.|
|1.0.3||1.11 MB||591.97 kB||94.75 kB||— **|
|1.0.2||1.11 MB||591.97 kB||94.75 kB||— **|
|1.0.1||1.11 MB||591.97 kB||94.75 kB||— **|
*Why this benchmark? Reproduceability, mostly. This test benchmarks the speed of the search algorithm by itself, rather than in the WASM runtime, removing any non-Stork-algorithm variance that might occur. On my computer, according to in-browser performance tests, the browser repaints with the search results roughly 10ms after the keypress event.
**Benchmarks from 1.0.0 and before were run on my personal computer. Benchmarks from 1.0.4 and afterwards were run on an AWS EC2 server t4.micro instance by running
cargo bench; the listed benchmark is for the
search::federalist::liberty benchmark. Benchmarks in between 1.0.0 and 1.0.4 were found to have been invalid.
- Added self-hosting support. Read the self-hosting documentation to learn more.
onResultsHidden- Callback that gets called when the results are hidden, when the user presses esc or clicks on the close button
onInputCleared- Callback that gets called when the input is cleared, when the user presses esc twice
showCloseButton- Boolean to determine whether the close button is visible or not
- Stork can now take in a configuration file that's piped into the
$ stork --buildcommand, instead of requiring that you pass in a file path.
html_selectoroption in the configuration file wasn't being parsed correctly, leading to the feature seemingly not working
- Better error message when there are no valid files
- Hyphens are now treated the same as spaces for indexing and searching purposes. In effect, you can now search for
avonand it will match the term
Stratford-upon-Avonin your index.
- Stork used to fail the entire indexing process if there was an error parsing a single file. Now, it will collect those errors and present them in the console, but still build an entire index with the remaining files.
- Fix highlight offsets when excerpt contains multi-byte characters. Requires that you re-build your index. (Thanks for reporting, @DanilaFe!)
- Describe which HTML selector cannot be found when an HTML document fails to index.
- Collect and display indexing errors instead of failing early (Thanks for suggesting, @fauno!)
- Some browsers wouldn't display results properly if you started typing before the WASM file had loaded (thanks, @reese!)
- The indexer wouldn't index words past a certain point if the document contained non-word space-delimited tokens. For example, if your document had the contents
hello - world, the word
worldwouldn't be indexed. (Thanks for reporting, @DanilaFe!)
- Fix a webpack bug that was only encountered on first install
Thanks for using Stork!
Stork 1.0.0 is here, with new features, stability improvements, and lots more speed. The major version bump signifies that Stork is officially out of beta, and that I (James, the developer) believe that it can be "production ready" on your site.
Index Generator Updates:
- Stork can parse HTML and Markdown input files
stork --testcommand will open a local webserver that lets you test a generated index, without having to build it
- Sped up index generation (Thanks @DenialAdams!)
- Separated index parsing from searching in generated WASM library, which sped up searches (Thanks @DenialAdams!)
- Reworked public interface for those using Stork as a Rust library
- Added dark.css theme
- New website documentation
Fixes a bug (#69) in which content wasn't being indexed after an all-punctuation word. Make sure to rebuild your index to see the effects of this fix.
Automatic CSS Improvements:
- Site styles were overriding Stork's Basic theme. The stylesheet has been updated to be more specific.
- Ubuntu build artifact uploaded to the Github release for easy
wget-ing during your static site's build process
cargo install, the downloaded binary was named
stork. This has been fixed.
Automatic JS improvements
- Start using Typescript (this accounts for the ballooning in JS filesize)
- You can now use the keyboard to interact with the search results! Up and down arrows to navigate, enter to select a result, and esc to hide the list of results
Rebuild your index to get:
- SRT Subtitle support
- Customizable stemming language
- Content in titles is now indexed and prioritized in search results
- Better error messages if search crashes, which it shouldn't.
Automatic JS improvements
Update the binary for:
- Improved UX if you provide deprecated configuration fields
- Improved console output if you run
- New site infrastructure
- WASM: Deprioritizes stop words in search results
- WASM: Searches for words with the same stem as the query
- WASM: Improves error messaging for corrupt indexes
- Indexer: Adds knowledge of word stems to index
- Rewrite search algorithm to support multi-word search queries, including sensible result ordering and query highlighting.
basic.cssto give results more room
Accidentally ran a bad version bump. There are no code changes from 0.5.1 to 0.5.2.