- Added highlighting support to Pagefind. Massive thanks to @Jothsa for pushing this across the line in #425! 🎉
- See 📘 Highlighting search terms for documentation on how to enable this new feature.
- Added a page size option to the Default UI. See 📘 UI Configuration > Page size.
- Added a
destroy()
function to the Pagefind JS API, allowing for a total re-initialization. See 📘 Re-initializing the search API. - Added a
destroy()
function to the Pagefind Default UI, allowing for a total re-initialization. See 📘 Re-initializing the Pagefind UI.
- Fixed a bug, resulting in a (very) large improvement to the NodeJS Indexing API performance (~100x).
- Fixed HTML entities being rendered escaped in metadata, filters, and custom page titles.
- Fixed a bug where
debouncedSearch
returnsnull
if any options object is passed to it. - Fixed a bug where a fully-qualified URL set via the NodeJS indexing API would be broken when returned as a search result.
- Fixed Pagefind's reporting of really fast indexing times (previously logged as slower than reality) — thanks to @danpls in #448.
- Fixed extracting sub-results when headings contain non-ascii text (especially RTL languages).
- Added Māori translations (PR #436 — thanks @Yoda-Soda !).
- Added Croatian translations (PR #440 — thanks @diomed !).
- Added Hungarian translations (PR #451 — thanks @adamlaki !).
- Added Bengali translations (PR #454 — thanks @marufmax !).
- Added Vietnamese translations (PR #467 — thanks @AREA44 !).
- Added Polish translations (PR #495 — thanks @KredensKuchenny !).
- Added Danish translations (PR #501 — thanks @jonassmedegaard !).
Hopefully the last hotfix for now — bugfixes only important for sites indexing Japanese or Chinese pages.
- Fixes a bug indexing ja/zh language pages containing weighting or element IDs
- Removes eager warnings being logged to the browser console if bunding the UI JavaScript files
- Adds a
.close()
function to the Pagefind NodeJS API, to help clean up the process when it is no longer required
- Fixes a bug when indexing some non-breaking spaces on ja/zh language pages in extended mode
Hotfix for Pagefind v1.0.0, restoring default-on support for multilingual word segmentation, and helping resolve packaging issues with new dependencies.
- Fixed the
pagefind
npm wrapper to use thepagefind_extended
release, as documented. - Changed
microjson
git dependency to apagefind_microjson
crate dependency (for now) — #421 .
Pagefind 1.0 is here! This release has been many months in the making, and we're thrilled to be bringing some great new features and improvements.
This release also marks a point in time for Pagefind's stability and maturity. Thanks to everyone who has helped out with contributions and feedback in the last year, we're now more confident than ever that Pagefind is a perfect fit with nearly any static website, staying performant and lean even as your site scales.
This 1.0 release includes one breaking change, and some notable non-breaking behavioral changes. A full writeup of these changes and their effects exists in the 📘 Migrating to Pagefind 1.0 guide on Pagefind's website.
- BREAKING:
- PREVIOUS: By default, Pagefind 0.x outputs files to a
_pagefind
in your site. - NEW: By default, Pagefind 1.x outputs files to a
pagefind
in your site. - More details on this change can be found in the [📘 migration guide].
- PREVIOUS: By default, Pagefind 0.x outputs files to a
- Changes to CLI options [📘 migration guide]:
- The
source
option has been renamed tosite
. - The
bundle-dir
option has been renamed tooutput-subdir
. - A new
output-path
option has been added.
- The
- Search indexing and ranking changes will cause result lists to differ from 0.x [📘 migration guide].
- The JS API initializes Pagefind at a different stage of execution [📘 migration guide].
Pagefind now supports configurable weighting for regions of content, which will be used when ranking results and generating excerpts.
Headings are automatically weighted higher, and custom weights can be inserted anywhere in your page.
See 📘 Weighting sections of the page higher or lower for documentation.
Pagefind now tracks headings and IDs when indexing your site. This can be used to show multiple results per page when searching your site, with direct links to the closest parent anchor.
See 📘 Showing multiple results per page for documentation.
The pagefind
package on npm can now be imported as a library into a NodeJS script, giving you programmatic control over indexing content from both disk and memory.
This feature is very open ended — be it integrating Pagefind into a static site generator, or indexing non-static and non-HTML content, we're excited to see what people come up with! Open a discussion on GitHub if you build anything unique that you would like to share!
See 📘 Indexing content using the NodeJS API for documentation.
Pagefind now takes inverse document frequency into account, meaning words that are unique across your site will rank higher than common words.
The ranking algorithm has also been improved across the board, which should result in better search relevance by default.
We're always looking to improve search relevance further, so open an issue on GitHub if you have any examples of searches that don't hit the mark.
Pagefind now better supports indexing various forms of compound words and code, meaning <MyComponent data-pagefind-body>
will now match searches for my, component, data, pagefind, and body.
- Pagefind now returns results for a prefix of the search word if there would otherwise be zero results:
- e.g. if searching for
bandwidth
would return zero results, you might get results forband
orban
.
- e.g. if searching for
- Pagefind can now search for emoji 🎉🚀✨.
- Do you filter more than you search? For those using the JS API directly, Pagefind now supports complex compound filtering:
- Pagefind is now smarter at generating excerpts, returning the most dense region of search hits on the page.
- Excerpt calculation also integrates with the new weighting feature, and will favor areas of the page with higher weighted words.
- The
pagefind
npm package no longer downloads binaries from a GitHub release, and instead has platform-specific dependencies that download only the relevant binary from npm.- This improves the installation speed of Pagefind through npm/npx, and also removes the need for any postinstall script making the entire process more reliable.
- The Default UI now supports being passed an HTMLElement directly, as an alternative to a selector (PR #331 — thanks @stefanprobst).
- The length of excerpts that Pagefind generates can now be customized:
- CLI: Fixed an issue where multiple
data-pagefind-body
tags on a page would conflict if one was nested deeper than the other. - CLI: Fixed builds for some Windows systems that were missing vcruntime.
- JS API: A new
pagefind.init()
function has been added, meaningpagefind.options()
can be called before loading assets, allowing you to change the path to load files from. - JS API: Performance searching very large sites for short terms should be improved.
- JS API: Passing an empty array for a filter value now behaves as if the filter was not supplied, instead of returning zero results.
- Default UI: Don't reset the browser-provided outlines when resetting UI styles.
- Default UI: Fixed an issue where titles containing HTML elements were not correctly escaped.
- Default UI: Improved the search input on mobile devices (PR #368 — thanks @valtlai !).
- Default UI: Fixed an issue where some UI strings would appear in English instead of the translated language.
- Added Indonesian translations (PR #273 — thanks @nixentric !).
- Added Serbian translations (PR #285 — thanks @DigitLib !).
- Added Italian translations (PR #323 — thanks @apjanco !).
- Added Hindi translations (PR #324 — thanks @Amitind !).
- Added Finnish translations (PR #366 — thanks @valtlai !).
- Added Turkish translations (PR #395 — thanks @taylanbildik !).
- Added Tamil translations (PR #402 — thanks @sanjaiyan-dev !).
Note: v0.12.0 will likely be the last feature release before an upcoming v1.0.0 that will contain a small handful of breaking changes. See the v1.0.0 milestone on GitHub for details and updates.
- CLI: Added a "Keep Index URL" setting. (PR #233 — thanks @kenpetti-toasttab !). See Pagefind CLI > Keep Index URL
- JS API: Added a
totalFilters
object to the search response, containing the total matches for the search term under each filter - JS API: Added an
unfilteredResultCount
key to the search response, containing the total matches for the search term if no filters were applied
- CLI: Stopped warning when encountering
data-pagefind-ignore="true"
instead ofdata-pagefind-ignore
- Search: Fixed merging filters from multiple indexes
- Default UI: Fixed filters sticking open once search input has been focused
- Default UI: Fixed the search input clearing when hitting the
Enter
key - Search / Default UI: Fixed HTML tags in Pagefind excerpts not being escaped. The
content
key remains unprocessed
- CLI: Improved
npx
wrapper compatibility on Windows, thanks @tylermercer! - JS API: Added a
debouncedSearch
function to the JS API. See Pagefind JS API > Debounced search - Default UI: Added a "Clear" button to the search input
- Default UI: Clear the search input on an
Esc
keypress - Default UI: Added UI translations for Swedish, thanks @mntzrr!
- Default UI: Added a
processTerm
hook that can normalize the search query. See Pagefind UI > Process term - Default UI: Added a
Clear
button to the search input - Default UI: Added functionality to clear the search input when
Esc
is pressed while the input is focused - Default UI: Published UI to npm under @pagefind/default-ui, as an alternative to using the files output by the Pagefind CLI
- Default UI: Fixed a syntax error in the CSS
- Work is underway on a new "Modular UI" that will live alongside the current "Default UI". Full support and documentation will be provided in a future release — the prerelease version can be found on npm under @pagefind/modular-ui
- As this package is still under development, some of the configuration may change in a future release. Make sure to pin your Pagefind versions for any production site relying on the Modular UI.
- Avoid using bsdtar in the release flow, as that will sometimes create sparse tar files that some packages cannot decompress. (Fixes lumeland/lume#362)
- Adds UI translations for Galician, Català & Spanish, thanks @pvillaverde!
- Fixes Pagefind failing on Safari due to an unsupported regex
- Fixed an issue where merging an index from a fully-qualified domain name would mangle the mapped URLs
- Fixed a corner case where a
data-pagefind-body
tag wouldn't be honored on pages with DOM nodes outside the mainhtml
element
- Fixed the Windows deployment target
- Updated only deployment targets:
- Pagefind now distributes an
aarch64-apple-darwin
build for M1 macOS machines- This will provide a speed benefit on these machines by skipping the Rosetta 2 emulation
- Pagefind now distributes an
aarch64-unknown-linux-musl
build for ARM Linux machines- Useful for ARM Docker images on M1 macOS
- Pagefind now distributes an
- Updated the npm/npx wrapper to reflect the newly available binaries
- Changed HTML parsing to a non-strict mode that will no longer error when encountering parsing ambiguities
- Updated the npm wrapper to respect an exit code returned from the main Pagefind binary
- Added the ability to exclude custom selectors via Pagefind config. See the exclude selectors documentation
- Fixed an issue where running a multi-site search through Pagefind UI wouldn't wait for all search indexes to be ready
- When the search term
null
is passed, Pagefind returns all results with filters applied.- In the case of a
null
search and an empty filters object, Pagefind would previously return zero results. - Pagefind will now return all results in this case.
- In the case of a
- Pagefind can now automatically read gzipped HTML files as its source
- Pagefind's automatic metadata now falls back to the
title
of a page if there is noh1
element - Fixed a couple of inconsistent url formatting issues on Windows
- Fix Windows release assets once more
- Removed
<header>
from the list of elements that Pagefind automatically ignores- If this element contains content you do not want to be indexed, you will now need to add
data-pagefind-ignore
- If this element contains content you do not want to be indexed, you will now need to add
- Added sorting functionality to Pagefind, see the Sorting documentation and the JS API Sorting usage
- Added the functionality to filter an index without searching, by passing
null
as the search query - Added support for custom Pagefind UI strings, see Using custom Pagefind UI strings
- Added a default debounce to the user input for Pagefind UI, and a corresponding
debounceTimeoutMs
option, see Debounce user input- Many thanks to @dprothero for the contribution! 💝
- Added a hook to process results before showing them in Pagefind UI, see the processResult documentation
- Fixed running Pagefind on Windows via the npx wrapper
- Pagefind now throws an error if a completely empty index is produced for whatever reason
- Fixed a bug where having exactly one known and one unknown language would drop the known language pages
- Fixed issue where
two<br/>words
would be indexed astwowords
rather than the correcttwo words
- Added
<style>
to the list of elements that Pagefind automatically ignores - Fixed the Pagefind UI
showEmptyFilters
option to work as expected - Fixed issue where adding a filter to a search with zero results would return all results for the filter
- Fixed uncommon bug around chunk boundaries
- For example: If your first search index chunk started with the word
hello
and you searched forh
, Pagefind would previously not load thehello
chunk and would instead return zero results.
- For example: If your first search index chunk started with the word
- Pagefind now gracefully skips pages that fail HTML parsing, and provides more context when these errors are hit.
- For those using the JS API directly, the
pagefind.options
function is now async. This will not break current usage, but using newer options may requireawait pagefind.options({ ... })
for them to be applied
- Added Multisite search support, allowing you to search multiple indexes from one Pagefind instance. See the new Multisite documentation for more information
- Added a preload function to the JS search API, allowing you to warm Pagefind up before a search, or while the user is typing. See the Preload documentation for more information
- Added a
timings
object to the JS search API response
- Passing a non-existent filter to the search function would previously be silently ignored. This will now return zero results
- Setting your baseURL to an external domain such as
https://example.com
would previously be prepended with a/
. This is now handled correctly and will link off-site - Pagefind would previously index search entities such as
"
without unescaping them. This is now fixed and these characters will be skipped - Searching for only punctuation would previously return all pages currently loaded into the Pagefind index. This will now return zero results
- Fixed a regression causing searches for hyphenated-phrases to return zero results
- Fixed Pagefind UI failing to match complex language codes such as
zh-hans-tw
to less complex translation files such aszh-tw
- Added French translations for Pagefind UI — thanks @nfriedli!
- Fixed standard & extended release archives attached to GitHub releases to be correctly assigned
- Multilingual support
- Pagefind now works out of the box for multilingual sites
- Pagefind UI is now translated into
af
,de
,en
,ja
,no
,pt
,ru
, &zh
- See the multilingual documentation for more information on this feature release
- This release adds a
pagefind_extended
binary release, which is larger than thepagefind
release but includes support for indexing Chinese and Japanese languagespagefind_extended
is now the default when runningnpx pagefind
. The smallerpagefind
release is still available via the GitHub Release attachments
- Improved Pagefind logging and added a
--verbose
flag with extra information - Added warnings when Pagefind encounters pages without outer
<html>
elements - Added a console warning when Pagefind detects that a cached
pagefind.js
file was loaded alongside a search index from a newer release
- Oops — fixes the npx wrapper on Windows
- Added prebuilt Windows binaries
- This adds Windows support for running Pagefind via
npx pagefind
- Windows binaries can also be downloaded via the GitHub Releases
- This adds Windows support for running Pagefind via
- Tēnā koutou katoa — Fixed an issue where Pagefind could not search for words containing some special characters.
- Support Apple Silicon Macs (via Rosetta 2)
- Fixed a form submission error in Pagefind UI causing trouble for some content security policies
- Fixed a visual quirk in the Pagefind UI filters on Safari
- The glob that Pagefind uses for finding files to index can now be configured. See Config > Glob
- Added a
data-pagefind-ignore="all"
option that does not process filters or metadata within the target element. See Indexing > Removing individual elements from the index - Added a
data-pagefind-default-meta
attribute that can provide fallback values for metadata that could not be found on the page. See Metadata > Defining default metadata - UI: Pagefind UI will now check for a
url
key in a page's metadata that should be used over the generated URL. See UI > Overriding the URL of a result - UI: Added a configuration option for hiding images from the Pagefind UI. See UI > Show images
- UI: Added a configuration option to hide filter values that have no available results given the search query and existing filters. See UI > Show empty filters
- UI: The Pagefind UI filter panel will now default to expanded if there are sufficiently few filters
- Server gzip support
- Pagefind implements its own gzip handling, but would fail in the rare case that a server detected the gzipped files and served them such that a browser would decompress them. Pagefind will now identify that these files have already been decompressed rather than error.
- UI: HTML entity improvements
- Fixed the remaining elements in the Pagefind UI that did not correctly render HTML entities.
- UI: Filter state improvements
- Previously, deleting a search term would reset the selected filter values and the open filter groups. This state is now preserved when the search input is empty.
- UI: The Pagefind UI JS and CSS files are now correctly minified
- Hash fragment contents using the entire file, to prevent stale content
- Use the image_alt correctly in Pagefind UI
- An automatic
image_alt
metadata value will be included when returning an automaticimage
metadata value. See Metadata > Default metadata - Multiple filter and metadata keys set can be set per element. See Metadata > Defining multiple keys
- A root selector can now be configured to further restrict Pagefind indexing. See Config > Root selector
- If re-running Pagefind over an output directory, existing hashed files will be reused if present, which will improve hot build speeds for large sites.
- Added latest version to the header of the documentation.
- Added
<template>
elements to the ignored text index list - Multiple exact matches on a page will no longer be returned as separate results
Tidied up the search API output to remove some not-yet-implemented fields.
Changed images in Pagefind UI to contain rather than cover. Relevant CSS variables have changed slightly.
Added a --serve
option to the Pagefind CLI that will host the site on a local development server after building the search index.
The first stable 0.x release of Pagefind.