* Update parseInterval to handle "0" correctly
When a parameter like "0ms" is passed in to parseInterval it gets parsed to 0.
Previously this would result in a return value of "undefined" because 0 is falsy
and thus the `return 0 || undefined` statements return undefined.
The purpose of the form `parseFloat(str) || undefined` was to return "undefined" if
parseFloat failed (parseFloat returns NaN, a falsy value, if it can't parse its
argument). Unfortunately, as mentioned, parseFloat can also succeed and return a
falsy value -- when the argument is "0" (or "0.0", etc.). So the new code, rather
than depending on the falsiness of the result of parseFloat, explicitly checks for
a NaN.
* Adds some semicolons
Adds some semicolons to parseInterval (and tests) for consistency.
* Add one more parseInterval test for "0"
Adds test test to make sure parseInterval works on "0".
* Adds functional tests for every, swap, settle, throttle, and delay
* Explcitly check that setTimeout values are > 0
These values come from user settings that are read from parseInterval,
so they could be a number or undefined.
If the value being checked is > 0 setTimeout will be called with some
associated function. If the value is 0 or 'undefined' the associated function
will be called immediately ('undefined' is not greater than 0).
* Change '!== undefined' to '> 0'
`pollInterval !== undefined` is a subtly different conditional than just `pollInterval` or `pollInterval > 0` (which are equivalent). Changes the conditional to `pollInterval > 0` so as to not change the behavior but also be more explicit in the test.
* apply updated patch with legacy WS/SSE removal
* remove legace WS/SSE from the test server also
* remove sse mention from getTriggerSpec()
* clean up sse.js implementation after removal of hx-sse
* Process sse-swap properly when swapped in
Previously `sse-swap` were only processed when `sse-connect` was on the
root of what was being swapped in.
This is now fixed so that the closest sseEventSource is found and used as the
event source
* Pass for code readabiltiy
remove the nested if statements as there is no difference in handling
between sseURL and LegacySSEUrl.
* Fix indenting
* Initial attempt at sse extension tests
* Make existing tests pass
* Add test case for #916
* backport test for #916 to hx-sse for completeness
* add a distinct test for sse-swap
* more tests, fix tests
+ more tests differentiating the implementation of sse-swap and
hx-trigger
* fix for "is closed after removal with no close and activity"
* Cleanup listener if element is gone
* Revert "Cleanup listener if element is gone"
This reverts commit bbe3715276302656d1a422fbc4778842776e35f6.
* Clean up listeners if element gone, fix indenting
* Incorportate feedback from first review
+ seperate eventSource creation logic and event registering logic
+ manually create event handling, still confused by how hx-trigger works
* `createEventSourceOnElement` now looks for event sources in children
* explicitly handle legacy sse handling instead of having extra
selectors in `querySelectorOnThisOrChildren`
+ a few readability changes
+ add regression check to make sure that sseEventSource is only created
on elements with sse-connect or equivalent
+ add test to make sure that sse-connect in the child of a swapped
element is handled
* Allow multiple listeners in eventsource mock
* Backport eventsource mocks to legacy tests
---------
Co-authored-by: 1cg <469183+1cg@users.noreply.github.com>
* Allow CSS selectors with whitespace in `hx-trigger`
Parsing of `hx-trigger` scans for whitespace, so if a CSS selector is used that contains whitespace, e.g. `form input`, a syntax error is raised.
A workaround is implemented by allowing such a CSS selector to be wrapped in either curly braces or parentheses.
* Add explanation whitespace in CSS selector to docs
* Tests for CSS selectors containing whitespace
* Use faster RegEx test, remove redundant variable declarations
* Added Descendant Combinator support to `root` and `target` modifiers
* Add missing semicolon
* Tests for descendant combinators in `root` and `target` modifiers
* Improve descendant combinator test coverage
* Current behavior testing
Testing current library behavior
* Test should remove correct handler
* Add question in htmx:confirm event detail
* Allow skipping window.confirm
* Additional test without hx-confirm value
* Wrap htmx.off in finally
* More correct assertion in case of no calls to confirm
* Remove erroneously added formatting
* Remove erroneously added formatting
* Documentation, fix loop
---------
Co-authored-by: mat <matt@techspace.asia>
* Make htmx IE11 compatible again + tests IE11 compatible
* IE11 compatible handmade socket mock for ws-ext tests
* Fallback when xpath isn't supported, hx-on wildcard now working on IE11
* Merge remote-tracking branch 'upstream/relative-url-in-hx-boost' into ie11-compatibility
This resolves a bug when file system URLs are used on Windows. It
reverts part of a previous change where URLs were normalized prior to
the request in order to reduce key collisions in the history cache.
Because htmx's history cache is based entirely on response URLs, not
request URLs, I am reasonably confident that reverting this line doesn't
affect that optimization at all, but in any case it causes a bug which
is more important.
New function to turn off the logger after using `logAll()`. I applied to
this to the tests as well so that `logAll()` is only used in the specfic
places where debug output is required. This makes the console output of
the tests dramatically more useful when run from the command line.
* Add support for hx-on:* attribute
* Use shorter XPath selector
* Add support for "hx-on::" shorthand
* Add note about mixing the two forms of `hx-on`
* Normalize URL used as a cache key for history, without breaking IE.
* Update test `boosted element hx-disinherit sanity check` to check that the fully resolved href of the boosted element matches request.detail.requestConfig.path
The `htmx:trigger` event is fired whenever an HTMX AJAX request would
be. Now, `hx-trigger` can also be specified on an element without an
AJAX request, which will simply fire the `htmx:trigger` event on the
action specified by the attribute, and do nothing else.
This pattern allows for specifying client-side actions that don't
require a network request, while remaining within the HTMX control
paradigm.
* Add hx-history="false" to prevent sensitive or session data entering the localStorage cache
* Correctly get value of hx-history attribute.
* Add test for hx-history attribute.
* Allow history cache to be disabled when multiple `hx-history` attributes are present in a document, and any one of them is set to false.
* Add documentation for `hx-history` attribute.
* Add `hx-history` attribute to reference doc
* Correct the order of the `hx-history` attr in the reference doc.
* Add events support for WebSockets
* Stop reconnection attempts after element has been removed
* Add tests for websockets
* Hide socker wrapper behind a more strict interface to avoid breaking changes in the future
* Fix legacy websocket tests interfering with new extension tests
* Minor doc fixes
* Add `wsBinaryType` configuration option
