KEML is a lightweight flexible alternative to HTMX, built around the concept of actions.
The ideas that HTMX has recently made popular are, frankly, the best that the web community collectively has ever had.
However, like most great ideas, they are far more important than any actual specific implementation.
Being small and fast and configuration/plugin free is not enough, when you are trying to compete with a well established and popular library.
KEML was born out of the classic 1-to-1 problem of the traditional jQuery-esque web application, that the HTMX api does nothing to address.
Consider the following "idiomatic" HTMX code:
<button hx-post="/clicked"
hx-trigger="click"
hx-target="#result"
hx-swap="innerHTML"
>
Click Me!
</button>
<div id="result"></div>Here the button element:
- can only react to exactly 1 event ("click")
- can only do 1 thing when that event happens (send a request to "/clicked")
- cannot delegate the request-sending to some other element/-s
- can only render the result into 1 (usually) target element
- has to know where that element is in the page, what its "id" and/or "class" attributes are
- has to decide for the target element the exact "hx-swap", of which there can only be 1
Out of these limitations arise:
- the need for a custom selector syntax built on top of the normal css-selector syntax
- the need for selectors in the first place, which you need to learn and understand to use the library effectively
- the implicit special handling of certain elements, like title and meta, present in the server response
- out of band swaps
- response selectors
- "hx-preserve"
- etc, etc, etc...
All which, are meant to solve real tangible application needs, but in the process of doing that over-complicate things that do not need to be complicated.
Consider the following KEML code, that works with the same backend:
<html>
<head>
<title render="result"></title>
</head>
<body>
<button on:click="handleClick doSomethingElse"
on:dblclick="handleDoubleClick"
>
Click Me!
</button>
<button on:click="handleClick">
Click Me, maybe?!
</button>
<input on="handleClick"
post="/clicked"
type="text"
name="input1"
result="result"
>
<input on="handleClick"
put="/notification"
type="text"
name="input2"
>
<div render="result"
position="replaceWith"
></div>
<p render="result"
position="append"
></p>
</body>
</html>- both buttons initiate the same "handleClick" action on "click"
- the first button actually initiates two independent actions on "click", that could both do completely different things
- the first button also reacts to the double click event
- neither of the buttons needs to know how those actions are being handled, if at all
- the two text inputs both handle the "handleClick" action, but send completely different requests
- the first input gives the server response a render-able name "result"
- neither of the inputs knows anything about the rendering, whether or not anything is even going to be rendered at all, where and how
- the div, p and title elements render the same server response differently
- the div will be completely replaced with the response
- the p will append the response after its last child
- the title will replace all of its children with the response
- there is nothing special about the title element at all
- there's no need for ids, classes or selectors and the location of each element in the document is completely unimportant
None of the problems that HTMX tries to solve with the complications listed above even exist in this paradigm!
While nothing is ever truly complete, the current feature-set should be able to cover 99% of actually useful HTMX features.
I'm not going to claim that it is completely bug free and supports every browser under the sun, because it still has a long way to go until that becomes a reality.
Thus, all constructive feedback and criticism are very welcome!
If a feature of HTMX is missing, that means one of the following:
- it is made unnecessary by a more flexible API
- it is downright evil and/or going against the spirit of HATEOAS (e.g JSON endpoints, local templates and most forms of local state)
- it wasn't implemented yet