ive just realized this is a lot more complicated if i want to actually mind all of the other

hx-

attributes during my manual swap... i can do a simple innerHTML replacement but that may not be what the srcElement specified

would be handy in the internalAPI

swap()

function was on the public API 😄

😑

the plot thickens... because the HTML being swapped in also has

hx-

attrs, it needs to be processed after it's swapped.. this is turning out to be more complicated than i had hoped

i got the swap working, but then the control is dead after the swap

I believe if you return the swapped elements they should be processed

ohh interesting, let me check that

that doesnt appear to be true when using the

onEvent

extension hook

maybe i should be using

handleSwap

🤔

nah nevermind that doesnt work because i dont get the original target selector -- only the first target element

i got it working using

onEvent

and calling

process

on the swapped node after i swap it 👍

Hi, does anyone here have experience with how to effectively use HTMX to dynamically replace the content by an Ajax request, but with the fact that the replaced HTML also contains ? I have the feeling from using HTMX in the past that HTMX removes tags from the replaced HTML, what is big problem for our use-cases. We also need to ensure that hx-boost or other content replacement functions updates the and add , or tags for CSS styles. Is here someone with recommendations related to this?

script tags in the content will be executed

merging the head tag entirely, including scripts, etc. doesn't currently work. I'd like to make it work at some point, but it's a ways off.

sorry 😑

Hotwire is laying the bricks

follow the lead

Thank you for your response @mysterious-toddler-20573. We want to prepare some plugins for the Astro framework for nice integration with HTMX. This functionality is absolutely critical. What do you see as the biggest risk/problem for implementing CSS/JS integration in ? One of the possible implementations that came to my mind is to implement an HTMX extension, which would check, as part of the content replacement, whether there are any , or tags in the new HTML (retrieved by Ajax) that have not yet been inserted in our page and if not, create these tags at the end of the HTML and add them to the DOM. In order to be able to effectively verify whether the given script (either inline or external via the src attribute) has already been inserted into the HTML, we would add some special HTML element with the ID "htmx-extras" as the last element in the . What do you think about this solution?

Thank you for your reaction @most-flag-2080. We researched HTMX, Unpoly and HotWire, and we like HTMX the most. Unpoly is very powerful, but the data is very large. However, all of these frameworks have their strengths and weaknesses, and unfortunately, not one of them is universal enough.

In the next few weeks, I will try to implement in them as extensions what we basically lack.

If you hook into the

htmx:afterRequest

event: https://htmx.org/events/#htmx:afterRequest you can grab the xhr, get the content out and have a go at the header. @most-flag-2080 is right in this case: Turbolinks I believe has been doing this for a long time. The boost functionality was added to htmx as an afterthought, I was more focused on making HTML more expressive and it felt too "magical" to me, so I kept it simple and effectively said "load all your JS up front or inline (w/ hyperscript. for example").

I started buckling under when I added support for the title tag, and I've accepted that at some point I'll need to do a full header merge, but I have been putting it off

I understand you @mysterious-toddler-20573. This is no longer a trivial task and will have its limits. It will not be possible to capture all situations perfectly. But I can imagine it as real. The newly discovered and tags (in loaded HTML by Ajax requests) for external styles could be inserted into the HTML before adding/replacing the HTML from the Ajax request. So, after replacing HTML, the content will be styled straight away and will not flicker. Scripts (both inline and external via the src attribute) would in turn be inserted in some clever way at the end of the HTML. Of course, in order to prevent the same JavaScript from being inserted and executed repeatedly, it would be possible to use some data attribute in the tag to set whether this javascript should be inserted and therefore interpreted again, even if it has already been inserted once before. To avoid solving this problem, Unpoly guides people to use

up.compiler()

(https://unpoly.com/up.compiler), which is quite an elegant way, but it also has certain limits for large projects, especially due to the impossibility of using lazy-loading techniques effectively. If we were not concerned with performance, of course we would load all the styles and JavaScripts for the entire website/application directly on the homepage, but this is not a suitable way.

Yep, the problem is interesting. I think we could do a "merge" where we compare the old and new content, probably w/ a dumb O(n^2) algorithm, merge in the new styles and scripts, remove the old ones

throw in a few events for people to hook into on add/remove (to maybe filter, or de-init javascript) and I think it'd be pretty clean and not a ton of code

happy to help with it, I like implementing it as an extension in htmx 1 and looking on integrating it into the core for htmx 2 (2023, unfortunately)

@mysterious-toddler-20573 - thank you very much for the consultation. In the coming weeks, during the implementation of one of our projects, I will try to prepare a functional proof-of-concept and I will consult it with you.

Thinking on it I wonder if it would dovetail with the idiomorph work, which implements morphing w proper node removal etc

I’m using htmx to update a div in a datatable. It’s kind working but I have to refresh the page the see the update. I googled the problem. It seems I need to use htmx.process function. But I couldn’t figure out where to put it. Has anyone had the same issue before?

Yep the problem is interesting I think