alt Hacker NewsThis isn't about the number of rendering logics. You'll have as many as you have variants, that's tautological. This is about where they happen.
In a classic app, there's one entity that keeps the state (the server), and one entity that keeps how it is rendered. This is very easy to reason about, and the contract is very clear. If I want to understand what happens, then I can open my frontend app and see "Hello <b>{{name}}</b>".
In HTMX, the logic is spread. What I see is a construct that says "Replace this with the mystery meat from the backend, whatever that is".
Assume there's a styling issue. The name looks too big. Is it because the frontend is adding a style that cascades, or is it because the backend returns markup with a class? Now any issue has an extra level of indirection, because you've spread your rendering logic into two places.
> which is what you just did by dismissing the reality that client-side requires the same 3 renderers that server-side requires!
But what's complex isn't the number of renderers, it's where the logic happens. The HTMX website is full of examples where the header of a table is rendered in the frontend, and the body is rendered in the backend. How's that considered sane when changing the order of the columns turns into a real ordeal?
> If I want to understand what happens, then I can open my frontend app and see "Hello <b>{{name}}</b>".
> In HTMX, the logic is spread
I disagree... with React, by definition, the logic is spread. Persistent data, and usually, business logic, is in some data store accessible via the app backend. And then a totally different entity, the front-end, renders that data (often implementing additional business logic) and manages state (which is often not yet recorded in the data store until various updates can be performed).
HTMX helps keep everything aligned. All the rendering logic is right there along with the data and the business logic. If I'm looking for a renderer, not only is it easy to find the template that produced "Hello <b>{{name}}</b>" but it is also easy to find the source of {{name}}. Which also makes it easy to alter {{name}}, say, from Smith, John to Mr. John Smith - because the data store and business logic are right there, it is low effort to switch the order and to begin including the salutation.
Your "front-end" is still all in one place, except it's on the server, and typically rendered via templates instead of React components. But the templates can often access native objects (including their properties and functions) instead of solely relying on JSON objects.
This comment is already long but regarding data tables... yea, highly dynamic pure data-based UI's such as charts and tables aren't HTMX's forte. But even then there are ways... the data- attribute is very useful, and since you are already using JS to handle sorting, filtering, re-ordering, showing tooltips, etc, it's very possible to render valid HTML fragments that can be properly rendered via that same JS (or contain data which can be).