Modern UI frameworks offer abstraction layers that make user interfaces declarative and reactive. However, the web platform itself exposes primitives that can be composed to achieve similar patterns without introducing a dedicated UI library. This article demonstrates an experimental approach for creating a reactive, declarative UI flow using only vanilla JavaScript, Web APIs, and Proxy-based state tracking.
The purpose of the experiment is to examine how far native capabilities can be pushed without framework-level abstractions and to illustrate architectural benefits of declarative behavior in UI code: improved clarity, maintainability, and reduced coupling.
The Target Behavior
The experiment focuses on a practical business scenario:
Display a modal dialog that performs periodic polling of an API endpoint. The dialog should remain open until a specific condition is met, then resolve or reject accordingly.
The modal dynamically:
- mounts itself into the DOM
- starts and manages a polling process
- exposes reactive internal state
- updates based on the polling result
- closes automatically when finished
- provides optional developer controls
The primary requirement is that consumer code defines what should happen, not how to wire it.
Declarative Usage Example
Example invocation:
uiModalEngine.showPollingDialog({
endpoint: `${getServiceBaseUrl()}/process/wait_for/confirmation`,
requestPayload: () => ({
taskId: currentTask.id,
mode: "rapid",
includeAudit: true,
}),
requestOptions: {
method: "POST",
headers: { "Content-Type": "application/json" },
},
shouldContinue: (response) => response.ok && response.pending === true,
intervalMs: 1000,
buildContent: (mountNode) => {
const contentBlock = uiModalEngine.createContentBlock({
title: "Waiting for confirmation...",
description: "This dialog will close automatically once the operation completes.",
})
mountNode.appendChild(contentBlock)
},
onResolved: ({ dialogNode, response }) => {
metrics.track("operation_confirmed")
dialogNode.remove()
},
onRejected: ({ dialogNode, error }) => {
logger.error("operation_polling_failed", error)
dialogNode.remove()
},
devToolsEnabled: false,
})Declarative Takeaways
| Concern | Ownership |
|---|---|
| UI behavior | Declarative configuration |
| UI rendering | Modal orchestrator |
| DOM structure | DOM utility layer |
| polling logic | polling helper |
| reactive state | Proxy-based tracker |
No framework is involved, yet responsibilities remain clearly segmented.
Core Building Block: DOM Utility Layer
To keep high-level code focused on behavior, DOM creation is delegated to a lightweight utility:
class DomToolkit {
constructor(doc) {
this.doc = doc
}
static getInstance(doc) {
if (!DomToolkit.instance) DomToolkit.instance = new DomToolkit(doc)
return DomToolkit.instance
}
createElement({ tag, classes, id, attrs = {}, styles = {}, html }) {
const el = this.doc.createElement(tag)
if (id) el.id = id
if (typeof classes === "string") el.classList.add(classes)
if (Array.isArray(classes)) classes.forEach(c => el.classList.add(c))
Object.entries(attrs).forEach(([k, v]) => el.setAttribute(k, v))
Object.entries(styles).forEach(([k, v]) => el.style[k] = v)
if (html != null) el.innerHTML = html
return el
}
}
const domToolkit = DomToolkit.getInstance(document)This removes boilerplate from business logic and centralizes standard element configuration.
Reactive State via Proxy
Next, the experiment introduces deep reactive state using the native Proxy object. This allows mutations at arbitrary depth to be observed without requiring explicit setters.
class DeepStateProxy {
constructor(target, { onSet, onDelete } = {}) {
this.onSet = onSet
this.onDelete = onDelete
return this.wrap(target, [])
}
wrap(node, path) {
if (!node || typeof node !== "object") return node
const handler = {
set: (target, key, value) => {
const fullPath = [...path, key]
target[key] = this.wrap(value, fullPath)
this.onSet?.(value, fullPath)
return true
},
deleteProperty: (target, key) => {
if (!(key in target)) return false
const fullPath = [...path, key]
delete target[key]
this.onDelete?.(fullPath)
return true
},
}
Object.keys(node).forEach(k => {
node[k] = this.wrap(node[k], [...path, k])
})
return new Proxy(node, handler)
}
}Usage example:
const state = new DeepStateProxy({
attempts: 0,
lastResponse: null,
}, {
onSet: (value, path) => console.debug("state changed:", path.join("."), value),
})This approach:
✔ enables deep mutation tracking
✔ does not require libraries
✔ keeps state as plain objects
✔ keeps consumer code minimal
Polling Logic as a Reusable Abstraction
To isolate asynchronous logic:
async function runPolling({ task, shouldStop, intervalMs }) {
while (true) {
const result = await task()
if (shouldStop(result)) return result
await new Promise(res => setTimeout(res, intervalMs))
}
}Isolating polling enables:
- testability (polling logic has no DOM dependencies)
- readability (consumer describes behavior declaratively)
- reusability (polling can be embedded into other flows)
Modal Orchestrator
The orchestrator integrates DOM utilities, polling, and reactive state into a coherent unit:
ModalOrchestrator.prototype.showPollingDialog = function (cfg) {
const {
endpoint, requestPayload, requestOptions,
shouldContinue, intervalMs,
buildContent, onResolved, onRejected,
devToolsEnabled = false,
} = cfg
const dialogNode = this.createDialogShell({ buildContent })
document.body.appendChild(dialogNode)
const state = new DeepStateProxy({
attempts: 0,
polling: true,
aborted: false,
lastResponse: { ok: false },
}, {
onSet: (value, path) => {
if (devToolsEnabled) console.debug("state:", path.join("."), value)
},
onDelete: () => { throw new Error("state mutation violation") },
})
state.polling = true
runPolling({
task: async () => {
const payload = requestPayload()
const res = await fetch(endpoint, { ...requestOptions, body: JSON.stringify(payload) })
.then(r => r.json())
.catch(err => ({ ok: false, error: err.message, errored: true }))
if (!shouldContinue(res) && !res.errored) state.polling = false
else state.attempts++
state.lastResponse = res
return res
},
shouldStop: () => !state.polling,
intervalMs,
})
.then(res => onResolved?.({ dialogNode, response: res }))
.catch(err => onRejected?.({ dialogNode, error: err }))
}Note the absence of framework-specific concepts such as:
- components
- hooks
- virtual DOM
- stores
Yet the intent remains clear and maintainable.
Observations & Takeaways
Key architectural observations include:
-
Declarative descriptions scale better than imperative wiring
The consumer code reads as a behavioral specification, not as a set of instructions. -
Reusable utilities reduce future cost
DOM plumbing and polling logic are built once and reused many times. -
Native Web APIs are powerful enough for complex flows
Proxy,fetch,Promise, and basic DOM operators enable experimentation without dependencies. -
Frameworks are optional, whereas abstraction is not
Frameworks package abstractions; they are not the only way to achieve them.
Conclusion
This experiment shows that declarative UI behavior and reactive state management do not strictly require third-party frameworks. While production systems benefit from established ecosystems, understanding how native patterns can replicate core ideas provides architectural insight and improves reasoning about frameworks.
By relying solely on the web platform, the experiment highlights the expressive power of vanilla JavaScript, clarifies why modern frameworks emphasize declarativity and reactivity, and reinforces the idea that good abstractions—framework or not—ultimately enable scalable UI code.