How to adapt an autocomplete/select field to work with server-side filtering and pagination

Introduction

In front-end development, there’s an extensive selection of component frameworks that offer easy solutions to most kinds of problems. Very often, though, you come across issues that require customization. Some frameworks allow this to a greater extent than others, and not all are equally easy to customize. Vuetify is one of the most feature-rich frameworks with very detailed documentation. In practice, though, investigating some seemingly trivial functionalities and coming up with an optimized solution can still take significant time.

Identifying the Challenge

Vuetify’s autocomplete component is amazing. It gives the user various options when it comes to customization, both visual and functional. While some modes can be triggered with a single property, others require a lot more effort, and the way to solution is not always straightforward. In this article, I will go over my solution for implementing server-side filtering and pagination, leveraging the concept of infinite scrolling. Additionally, the techniques discussed here can also be applied to the v-select component.

The Solution: Server-side Enhancements

In this chapter we will outline our solution for boosting v-autocomplete with server-side logic. First, we will wrap it into our own custom component that will be used to make further adjustments. Using the built-in append-item slot in combination with Vuetify’s v-intersect directive we will implement the so-called infinite scroll. That means we will only load a handful of records at first. With the aforementioned combination, we will detect when the bottom of the list is reached. At that point, we will automatically send a follow-up request to load the next page of records, until we finally reach the bottom.

After that, we will expand our solution to include filtering by making adjustments to v-autocomplete‘s properties, disabling front-end filtering, adding adequate indicators and handling scroll positions to ensure a smooth and intuitive experience for the end-user. We will end up with something like this:

Setting Things Up

The technical implementation will be demonstrated with Vue, my preferred framework for everyday work, combined with Vuetify, a very robust and highly customizable component framework commonly used in the Vue ecosystem. Note that concepts used here can be applied using other combinations of popular JavaScript technologies.

The solution will differ slightly depending on the versions of Vue and Vuetify. Since versions 3.x of both have been out for quite some time and are now industry standards, I will be using them. I will, however, leave important notes for Vue 2/Vuetify 2, as many active projects still use them. The differences are usually minor, except when it comes to accessing inner Vuetify elements (which is harder to do in Vue 3, as $refs isn't supported).

To begin, we will create a new blank project. You can skip this paragraph if you’re looking to add the solution to an existing project. Using Node Package Manager (NPM), we will create the project with the command: npm create vue@latest. The default settings are fine for our purposes, but if you prefer, you can change them. I enabled ESLint and Prettier options. There are other ways to initiate a Vue project, but I prefer this one because it uses Vite as the development server by default.

Next, we need to add Vuetify and the basic dependencies not included in it. Unless you chose another icon font or prefer another option for CSS, you can run this: npm install vuetify @mdi/font sass. Following the official documentation you can set up Vuetify in the main.js file. Don’t forget the font line if you’re using MDI icons like I am.

// file: main.js
import './assets/main.css';

import { createApp } from 'vue';
import App from './App.vue';

import '@mdi/font/css/materialdesignicons.css';
import 'vuetify/styles';
import { createVuetify } from 'vuetify';
import { VAutocomplete } from 'vuetify/components';
import { Intersect } from 'vuetify/directives';

const vuetify = createVuetify({
  components: { VAutocomplete },
  directives: { Intersect }
});

createApp(App).use(vuetify).mount('#app');

For our back-end, I opted to use a free API service with fake data called JSON Placeholder. While it’s not something you would use in a production app, it’s a simple and free service that can provide us with everything we need with minimal adjustments.

Now, let’s dive into the actual coding process. Create a new Vue file inside the components directory. Name it however you want - I chose PaginatedAutocomplete.vue. Add a template section containing a single v-autocomplete element. To populate this element with data, we'll define a records property that will be passed to the component.

For some minor styling adjustments, consider adding classes or props to limit the width of the autocomplete field and its dropdown menu to around 300px, preventing it from stretching across the entire window width.

<!-- file: PaginatedAutocomplete.vue -->
<template>
  <v-autocomplete :items="items" :menu-props="{ maxWidth: 300 }" class="autocomplete">
    <!--  -->
  </v-autocomplete>
</template>

<script setup>
defineProps({
  items: {
    type: Array,
    required: true
  }
});
</script>

<style lang="scss" scoped>
.autocomplete {
  width: 300px;
}
</style>

In the App.vue file, we can delete or comment out the header and Welcome components and import our newly created PaginatedAutocomplete.vue. Add the data ref that will be used for it: records, and set its default value to an empty array.

<!-- file: App.vue -->
<script setup>
import { ref } from 'vue';

import PaginatedAutocomplete from './components/PaginatedAutocomplete.vue';

const records = ref([]);
</script>

<template>
  <main>
    <PaginatedAutocomplete :items="records" />
  </main>
</template>

Adjust global styles if you prefer. I changed the color scheme from dark to light in base.css and added some centering CSS to main.css.

That completes the initial setup. So far, we only have a basic autocomplete component with empty data.

Controlling Data Flow with Infinite Scroll

Moving forward, we need to load the data from the server. As previously mentioned, we will be utilizing JSON Placeholder, specifically its /posts endpoint. To facilitate data retrieval, we will install Axios with npm install axios.

In the App.vue file, we can now create a new method to fetch those records. It’s a simple GET request, which we follow up by saving the response data into our records data property. We can call the function inside the onMounted hook, to load the data immediately. Our script section will now contain this:

<!-- file: App.vue -->
<script setup>
import { ref, onMounted } from 'vue';
import axios from 'axios';

import PaginatedAutocomplete from './components/PaginatedAutocomplete.vue';

const records = ref([]);

function loadRecords() {
  axios
    .get('https://jsonplaceholder.typicode.com/posts')
    .then((response) => {
      records.value = response.data;
    })
    .catch((error) => {
      console.log(error);
    });
}

onMounted(() => {
  loadRecords();
});
</script>

To improve the visual user experience, we can add another data prop called loading. We set it to true before sending the request, and then revert it to false after the response is received. The prop can be forwarded to our PaginatedAutocomplete.vue component, where it can be tied to the built-in v-autocomplete loading prop. Additionally, we can incorporate the clearable prop. That produces the following code:

<!-- file: Paginated Autocomplete.vue -->
<template>
  <v-autocomplete
    :items="items"
    :loading="loading"
    :menu-props="{ maxWidth: 300 }"
    class="autocomplete"
    clearable
  >
    <!--  -->
  </v-autocomplete>
</template>

<script setup>
defineProps({
  items: {
    type: Array,
    required: true
  },

  loading: {
    type: Boolean,
    required: false
  }
});
</script>

// file: App.vue
// ...
const loading = ref(false);

function loadRecords() {
  loading.value = true;

  axios
    .get('https://jsonplaceholder.typicode.com/posts')
    .then((response) => {
      records.value = response.data;
    })
    .catch((error) => {
      console.log(error);
    })
    .finally(() => {
      loading.value = false;
    });
}
// ...

<!-- file: App.vue -->
<!-- ... -->
<PaginatedAutocomplete :items="records" :loading="loading" />
<!-- ... -->

At this point, we have a basic list of a hundred records, but it’s not paginated and it doesn’t support searching. If you’re using Vuetify 2, the records won’t show up correctly - you will need to set the item-text prop to title. This is already the default value in Vuetify 3. Next, we will adjust the request parameters to attain the desired behavior. In a real project, the back-end would typically provide you with parameters such as page and search/query. Here, we have to get a little creative. We can define a pagination object on our end with page: 1, itemsPerPage: 10 and total: 100 as the default values. In a realistic scenario, you likely wouldn’t need to supply the first two for the initial request, and the third would only be received from the response. JSON Placeholder employs different parameters called _start and _limit. We can reshape our local data to fit this.

// file: App.vue
// ...
const pagination = ref({
  page: 1,
  perPage: 10,
  total: 100
});

function loadRecords() {
  loading.value = true;

  const params = {
    _start: (pagination.value.page - 1) * pagination.value.perPage,
    _limit: pagination.value.perPage
  };

  axios
    .get('https://jsonplaceholder.typicode.com/posts', { params })
    .then((response) => {
      records.value = response.data;
      pagination.value.total = response.headers['x-total-count'];
    })
    .catch((error) => {
      console.log(error);
    })
    .finally(() => {
      loading.value = false;
    });
}
// ...

Up to this point, you might not have encountered any new concepts. Now we get to the fun part - detecting the end of the current list and triggering the request for the next page of records. Vuetify has a directive called v-intersect, which can inform you when a component you attached it to enters or leaves the visible area in your browser. Our interest lies in its isIntersecting return argument. The detailed description of what it does can be found in MDN Web Docs. In our case, it will allow us to detect when we’ve reached the bottom of the dropdown list. To implement this, we will attach the directive to our v-autocomplete‘s append-item slot.

To ensure we don’t send multiple requests simultaneously, we display the element only when there’s an intersection, more records are available, and no requests are ongoing. Additionally, we add the indicator to show that a request is currently in progress. This isn’t required, but it improves the user experience. Vuetify’s autocomplete already has a loading bar, but it might not be easily noticeable if your eyes are focused on the bottom of the list. We also need to update the response handler to concatenate records instead of replacing them, in case a page other than the first one was requested.

To handle the intersection, we check for the first (in Vuetify 2, the third) parameter (isIntersecting) and emit an event to the parent component. In the latter, we follow this up by sending a new request. We already have a method for loading records, but before calling it, we need to update the pagination object first. We can do this in a new method that encapsulates the old one. Once the last page is reached, we shouldn’t send any more requests, so a condition check for that should be added as well. With that implemented, we now have a functioning infinite scroll.

<!-- file: PaginatedAutocomplete.vue -->
<template>
  <v-autocomplete
    :items="items"
    :loading="loading"
    :menu-props="{ maxWidth: 300 }"
    class="autocomplete"
    clearable
  >
    <template #append-item>
      <template v-if="!!items.length">
        <div v-if="!loading" v-intersect="handleIntersection" />

        <div v-else class="px-4 py-3 text-primary">Loading more...</div>
      </template>
    </template>
  </v-autocomplete>
</template>

<script setup>
defineProps({
  items: {
    type: Array,
    required: true
  },

  loading: {
    type: Boolean,
    required: false
  }
});

const emit = defineEmits(['intersect']);

function handleIntersection(isIntersecting) {
  if (isIntersecting) {
    emit('intersect');
  }
}
</script>

// file: App.vue
// ...
function loadRecords() {
  loading.value = true;

  const params = {
    _start: (pagination.value.page - 1) * pagination.value.perPage,
    _limit: pagination.value.perPage
  };

  axios
    .get('https://jsonplaceholder.typicode.com/posts', { params })
    .then((response) => {
      if (pagination.value.page === 1) {
        records.value = response.data;
        pagination.value.total = response.headers['x-total-count'];
      } else {
        records.value = [...records.value, ...response.data];
      }
    })
    .catch((error) => {
      console.log(error);
    })
    .finally(() => {
      loading.value = false;
    });
}

function loadNextPage() {
  if (pagination.value.page * pagination.value.perPage >= pagination.value.total) {
    return;
  }

  pagination.value.page++;

  loadRecords();
}
// ...

Efficiency Meets Precision: Moving Search to the Back-end

To implement server-side searching, we begin by disabling from-end filtering within the v-autocomplete by adjusting the appropriate prop value (no-filter). Then, we introduce a new property to manage the search string, and then bind it to v-model:search-input (search-input.sync in Vuetify 2). This differentiates it from the regular input. In the parent component, we capture the event, define a query property, update it when appropriate, and reset the pagination to its default value, since we will be requesting page one again. We also have to update our request parameters by adding q (as recognized by JSON Placeholder).

<!-- file: PaginatedAutocomplete.vue -->
<template>
  <v-autocomplete
    :items="items"
    :loading="loading"
    :menu-props="{ maxWidth: 300 }"
    class="autocomplete"
    clearable
    no-filter
    v-model:search-input="search"
    @update:search="emitSearch"
  >
    <template #append-item>
      <template v-if="!!items.length">
        <div v-if="!loading" v-intersect="handleIntersection" />

        <div v-else class="px-4 py-3 text-primary">Loading more...</div>
      </template>
    </template>
  </v-autocomplete>
</template>

<script setup>
import { ref } from 'vue';

defineProps({
  items: {
    type: Array,
    required: true
  },

  loading: {
    type: Boolean,
    required: false
  }
});

const emit = defineEmits(['intersect', 'update:search-input']);

function handleIntersection(isIntersecting) {
  if (isIntersecting) {
    emit('intersect');
  }
}

const search = ref(null);

function emitSearch(value) {
  emit('update:search-input', value);
}
</script>

<!-- file: App.vue -->
<script>
// ...
const query = ref(null);

function handleSearchInput(value) {
  query.value = value;

  pagination.value = Object.assign({}, { page: 1, perPage: 10, total: 100 });

  loadRecords();
}

onMounted(() => {
  loadRecords();
});
</script>

<template>
  <main>
    <PaginatedAutocomplete
      :items="records"
      :loading="loading"
      @intersect="loadNextPage"
      @update:search-input="handleSearchInput"
    />
  </main>
</template>

// file: App.vue
// ...
function loadRecords() {
  loading.value = true;

  const params = {
    _start: (pagination.value.page - 1) * pagination.value.perPage,
    _limit: pagination.value.perPage,
    q: query.value
  };
// ...

If you try the search now and pay attention to the network tab in developer tools, you will notice that a new request is fired off with each keystroke. While our current dataset is small and loads quickly, this behavior is not suitable for real-world applications. Larger datasets can lead to slow loading times, and with multiple users performing searches simultaneously, the server could become overloaded. Fortunately, we have a solution in the Lodash library, which contains various useful JavaScript utilities. One of them is debouncing, which allows us to delay function calls by leaving us some time to call the same function again. That way, only the latest call within a specified time period will be triggered. A commonly used delay for this kind of functionality is 500 milliseconds. We can install Lodash by running the command npm install lodash. In the import, we only reference the part that we need instead of taking the whole library.

// file: PaginatedAutocomplete.vue
// ...
import debounce from 'lodash/debounce';
// ...

// file: PaginatedAutocomplete.vue
// ...
const debouncedEmit = debounce((value) => {
  emit('update:search-input', value);
}, 500);

function emitSearch(value) {
  debouncedEmit(value);
}
// ...

Now that’s much better! However, if you experiment with various searches and examine the results, you will find another issue - when the server performs the search, it takes into account not only post titles, but also their bodies and IDs. We don’t have options to change this through parameters, and we don’t have access to the back-end code to adjust that there either. Therefore, once again, we need to do some tweaking of our own code by filtering the response data. Note that in a real project, you would discuss this with your back-end colleagues. Loading unused data isn’t something you would ever want!

// file: App.vue
// ...
.then((response) => {
      const recordsToAdd = response.data.filter((post) => post.title.includes(params.q || ''));

      if (pagination.value.page === 1) {
        records.value = recordsToAdd;
        pagination.value.total = response.headers['x-total-count'];
      } else {
        records.value = [...records.value, ...recordsToAdd];
      }
    })
// ...

To wrap up all the fundamental functionalities, we need to add record selection. This should already be familiar to you if you’ve worked with Vuetify before. The property selectedRecord is bound to model-value (or just value in Vuetify 2). We also need to emit an event on selection change, @update:model-value, (Vuetify 2: @input) to propagate the value to the parent component. This configuration allows us to utilize v-model for our custom component.

Because of how Vuetify’s autocomplete component works, both record selection and input events are triggered when a record is selected. Usually, this allows more customization options, but in our case it’s detrimental, as it sends an unnecessary request and replaces our list with a single record. We can solve this by checking for selected record and search query equality.

// file: App.vue
// ...
function handleSearchInput(value) {
  if (selectedRecord.value === value) {
    return;
  }

  query.value = value;

  pagination.value = Object.assign({}, { page: 1, perPage: 10, total: 100 });

  loadRecords();
}

const selectedRecord = ref(null);
// ...

<!-- file: App.vue -->
<template>
  <main>
    <PaginatedAutocomplete
      v-model="selectedRecord"
      :items="records"
      :loading="loading"
      @intersect="loadNextPage"
      @update:search-input="handleSearchInput"
    />
  </main>
</template>

<!-- file: PaginatedAutocomplete.vue -->
<!-- ... -->
  <v-autocomplete
    :items="items"
    :loading="loading"
    :menu-props="{ maxWidth: 300 }"
    :model-value="selectedItem"
    class="autocomplete"
    clearable
    no-filter
    v-model:search-input="search"
    @update:model-value="emitSelection"
    @update:search="emitSearch"
  >
<!-- ... -->

// file: PaginatedAutocomplete.vue
// ...
const emit = defineEmits(['intersect', 'update:model-value', 'update:search-input']);

function handleIntersection(isIntersecting) {
  if (isIntersecting) {
    emit('intersect');
  }
}

const selectedItem = ref(null);

function emitSelection(value) {
  selectedItem.value = value;

  emit('update:model-value', value);
}
// ...

Almost done, but if you are thorough with your testing, you will notice an annoying glitch - when you do a search, scroll down, then do another search, the dropdown scroll will remain in the same place, possibly causing a chain of new requests in quick succession. To solve this, we can reset the scroll position to the top whenever a new search input is entered. In Vuetify 2, we could do this by referencing the internal v-menu of v-autocomplete, but since that’s no longer the case in Vuetify 3, we need to get creative. Applying a unique class name to the menu allows us to select it through pure JavaScript and then follow up with necessary adjustments.

<!-- file: PaginatedAutocomplete.vue -->
<!-- ... -->
<v-autocomplete
  ...
  :menu-props="{ maxWidth: 300, class: `dropdown-${uid}` }"
  ...
>
<!-- ... -->

// file: PaginatedAutocomplete.vue
// ...
const debouncedEmit = debounce((value) => {
  emit('update:search-input', value);

  resetDropdownScroll();
}, 500);

function emitSearch(value) {
  debouncedEmit(value);
}

const uid = Math.round(Math.random() * 10e4);

function resetDropdownScroll() {
  const menuWrapper = document.getElementsByClassName(`dropdown-${uid}`)[0];
  const menuList = menuWrapper?.firstElementChild?.firstElementChild;

  if (menuList) {
    menuList.scrollTop = 0;
  }
}
// ...

There we have it, our custom autocomplete component with server side filtering and pagination is now complete! It was rather simple in the end, but I’m sure you would agree that the way to the solution was anything but with all these little tweaks and combinations we had to make.

If you need to compare anything with your work, you can access the source files through a GitHub repository here.

Overview and Conclusion

The journey doesn’t have to end here. If you need further customization, you can search the Vuetify documentation for ideas. There are countless possibilities waiting to be explored. For instance, you can try working with multiple values at once. This is already supported by Vuetify, but may need additional adjustments to be combined with our solutions. Still, this is a useful thing to have in many projects. Or, you can try template customization. You have the power to redefine the look and feel of your selection templates, list templates, and more. This opens the door to crafting user interfaces that align perfectly with your project's design and branding.

There are many other options in addition to those. In fact, the depth of customization available warrants the creation of additional articles to cover these advanced topics comprehensively. In the end, the Vue + Vuetify stack isn’t the only one that supports something like this. If you work with other frameworks, I encourage you to try to develop an equivalent to this on your own.

In conclusion, we transformed a basic component into a specialized solution that fits our needs. You've now armed yourself with a versatile tool that can be applied to a wide range of projects. Whenever you find yourself working with extensive lists of records, a server-side pagination and filtering solution becomes your go-to strategy. It not only optimizes the performance from the server's perspective, but also ensures a smoother rendering experience for your users. With some tweaks, we addressed some common issues and opened up new possibilities for further adjustments.