Passing Search A Query Without Using The SearchBox Component In Algolia

You probably haven't noticed (and you'd be forgiven if this is the case!) that my site now has the ability to search through posts. This is a strange turn of events for me as I decided to remove search capability from my site many years ago as I didn't feel it added any benefits for the user. This became evident from Google Analytics stats where searches never hit high enough numbers to warrant having it. The numbers don't lie!

So what caused this turnaround?

I've noticed that I'm regularly referring back through posts to refresh myself on things I've done in the past and to find solutions to issues I know I've previously written about. Having a search would make trawling through my few hundred posts a lot easier. So this is more of a personal requirement than commercial. But there is an exciting aspect to this as well - experimenting with Algolia. Using Algolia search is something I've been meaning to look into for a long time and integrating with GatbsyJS.

The thought of having the good ol' magnifying glass back in the navigation makes me nostalgic!

Note: In this post, I won't be covering the basic Algolia setup or the plugins needed to install as there is already a great wealth of information online. Check out my "Useful Links" section at the end of the post.

Basic Setup

Integrating Algolia into GatbsyJS was relatively straight-forward due to the wealth of information that others have already written and also the plugins themselves. The plugins make light work of rendering search results quickly allowing enough customisations to the HTML markup for easy implementation within any site. By default, the plugins contain the following components:

  • InstantSearch
  • SearchBox
  • Hits
import algoliasearch from 'algoliasearch/lite';
import PropTypes from 'prop-types';
import { Link } from 'gatsby';
import { InstantSearch, Hits, Highlight, SearchBox } from 'react-instantsearch-dom';
import React from 'react';

// Get API keys from the environment file.
const appId = process.env.GATSBY_ALGOLIA_APP_ID;
const searchKey = process.env.GATSBY_ALGOLIA_SEARCH_KEY;
const searchClient = algoliasearch(appId, searchKey);

const SearchPage = () => (
  <InstantSearch
    searchClient={searchClient}
    indexName={process.env.GATSBY_ALGOLIA_INDEX_NAME}
  >
    <SearchBox />
    <Hits hitComponent={Hit} />
  </InstantSearch>
);

function Hit(props) {
  return (
    <article className="hentry post">
      <h3 className="entry-title">
        <Link to={props.hit.fields.slug}>
          <Highlight attribute="title" hit={props.hit} tagName="mark" />
        </Link>
      </h3>
      <div className="entry-meta">
        <span className="read-time">{props.hit.fields.readingTime.text}</span>
      </div>
      <p className="entry-content">
        <Highlight hit={props.hit} attribute="summary" tagName="mark" />
      </p>
    </article>
  );
}

Hit.propTypes = {
  hit: PropTypes.object.isRequired,
};

export default SearchPage;

The InstantSearch is the core component that directly interacts with Algolia's API and takes in two properties, "searchClient" and "indexName" containing the Application ID and Search Key that is acquired from the Algolia account setup. This component contains two child components, SearchBox is the search textbox and Hits that displays results from the search query.

It is the Hits component where we can customise the HTML with our own markup by using it's "hitComponent" attribute. In my case, I created a function to generate HTML where I access the properties from the search index. What's really cool is here is we have the ability to also highlight our search term where they may occur in the results by using the Highlight component (also provided by the Algolia plugin) and adding a "tagName" attribute.

Removing The SearchBox Component

The standard implementation may not suit all scenarios as you may want a search term to be sent to the InstantSearch component differently. For example, it could be from a custom search textbox or (as in my case) read from a query-string parameter. It wasn't until I started delving further into the standard setup I realised you cannot just remove the SearchBox component and pass a value directly, but there is a workaround.

I have expanded upon the code-snippet, above, to demonstrate how my search page works...

import algoliasearch from 'algoliasearch/lite';
import PropTypes from 'prop-types';
import { Link } from 'gatsby';
import { InstantSearch, Hits, Highlight, connectSearchBox } from 'react-instantsearch-dom';
import Layout from "../components/global/layout";
import React, { Component } from "react";

// Get API keys from the environment file.
const appId = process.env.GATSBY_ALGOLIA_APP_ID;
const searchKey = process.env.GATSBY_ALGOLIA_SEARCH_KEY;
const searchClient = algoliasearch(appId, searchKey);
const VirtualSearchBox = connectSearchBox(() => <span />);

class SearchPage extends Component { 
  state = {
    searchState: {
      query: '',
    },
  };

  componentDidMount() {   
    // Get "term" query string parameter value.
    let search = window.location.search;
    let params = new URLSearchParams(search);
    let searchTerm = params.get("term");

    // Send the query string value to a "searchState" object used by Algolia.
    this.setState(state => ({
      searchState: {
        ...state.searchState,
        query: searchTerm,
      },
    }));
 }

  render() {
      // Default "instantSearch" HTML to prompt user to enter a search term.
      var instantSearch = null;
      
      // If there is a search term, utilise Algolia's instant search.
      if (this.state.searchState.query) {
        instantSearch = <div className="entry-content">
                          <h2>You've searched for "{this.state.searchState.query}".</h2>
                          <div className="post-list archives-list">
                          <InstantSearch
                              searchClient={searchClient}
                              indexName={process.env.GATSBY_ALGOLIA_INDEX_NAME}
                              searchState={this.state.searchState}
                            >
                              <VirtualSearchBox />
                              <Hits hitComponent={Hit} />
                            </InstantSearch>  
                          </div>
                        </div>
      }
      else {
        instantSearch = <div className="entry-content">
                          <h2>You haven't entered a search term.</h2>
                          <p>Carry out a search by clicking the <em>magnifying glass</em> in the navigation.</p>
                        </div>
      }

      return (
        <Layout>
          <header className="page-header">
            <h1>Search</h1>
            <p>Search the knowledge-base...</p>
          </header>
          <div id="primary" className="content-area">
            <div id="content" className="site-content" role="main">
                <div className="layout-fixed">
                    <article className="page hentry">
                      {instantSearch}
                    </article>
                </div>
            </div>
          </div>
      </Layout>
    )
  }
}

function Hit(props) {
  return (
    <article className="hentry post">
      <h3 className="entry-title">
        <Link to={props.hit.fields.slug}>
          <Highlight attribute="title" hit={props.hit} tagName="mark" />
        </Link>
      </h3>
      <div className="entry-meta">
        <span className="read-time">{props.hit.fields.readingTime.text}</span>
      </div>
      <p className="entry-content">
        <Highlight hit={props.hit} attribute="summary" tagName="mark" />
      </p>
    </article>
  );
}

Hit.propTypes = {
  hit: PropTypes.object.isRequired,
};

export default SearchPage

My code is reading from a query-string value and passing that to a "searchState". The searchState object is created by React InstantSearch internally. Every widget inside the library has its own way of updating it. It contains parameters on the type of search that should be performed, such as query, sorting and pagination, to name a few. All we're interested in doing is updating the query parameter of this object with our search term.

If the query parameter from the "searchState" object is empty, render search results, otherwise, display a message stating a search term is required.

One thing to notice is the SearchBox has been replaced with a VirtualSearchBox, which uses the connector of the search box to create a virtual widget - in our case an empty span tag. This will link the InstantSearch component with the query. Having some form of search box component is compulsory.

Conclusion

I prefer not to use the out-of-the-box search box component as I can potentially save requests to Algolia's API, as searches aren't being made on the fly as a user enters a search term. This is the plugins default behaviour.

Passing a search term through a query-string may come across as a little backwards, especially when it's rather nice to see search results change before your eyes as you type letter-by-letter. However, this approach misses one key element: Tracking in Google Analytics. Even though I will be primary the person making the most use of my site search, it'll be interesting to see who else uses it and what search keywords are used.

Useful Links

Lazyload and Responsively Serve External Images In GatsbyJs

For the Gatsby version of my website, currently in development, I am serving all my images from Imagekit.io - a global image CDN. The reasons for doing this is so I will have the ultimate flexibility in how images are used within my site, which didn’t necessarily fit with what Gatsby has to offer especially when it came to how I wanted to position images within blog post content served from markdown files.

As I understand it, Gatsby Image has two methods of responsively resizing images:

  1. Fixed: Images that have a fixed width and height.
  2. Fluid: Images that stretch across a fluid container.

In my blog posts, I like to align my images (just take look at my post about my time in the Maldives) as it helps break the post up a bit. I won’t be able to achieve that look by the options provided in Gatsby. It’ll look all a little bit too stacked. The only option is to serve my images from Imagekit.io, which in the grand scheme isn’t a bad idea. I get the benefit of being able to transform images on the fly, optimisation (that can be customised through Imagekit.io dashboard) and fast delivery through its content-delivery network.

To meet my image requirements, I decided to develop a custom responsive image component that will perform the following:

  • Lazyload image when visible in viewport.
  • Ability to parse an array “srcset" sizes.
  • Set a default image width.
  • Render the image on page load in low resolution.

React Visibility Sensor

The component requires the use of "react-visibility-sensor” plugin to mimic the lazy loading functionality. The plugin notifies you when a component enters and exits the viewport. In our case, we only want the sensor to run once an image enters the viewport. By default, the sensor is always fired every time a block enters and exits the viewport, causing our image to constantly alternate between the small and large versions - something we don't want.

Thanks for a useful post by Mark Oskon, he provided a solution that extends upon the react-visibility-sensor plugin and allows us to turn off the sensor after the first reveal. I ported the code from Mark's solution in a newly created component housed in "/core/visibility-sensor.js", which I then reference into my LazyloadImage component:

import React, { Component } from "react";
import PropTypes from "prop-types";
import VSensor from "react-visibility-sensor";

class VisibilitySensor extends Component {
  state = {
    active: true
  };

  render() {
    const { active } = this.state;
    const { once, children, ...theRest } = this.props;
    return (
      <VSensor
        active={active}
        onChange={isVisible =>
          once &&
          isVisible &&
          this.setState({ active: false })
        }
        {...theRest}
      >
        {({ isVisible }) => children({ isVisible })}
      </VSensor>
    );
  }
}

VisibilitySensor.propTypes = {
  once: PropTypes.bool,
  children: PropTypes.func.isRequired
};

VisibilitySensor.defaultProps = {
  once: false
};

export default VisibilitySensor;

LazyloadImage Component

import PropTypes from "prop-types";
import React, { Component } from "react";
import VisibilitySensor from "../core/visibility-sensor"

class LazyloadImage extends Component {
    render() {
      let srcSetAttributeValue = "";
      let sanitiseImageSrc = this.props.src.replace(" ", "%20");

      // Iterate through the array of values from the "srcsetSizes" array property.
      if (this.props.srcsetSizes !== undefined && this.props.srcsetSizes.length > 0) {
        for (let i = 0; i < this.props.srcsetSizes.length; i++) {
          srcSetAttributeValue += `${sanitiseImageSrc}?tr=w-${this.props.srcsetSizes[i].imageWidth} ${this.props.srcsetSizes[i].viewPortWidth}w`;

          if (this.props.srcsetSizes.length - 1 !== i) {
            srcSetAttributeValue += ", ";
          }
        }
      }

      return (
          <VisibilitySensor key={sanitiseImageSrc} delayedCall={true} partialVisibility={true} once>
            {({isVisible}) =>
            <>
              {isVisible ? 
                <img src={`${sanitiseImageSrc}?tr=w-${this.props.widthPx}`} 
                      alt={this.props.alt}
                      sizes={this.props.sizes}
                      srcSet={srcSetAttributeValue} /> : 
                <img src={`${sanitiseImageSrc}?tr=w-${this.props.defaultWidthPx}`} 
                      alt={this.props.alt} />}
              </>
            }
          </VisibilitySensor>
      )
    }
}

LazyloadImage.propTypes = {
  alt: PropTypes.string,
  defaultWidthPx: PropTypes.number,
  sizes: PropTypes.string,
  src: PropTypes.string,
  srcsetSizes: PropTypes.arrayOf(
    PropTypes.shape({
      imageWidth: PropTypes.number,
      viewPortWidth: PropTypes.number
    })
  ),
  widthPx: PropTypes.number
}

LazyloadImage.defaultProps = {
  alt: ``,
  defaultWidthPx: 50,
  sizes: `50vw`,
  src: ``,
  widthPx: 50
}

export default LazyloadImage

Component In Use

The example below shows the LazyloadImage component used to serve a logo that will serve a different sized image with the following widths - 400, 300 and 200.

<LazyloadImage 
                src="https://ik.imagekit.io/surinderbhomra/Pages/logo-me.jpg" 
                widthPx={400} 
                srcsetSizes={[{ imageWidth: 400, viewPortWidth: 992 }, { imageWidth: 300, viewPortWidth: 768 }, { imageWidth: 200, viewPortWidth: 500 }]}
                alt="Surinder Logo" />

Useful Links

https://alligator.io/react/components-viewport-react-visibility-sensor/ https://imagekit.io/blog/lazy-loading-images-complete-guide/ https://www.sitepoint.com/how-to-build-responsive-images-with-srcset/

"Invalid response for blob" Error When Making XHR Request

Whilst making a request to one of my API endpoints for an iOS application, I came across a very unhelpful error: "Invalid response for blob". I couldn't really understand why React Native was complaining about this single API endpoint, since all my other endpoints did not encounter this error.

React Native: Invalid Response For Blob

The API endpoint in question is a pretty simple email address validator. If the users email address is unique and passes all verification checks, the endpoint will either return a 200 (OK) or 400 (Bad Request) along with a response containing the error. For those who understand ASP.NET Web API development, my endpoint is as follows:

/// <summary>
/// Check if a user's email address is not already in user along with other string validation checks.
/// </summary>
/// <param name="email"></param>
/// <returns></returns>
[HttpPost]
[AllowAnonymous]
[Route("EmailAddressValidator")]
public HttpResponseMessage EmailAddressValidator(string email)
{
    if (UserLogic.IsEmailValid(email, out string error)) // UserLogic.IsEmailValid() method carries out email checks...
        return Request.CreateResponse(HttpStatusCode.OK);

    return Request.CreateResponse(HttpStatusCode.BadRequest, new ErrorModel { Error = error });
}

So pretty simple stuff!

Weirdly enough the "Invalid response for blob" issue did not occur within my endpoint when the users email address did not pass the required checks, thus returning a 400 error and a response detailing the error. It was only when a 200 response was returned without a value.

There seems to be a bug in the React Native environment when it comes to dealing with empty API responses. The only way I could get around this, is to always ensure all my API endpoints returned some form of response. Very strange! I suggest all you fellow React Native developers do the same until a fix is put in place.

The issue has already been logged so I will be keeping an eye on a release for a fix.

Update (16/07/2018)

I wasn't too sure to whether anything had been done in regards to the fix since writing this post as there was no update to the Github issue that was first logged on the 5th March. So I decided to share this very post to a React Native group on Reddit to get some form of answer. Within a short period of time, I was told this issue has been fixed in React Native version 0.56.

Is ReactJS A Worthwhile Addition To An ASP.NET MVC Application?

ReactJSI've been meddling around with ReactJS over the last week or so, seeing if this is something viable to use for future client projects. I am always constantly on the lookout to whether there are better alternatives on how my fellow developers and I develop our sites.

Throughout the sample applications I've been building, I constantly asked myself one question: Why Would I Use ReactJS In My Day To Day Development? I am ASP.NET developer who build websites either using Web Forms or MVC Razor. So I am finding it difficult to comprehend whether using ReactJS is viable in these frameworks, especially MVC.

ReactJS is primarily a view framework where you have the ability to write component-based web applications directly into your JavaScript that then gets output to the DOM virtually - making for a very fast and responsive webpage. It's a different approach to developing websites that I quite like. But for the moment, I just don't see how it can benefit me when the full MVC framework does a pretty good job with all the bells and whistles.

For example, I segregate all my key HTML markup into partial views in order to increase re-use throughout my web application, which works really well when making AJAX calls where the markup needs to be displayed on the page asynchronously as well as server-side. I can just see by implementing ReactJS, I will be duplicating this process at JavaScript and CSHTML level if a markup change ever needed to be made. If partial views does the job effectively, I'm not too sure the need for ReactJS in my future ASP.NET MVC creations.

Don't get me wrong - I really like ReactJS. It makes writing JavaScript an even more enjoyable experience purely due to the JSX syntax. Long gone are the days where you have to concatenate strings to form HTML. More importantly, it's readable and truly scalable.

Unfortunately, it doesn't look like ReactJS is a viable option for me at this moment in time. I can see how it would be a very useful framework for building web applications where there is a requirement for the view to be created strictly client-side along with heavy use of AJAX calls from an API layer to serve data to your application. But in situations where you have two frameworks that provide the ability to create views, in this case ReactJS and ASP.NET MVC, it doesn't make sense.

I could be talking absolute nonsense and missing the whole point. If this is the case (most likely!), please leave a comment.