Sådan bruges API-første platforme til at opbygge dine websteder hurtigere

Sådan bruges API-første platforme til at opbygge dine websteder hurtigere

Værktøjer som Jekyll, Hugo eller Hexo har populariseret statiske websteder i de senere år. Den såkaldte JAMstack giver dig mulighed for at levere meget dynamisk indhold uden noget back-end-lag overhovedet. Derudover gjorde udvikler-første API'er det muligt for front-end-udviklere at opbygge endnu mere kompleks funktionalitet. Dette kan de gøre uden at forlade browsersandkassen. Lad os se, hvordan du kan udnytte moderne API-første platforme til at sende en solid prototype af en forretningsapplikation. Den tilgang, der præsenteres i denne artikel, kan blive et nyttigt aktiv i din løsningsarkitters værktøjskasse.

Vejledningen består af 2 dele:

  • Den første viser dig, hvordan du designer applikationen for at få en såkaldt happy path. Vi bygger en semi-automatiseret prototype, som du kan bruge til at få brugerfeedback på en demo-session
  • Den anden forklarer, hvordan man styrker automatisering af forretningsprocesser, så appen kan håndtere tidligt trafik i produktionen

Hvad er en API-første platform?

Som Ed Shelley fra ChartMogul beskriver det, er der et par ret svære at gå glip af funktioner i en sådan tjeneste:

  • Der er INGEN brugergrænseflade (GUI). Eller i nogle tilfælde er der en GUI, men den er sekundær i forhold til kerneproduktet.
  • Interaktion med tjenesten sker via en webbaseret API. Dette er en programmatisk måde at forbinde tjenester og overføre data over internettet på en maskinlæsbar måde.
  • Tjenestens værdi er normalt i de data, der leveres (via API).
  • Prissætning er ofte brugsbaseret, hvilket betyder, at prisen er baseret på antallet af anmodninger til API'en.

Dybest set er det, de tilbyder, et sæt byggesten, normalt i SaaS-modellen. Disse kan du bruge til at konstruere en bestemt funktionalitet med mindre kode. En af de første og sandsynligvis mest bemærkelsesværdige repræsentanter for dette er Stripe.Stripe hjælper med at behandle betalinger. Imidlertid har du måske hørt om andre store fisk for nylig, der er kommet ud af markedet, som Twilio eller Algolia.

Hvorfor bruge en API-første platform?

Lad os starte med en lille ansvarsfraskrivelse. Denne tutorial beskriver, hvordan man udvikler applikationer uden serverside overhovedet. Vi mener dog, at det ikke er en pragmatisk tilgang til softwarearkitektur.

Vi ønsker snarere at fremhæve nogle dele af dit back-end-maskiner, som du ikke har brug for at implementere fra bunden. Dette gælder især når forretningskravene til en bestemt funktion ikke er sat i sten, og dit mål er faktisk at finde ud af dem. Med andre ord at finde ud af, om funktionaliteten får et positivt svar fra brugerne og til sidst har en plads i dit produkt.

På samme tid ønsker du ikke at låse dit produkt sammen med en leverandør, der leverer en out-of-the-box-løsning. Dette skyldes, at du ved, at det før eller senere vil føre til et ”helvedes helvede”. Og som du har lært, er det svært at komme tilbage derfra.

For at give dig et eksempel, forestil dig at din virksomhed ønsker at oprette en blog. Desuden har de allerede erklæret, at de ønsker at udvide det og tjene penge på det. Der er to implicitte krav, du skal tage i betragtning, før du kommer med en tech stack i et sådant scenario:

  • Du vil hurtigt sende blogfunktionaliteten - virksomheder kan ikke vente i evigheder på en simpel blog.
  • Du vil ikke ende med at jonglere Wordpress-plugins.

Den type værktøjer, vi vil præsentere, kan være svaret. De giver dig nogle funktionelle byggesten, og din eneste opgave er at tilpasse dem til din virksomhed.

Du bliver glad, fordi du har fuld kontrol over din kodebase. Derudover er ledelsen også glad, fordi de får værdi fra dag 1.Plus, de behøver ikke betale forud!

Så lad os nu vise dig, hvordan disse værktøjer kan spare uges tekniske tid, mens du stadig holder din kodebase åben for ændringer.

Bemærk: De værktøjer, vi skal bruge, fungerer også i serversiden. De tilbyder faktisk mere funktionalitet, når de er forbundet ved hjælp af sikre API-nøgler. Så vi synes, det er mere pragmatisk at have det integreret på serversiden. Ikke desto mindre vil vi som eksperiment kun bruge klientsidesfunktionaliteten. Derudover bruger vi noget Zapier-lim til hurtigt at automatisere forskellige forretningsprocesser.

nostalgi.io

Vi bygger en markedsplads for de ældre webteknologikonsulenter - nostalgia.io. Hvis du med en chance søger hjælp til et gammelt system baseret på Struts eller Google Web Toolkit, er dette stedet at gå. I den første del af denne vejledning lærer vi, hvordan man udnytter flere API-første platforme til at sende følgende funktionalitet:

  • Gennemse ældre teknologier
  • Fuldtekstsøgning og filtrering af eksperter
  • Reservation af møder med eksperter
  • Rabat med kuponer

Den tekniske stak består af:

  • Tilfreds - som en database for teknologier og eksperter
  • Algolia - til søgning i fuld tekst
  • Timekit - for tilgængelighedskontrol og reservation
  • Typeform - til formularer
  • Voucherify - til kuponadministration (ansvarsfraskrivelse: dette er vores produkt)

Bemærk: Vi dækker ikke selve godkendelse og betalingsbehandling. Du kan prøve at implementere dem selv som hjemmearbejde (tip: auth0 og Stripe kan være nyttige).

Lad os hoppe ind i koden.

Bemærk 1: for kortfattethed beskriver vi ikke en detaljeret trin-for-trin guide. Du bør slå de manglende dele i specifikationerne op - heldigvis har API-første udbydere tendens til at have udviklervenlige dokumenter, omfattende API-reference og snesevis af nyttige vejledninger.

Bemærk 2: der er mange måder, du kan være vært for dit statiske websted. Vi bruger glitch-udviklingsplatform, så du nemt kan remixe den og lege med den selv.

Note 3: Vi er ligeglad med applikationens udseende og følelse for ikke at skjule integrationsdelen, plus det på en eller anden måde passer til forretningstemaet, ikke? :)

Datamodel - tilfreds

Normalt starter applikationens design med en dataforholdsmodel. Dette bør også være den første bekymring for os. Men lad os springe over databaseleverandørens diskussion et øjeblik og springe direkte ind i modeller. Hvordan det?

Mød Contentful - et hovedløst CMS. Ved hjælp af en skarp forenkling kan du tænke på det som en Wordpress uden front-end.

Det muliggør:

  • udviklere til at levere indholdet justeret til mediet, det være sig website, mobilapp eller VR-enhed - dette gøres via RESTful API
  • marketingfolk til at oprette, administrere og udgive indhold uden at skulle beskæftige sig med formateringen - med støtte fra indholdsmodelleringsdashboardet og rich-text editor

Vi bruger Contentful til at oprette 2 basale enheder - teknologi og ekspert. En ekspert kender en eller flere teknologier. Lad os se, hvor let det er at oprette sådanne enheder, tilføje nogle rigtige objekter og vise dem på en statisk side.

Teknologibrowser

Med Contentfuls modeladministrator er det lige så let at designe en enhed som at trække nye felter ind i dataindholdsmodeladministratoren. Der er 8 forskellige typer. Disse inkluderer standard, som en streng eller et tal. Der som også nogle specifikke typer, som Location eller Media, som kommer med nyttige egenskaber.

Opret en gratis konto. Følg derefter on-boarding guide for at skabe et rum.

Til sidst skal du oprette din første model, svarende til hvad du kan se i skærmbilledet nedenfor:

Now that you have the Technology model, go to the Content tab to create a couple of instances. As you can see, Contentful provides an intuitive editor for data entry. It takes care of data validation, localization, publishing status, version control, and much more. It’s first and foremost a developer-first platform. Yet these features satisfy marketers and content managers too.

Enough clicking, let’s get to coding. The first task is to display the technologies we’ve just created. To do so, we’ll use Contentful JavaScript SDK.

It makes fetching technologies easy and comes down to 3 steps:

  • Create a new glitch website project, load the contentful.js script, and initialize it with the credentials you can find in the API section.

Note: there are 2 types of keys available in Contentful. One is for content management and one is for content delivery.

The first type can be used to create, update, or delete new models or their instances programmatically.

The second gives you a way to deliver your content to your website or app.

This distinction has been made for security reasons. You don’t want to publish your content management keys on your website, do you? The same applies to the other API-first platforms we’re using in this tutorial.

  • Call getEntries method. This loads the content according to your query parameters. In our case, we want to load only the “Technology” entities. Build some front-end on top of the data . What you get from Contentful is pure JSON (example). Now you can display it to your users as you want. That’s one of the biggest pros when you want or have to adapt your content to multiple devices.

Take a look at this gist:

const client = contentful.createClient({ space: SPACE_ID, accessToken: ACCESS_TOKEN }) const techCards = document.querySelector('#cards'); function fetchTechnologies () { return client.getEntries({ content_type: "technology" }) .then((response) => response.items) .catch((error) => { console.log(`\nError occurred while fetching Entries for Technology:`) console.error(error) }) } fetchTechnologies().then((technologies) => { techCards.innerHTML += technologies.map(technology => ` 

${technology.fields.name}

${technology.fields.description}

`).join('') })

Short and sweet, isn’t it? You can see the overall effect here.

Adding Experts and the Search

So, now we want to display the list of experts when somebody chooses a particular tech. That should be similar to what we’ve just done with Technology a second ago. But let’s make it a bit more advanced. What if we want to make experts searchable? Think of full-text search in their profiles and also a price filter.

Certainly, you can build it on top of Contentful. For example, add another entity, configure search mechanics and UI with getEntries, but there’s a faster way. And by saying faster I mean in both implementation time and the speed of loading search results.

We’ll use another API provider — Algolia. Their platform makes it easy to build and maintain super-fast full-text search. They take care of typo-tolerance, synonyms, geosearch and other little issues. These issues you would most likely bump into when your search feature goes to production.

How does it work? You just use a RESTful API to feed their engine with the data. Then, you configure what attributes should be searchable and how the results should be ranked. Finally, using their JavaScript SDK, you can deliver the instant search experience to any website. Let’s make our experts searchable now!

We’ll start off by creating a data model in Contentful and creating a relationship with Technology entity. Then, we’ll build an Algolia index and add our entities (JSON format) to it.

Add another content model with the fields you can see below:

Notice that we’ve made a one-to-many relationships using the Reference field type. We just want to reflect that any expert might know more than one tech. Once ready, add some experts and assign them to their technologies manually. Use multiple technologies for one of the experts.

You should end up having a similar list:

And the JSON structure looks like:

 { "sys":{ "space":{ "sys":{ "type":"Link", "linkType":"Space", "id":"n763nxcwuf4y" } }, "id":"1mn1mwlwAcQWqgQamsIEmW", "type":"Entry", "createdAt":"2017-12-05T11:29:35.202Z", "updatedAt":"2017-12-13T10:04:52.381Z", "revision":7, "contentType":{ "sys":{ "type":"Link", "linkType":"ContentType", "id":"expert" } }, "locale":"en-US" }, "fields":{ "name":"Javier Hernandez", "technologies":[ { "sys":{ "type":"Link", "linkType":"Entry", "id":"5oKmKwfdjGO2cCaCkwamKW" } }, { "sys":{ "type":"Link", "linkType":"Entry", "id":"7Dtej0GnXqw6cSIMmA6Cko" } } ], "image":{ "sys":{ "type":"Link", "linkType":"Asset", "id":"4RZoQOCwvCMEWMMCuqA0ey" } }, "description":"Lorem ipsum dolor sit amet, consectetur adipiscing elit, sed do eiusmod tempor incididunt ut labore et dolore magna aliqua. Sed faucibus turpis in eu mi bibendum. Mauris in aliquam sem fringilla ut. Tincidunt nunc pulvinar sapien et ligula. ", "projects":53, "price":40, "city":{ "lon":2.2247314453125, "lat":41.36933709640475 } } }

Let’s upload our experts to Algolia. Sign up for a free account, go to the Indices section and run NEW INDEX.

Now we need to transfer our entities from Contentful to Algolia. We could’ve used a dedicated migrator. This is a fantastic tool that automatically loads your content. It then removes, in this case redundant, the Contentful system information (see the gist above) from effective JSONs. It can also resolve relationships. For example instead of IDs, you’ll send the actual names when it comes to the “technologies” field. Finally it syncs with the Algolia index.

But we’ll deliberately do it manually. We need a small improvement in the way we build our index. Therefore, syncing one-to-one with the migrator isn’t an option in our case.

When we use search input in a technology site, naturally we want to include only the experts of chosen technology in the search results. As you can see in the Expert JSON example, technologies are represented as an array of objects. The problem is that you can’t build a facet which filters the data based on a nested array of objects with Algolia.

What they suggest is to split the expert object into as many sub-objects as the number of technologies. So, in the case of Javier Hernandez, who knows 2 frameworks, we should add 2 objects:

{ name: "Javier Hernandez", technologies: { name: “Google Web Toolkit” … // other properties } … // other properties } { name: "Javier Hernandez" technologies: { name: “Apache Struts 1” … // other properties } … // other properties }

As an exercise, you can create a script that splits experts and adds them to the index through the Algolia API. You’ll need Algolia server-side authentication keys. Here’s a snippet which handles the split logic. Notice that the script also purges Contentful’s system info.

This makes the objects lighter will make the search faster:

client.getEntries({ content_type: "expert" }) .then((response) => { const denormalized = [].concat(...response.items.map(item => { let arr = [] item.fields.contentfulID = item.sys.id delete item.sys item.fields.technologies.forEach(tech => { const i = JSON.parse(JSON.stringify(item)) i.fields.technologies = tech.fields i.fields.image = item.fields.image.fields arr.push(i.fields) }) return arr })) console.log(JSON.stringify(denormalized)) }) .catch((error) => { console.log(`\nError occurred while fetching Entries for Expert:`) console.error(error) })

As we have 6 experts and 2 of them know 2 technologies, we should end up with 8 objects in the index. As an alternative to the API insert method, you can upload them with the dashboard. Once uploaded, you can try to use the search in the Dashboard to see how fast Algolia filters the data.

Now, we’re almost ready to connect our search to Algolia. Almost — because we need to create a facet which will allow us to filter results by technology and price. Go to DISPLAY and select technology.name and price in “Attributes for faceting”, then Save.

Finally, we can connect our search to our index so that it retrieves and displays the results. Algolia comes with an advanced JavaScript library which makes it easy as pie.

Take a look at this code:

const isConfig = { appId: 'N675AF3ESI', apiKey: '14b65c352deb9a505131d3d00cba2f6c', indexName: 'experts', urlSync: false } isConfig.searchParameters = { filters: `technologies.name:"${selectedTechnology}"` } const search = instantsearch(isConfig) search.addWidget( instantsearch.widgets.searchBox({ container: '#search-input' }) ) search.addWidget( instantsearch.widgets.hits({ container: '#hits', hitsPerPage: 10, templates: { item: document.getElementById('hit-template').innerHTML, empty: "We didn't find any results for the search \"{{query}}\"" }, cssClasses: { root: 'row', item: 'col-md-4' } }) ) search.addWidget( instantsearch.widgets.rangeSlider({ container: '#price-refinement', attributeName: 'price', tooltips: { format: function(rawValue) { return '$' + Math.round(rawValue).toLocaleString(); } } }) ) search.start()

Notice how we configure the search to use the technology filter in lines 8–10. See how easy it is to adjust the result page to a respective container — line 28 (though it’s hard to find in the docs).

Overall, with about 2-dozen lines and you get this:

So far, we’ve built a simple expert browser supporting full-text search and price sliders. Adding new experts is troublesome at this stage because you have to first manually create them in Contentful and then sync it with Algolia. We’ll automate this in the second part.

The good news is that you can already use this prototype to get some early feedback for the technology browsing and experts filtering. The next step is to create the expert profile page and enable booking.

The search demo code can be found in experts.html.

Bookings

As you might’ve guessed, we won’t be implementing calendar functionality from scratch either. We’ll use Timekit. They offer the API + dashboard to manage calendars and bookings for people and resources. Think of it as a Google/Outlook calendar engine exposed with REST API.

The process of making experts bookable with Timekit is as follows:

  • Create a Resource entity and an assigned Calendar entity
  • Store the resource and calendar IDs in a corresponding expert entity in Contentful
  • Use Timekit JS SDK to display the calendar on an expert’s profile page

And that’s it, you’ve just got bookings up-and-running! Don’t believe me? Read on:

  • Create an account and start a free trial (there’s no free version).
  • Create a Project in which you’ll define the basic calendar mechanics . For instance, event duration, minimum notice, and reminders.
  • Define if the booking requests should be accepted automatically or must be confirmed manually.

For each expert, create a resource and within this resource, create a calendar. Notice that one resource might have more than one calendar.

This is a nice feature to keep in mind when we plan some upgrades in Nostalgia’s business model.

Now, we have to store the resource’s email, newly created calendar ID, and the client-side API key in the corresponding expert entity in Contentful.

You can edit the expert content model and add a JSON field named timekit. Then, modify the expert entities to add timekit details.

The last step is to display the actual calendar in the expert’s profile page. You already know the process. Include an SDK script and configure it properly to render the widget.

But this time we need to load 2 libraries:

  • Contentful — to load customers details, including Timekit credentials
  • Timekit — to place the calendar assigned to a given expert

Here’s the code:

const widget = new TimekitBooking() const client = contentful.createClient({ space: SPACE_ID, accessToken: ACCESS_TOKEN }) client.getEntries({'sys.id': expertId}).then((response) => { const e = response.items[0].fields expertWidget.innerHTML= ` 

${e.name}

$${e.price}

projects: ${e.projects}

${e.description}

` const timekitConf = e.timekit widget.init({ targetEl: '#bookingjs', app: 'nostalgia-4592', apiToken: timekitConf.apiToken, email: timekitConf.email, calendar: timekitConf.calendar, name: 'Jane Doe', timekitFindTime: { length: '3 hours', start: 'tomorrow', filters: { and: [ { specific_time: { start: '8', end: '17' }} ] } }, fullCalendar: { defaultView: 'month' } }) }) .catch((error) => { console.log(`\nError occurred while fetching Entries for Expert:`) console.error(error) })

Notice how we can adjust the booking details such as time slots (line 39). Timekit offers even more customization capabilities, so make sure to read the booking.js spec.

The effect just blows us away. Twenty lines of code and we have our booking widget in place. Timekit oversees the whole process for you. It helps with resolving conflicts and sends email confirmations to experts and customers.

The most important thing is that this approach is highly flexible. It’s all in code. Every single piece of this mechanism can be adjusted through the API.

For example, let’s suppose we want to review a booking request before accepting. It just so happens that Timekit makes it possible with a single flag. Such options are the real power of API-first solutions. Make sure you read the tutorials and docs to learn all the features.

Coupons

Nostalgia isn’t a well-known business yet. We need to find some way to attract early adopters. One of the oldest and most successful methods is discounts. A discount might be applied either after redeeming a coupon or because of the volume of products in the cart. To implement both cases, you might want to use Voucherify.

Why Voucherify? There are a few basic things you should get right when you want to handle coupons properly to save yourself tons of engineering time:

  • Uniqueness of coupon codes — To reduce fraud and get precise tracking of your promotional campaigns
  • Extensible coupon validation mechanism — This is a generic approach which enables adding/removing/expiring multiple coupon codes
  • Easy monitoring of redemption — This will answer marketing and customer service department questions off the bat

You can take care of these 3 things yourself. However, you can get the same result with a couple of lines using Voucherify API endpoints. By doing so, you can immediately forget about coupon misuse, maintaining the “if” ladder validating whether the code is active and valid. You can also forgo providing marketing teams with coupon campaigns results. Nor will you be drilling down the logs to understand why a customer’s redemption failed.

Let’s create a bulk 1000 coupons. These we’ll send to our early adopters. Finally let’s give the customers the possibility to actually use them in our website to enjoy discounted prices.

Sign up for a Voucherify account and go to the campaign manager to create the first batch of coupon codes. Let’s say each coupon carries 25% off.

In the manager, you can specify the discount details and other business limits. For example, specify the expiration date, max total amount, or a specific customer segment eligible for the discount.

When the manager is done, you can start distributing coupons through various channels. Voucherify offers email, sms, push notification, intercom or braze out of the box. But there’s a lot of other ways available thanks to REST API and webhooks.

Before you send them out, you should give customers an option to redeem them. This can be achieved by using the redemption endpoint from the API. Yet, you can also use the pre-built widget from voucherify.js.

Voucherify allows you to either validate or redeem the coupon.

The validation checks if:

  • the coupon comes from your Voucherify account
  • is not expired or disabled
  • it matches all business rules

The redemption does the validation first and subsequentlymarks the coupon as used. In this part, we’ll wire in validation only to show customers a discounted price. In the second article, we’ll send a redemption request when the booking is confirmed.

Include voucherify.js snippet and optionally the corresponding CSS file for a better look and feel. Then, put in the following code:

Voucherify.initialize( "4dde7477-d8d1-4057-8f91-8a9e7137acee", "404c6c0b-4445-4f14-84b1-f4a58f1da2f6" ) Voucherify.render("#voucher-widget", { textPlaceholder: "Your coupon...", onValidated: function(response) { if (response) { const priceTag = document.querySelector('#priceTag') priceTag.innerHTML = Voucherify.utils.calculatePrice(parseInt(priceTag.innerHTML), response) } } })

The library will render a coupon widget which automatically validates the code against the Voucherify API.

You can test it out with the codes we’ve pre-generated with the campaign manager:

* 25% off: nstlg-CCAMIDFf, nstlg-wZK4CoLs, nstlg-V8eV9A3p

* $5 off: uub-nstlg, afl-nstlg, yeq-nstlg

* expired code: VuFF2Wyy

Notice that you can easily customize codes patterns, prefixes and post-fixes can be useful for tracking and reporting.

Now, paste any coupon code into the widget and see the corresponding discount applied:

n the 2nd article, we’ll show you how to monitor the successful and failed coupon redemptions to see if your promo campaign is on track.

Voucherify offers much more than that. Check out the docs and examples to find out how to build advanced promotions and referral programs in days instead of months.

You can find the booking page code here (scheduler.html).

Recap

We planned to build a proof of concept for a new business application — Nostalgia.io. A prototype we can use to pick early users’ brains. Something we can deliver in a decent timeframe yet not a total throwaway.

Hopefully, we’ve convinced you that with developer-first tools like Contentful, Algolia, Timekit, or Voucherify you can achieve it. Even more importantly, you can do it without setting up any back-end layer at all.

It still requires some manual work to keep data in sync among tools. Yet, the flexibility and speed of iteration of these API-first tools at your fingertips definitely makes up for it.

Certainly, these tools aren’t all light and bright. For example, we bumped into these several issues when going through this article:

  • Contentful getEntry()method doesn’t resolve links. We had to use getEntries() instead to get a single expert entity with profile image URL
  • It took us more than a little while to understand how we can display results using column layout (the default is rows)
  • Timekit doesn’t allow for fetching the calendar instance config using an external id. That’s why we need to store calendar tokens in expert entity in Contentful
  • The Voucherify widget doesn’t enable you to try another valid code without refreshing the website

I’m sure there are many more of them. But you can work these little issues around in far less time than you’d spend to build these features from scratch. On top of that, you avoid the serious and time-consuming architectural mistakes the teams at these platforms have made before you.

The source code of the project can be found here. And the demo is live here!

Hardening and scaling

As you can see, some processes are still manual and thus tedious:

  • Adding new experts
  • Making experts searchable
  • Creating calendars for experts

I den næste del limer vi disse tjenester ved hjælp af Zapier. Zapier er en platform, der letter tilslutning af API-første platforme. På denne måde reducerer vi det manuelle arbejde, der er nødvendigt for at køre de ovennævnte forretningsstrømme. Eksempelvis vil eksperter være i stand til at tilmelde sig selv. Derudover opretter platformen alle nødvendige enheder programmatisk.

Endelig vil vi skubbe prototypen til produktion. Det vil stadig være en tidlig applikation, men det vil være mere robust og klar til at betjene rigtige kunder. Bliv hængende!

Opdatering : du kan finde den anden del er her.