Merge pull request #6 from StephixOne/topic

Return likely predicted topics for posts.

Author: StephixOne

package.json

@@ -19,7 +19,7 @@
     "nodemon": "1.18.9",
     "npm-run-all": "4.1.5",
     "ts-node": "4.1.0",
-    "typescript": "2.9.2"
+    "typescript": "3.3"
   },
   "dependencies": {
     "apollo-errors": "1.9.0",

src/resolvers/index.ts

@@ -1,6 +1,6 @@
 import * as scuid from 'scuid';
 
-import { generateAuthToken, getUserId } from '../utils';
+import { generateAuthToken, getUserId, likelyTopics } from '../utils';
 
 const DEFAULT_COUNT = 25;
 
@@ -66,37 +66,57 @@ export default {
     },
 
     // refactor user relation into Class
-    Post: (parent, { id }, { faker }) => ({
-      id,
-      title: faker.random.words(),
-      body: faker.lorem.paragraphs(),
-      published: faker.random.boolean(),
-      createdAt: faker.date.past(),
-      author: {
-        id: scuid(),
-        firstName: faker.name.firstName(),
-        lastName: faker.name.lastName(),
-        email: faker.internet.email(),
-        avatar: faker.image.avatar()
-      }
-    }),
-
-    // refactor user relation into Class
-    allPosts: (parent, { count }, { faker }) => {
-      return new Array(count).fill(0).map(_ => ({
-        id: scuid(),
-        title: faker.random.words(),
-        body: faker.lorem.sentences(),
+    Post: (parent, { id }, { faker }) => {
+      const title = faker.random.words();
+      const body = faker.lorem.paragraphs();
+      const firstName = faker.name.firstName();
+      const lastName = faker.name.lastName();
+      return {
+        id,
+        title,
+        body,
         published: faker.random.boolean(),
         createdAt: faker.date.past(),
         author: {
           id: scuid(),
-          firstName: faker.name.firstName(),
-          lastName: faker.name.lastName(),
+          firstName,
+          lastName,
           email: faker.internet.email(),
           avatar: faker.image.avatar()
+        },
+        likelyTopics: likelyTopics(
+          `${firstName} ${lastName}`,
+          title, body
+        )
+      };
+    },
+
+    // refactor user relation into Class
+    allPosts: (parent, { count }, { faker }) => {
+      return new Array(count).fill(0).map(_ => {
+        const title = faker.random.words();
+        const body = faker.lorem.paragraphs();
+        const firstName = faker.name.firstName();
+        const lastName = faker.name.lastName();
+        return {
+          id: scuid(),
+          title,
+          body,
+          published: faker.random.boolean(),
+          createdAt: faker.date.past(),
+          author: {
+            id: scuid(),
+            firstName,
+            lastName,
+            email: faker.internet.email(),
+            avatar: faker.image.avatar()
+          },
+          likelyTopics: likelyTopics(
+            `${firstName} ${lastName}`,
+            title, body
+          )
         }
-      }));
+      });
     }
   },
 

src/schema.graphql

@@ -22,13 +22,19 @@ type Todo {
   completed: Boolean!
 }
 
+type Topic {
+  label: String!
+  likelihood: Float
+}
+
 type Post {
   id: ID!
   title: String!
   body: String!
   published: Boolean!
   createdAt: String!
   author: User!
+  likelyTopics: [Topic]
 }
 
 type Query {

src/utils.ts

@@ -16,3 +16,49 @@ export const getUserId = ctx => {
 
   throw new Error('Not Authorized');
 };
+
+// Ten randomly-chosen but consistent topic nouns.
+export const topicWords = [
+  'wedding',
+  'sport',
+  'fishing',
+  'shopping',
+  'security',
+  'management',
+  'community',
+  'celebrity',
+  'birthday',
+  'potato'
+];
+
+export const likelyTopics = (user, title, body) => {
+  // create a custom order over the topics, based on letters as they appear in
+  // the user's name, post title, and post body, for variability
+  const alphabet = `${title} ${user} ${body}`;
+  const topicLabels = topicWords.sort((a, b) => {
+    for (let i = 0; i < Math.min(a.length, b.length); i++) {
+      const aValue = alphabet.indexOf(a[i]);
+      const bValue = alphabet.indexOf(b[i]);
+      if (aValue !== bValue) {
+        return bValue - aValue;
+      }
+    }
+    return b.length - a.length;
+  })
+  // compute a biased probability distribution
+  const likelihoods = [];
+  let total = 0;
+  for (let i = 0; i < topicLabels.length; i++) {
+    const weight = Math.random() * Math.log(i + 2);
+    total += weight;
+    likelihoods.push(weight);
+  }
+  likelihoods.reverse();
+  // and associate each likelihood with a probability
+  return likelihoods.map((weight, i) => {
+    return {
+      likelihood: weight / total,
+      label: topicLabels[i]
+    }
+  }).sort((a, b) => b.likelihood - a.likelihood);
+};