From bdc2c0c1da4f2e1970c34dd33b5a5ab6e2c59680 Mon Sep 17 00:00:00 2001
From: Adam Pash <adam.pash@gmail.com>
Date: Fri, 9 Sep 2016 10:25:12 -0400
Subject: [PATCH] feat: can now fetch attrs in RootExtractor's select method

---
 TODO.md                                 |  1 +
 src/extractor/constants.js              |  1 +
 src/extractor/custom/nymag.com/index.js | 13 +++++++
 src/extractor/root-extractor.js         | 16 +++++++--
 src/extractor/root-extractor.test.js    | 45 +++++++++++++++++++++++--
 5 files changed, 71 insertions(+), 5 deletions(-)
 create mode 100644 src/extractor/constants.js

diff --git a/TODO.md b/TODO.md
index 38a71233..f5c7f8bf 100644
--- a/TODO.md
+++ b/TODO.md
@@ -12,6 +12,7 @@ TODO:
 - Separate constants into activity-specific folders (dom, scoring)
 
 DONE:
+x add option to fetch attrs in RootExtractor's select method
 x get custom datePublished selector to convert to date object (prob through cleaner)
 x extract and generalize cleaners
   x move arguments to cleaners to object
diff --git a/src/extractor/constants.js b/src/extractor/constants.js
new file mode 100644
index 00000000..b6fc067b
--- /dev/null
+++ b/src/extractor/constants.js
@@ -0,0 +1 @@
+export const ATTR_RE = /\[([\w-]+)\]/
diff --git a/src/extractor/custom/nymag.com/index.js b/src/extractor/custom/nymag.com/index.js
index 6b709231..d96a4bc6 100644
--- a/src/extractor/custom/nymag.com/index.js
+++ b/src/extractor/custom/nymag.com/index.js
@@ -39,6 +39,19 @@ const NYMagExtractor = {
       'h1.headline-primary',
       'h1',
     ]
+  },
+
+  author: {
+    selectors: [
+      '.by-authors',
+    ]
+  },
+
+  datePublished: {
+    selectors: [
+      'time.article-timestamp[datetime]',
+      'time.article-timestamp',
+    ]
   }
 }
 
diff --git a/src/extractor/root-extractor.js b/src/extractor/root-extractor.js
index c6f53480..5635dfd3 100644
--- a/src/extractor/root-extractor.js
+++ b/src/extractor/root-extractor.js
@@ -3,6 +3,7 @@ import 'babel-polyfill'
 import GenericExtractor from './generic'
 import Cleaners from '../cleaners'
 import { convertNodeTo, stripTags } from './utils/dom'
+import { ATTR_RE } from './constants'
 
 const RootExtractor = {
   extract(extractor=GenericExtractor, opts) {
@@ -44,7 +45,7 @@ function extract(opts) {
     GenericExtractor[type](opts)
 }
 
-function select(opts) {
+export function select(opts) {
   const { $, type, extractionOpts, extractHtml=false } = opts
   // Skip if there's not extraction for this type
   if (!extractionOpts) return
@@ -75,8 +76,17 @@ function select(opts) {
 
     return $.html($content)
   } else {
-    // return stripTags($(matchingSelector).text(), $)
-    return Cleaners[type]($(matchingSelector).text(), opts)
+    // if selector includes an attr (e.g., img[src]),
+    // extract the attr
+    const attr = matchingSelector.match(ATTR_RE)
+    let result
+    if (attr) {
+      result = $(matchingSelector).attr(attr[1])
+    } else {
+      // otherwise use the text of the node
+      result = $(matchingSelector).text()
+    }
+    return Cleaners[type](result, opts)
   }
 }
 
diff --git a/src/extractor/root-extractor.test.js b/src/extractor/root-extractor.test.js
index 1b86ebb7..aad473fe 100644
--- a/src/extractor/root-extractor.test.js
+++ b/src/extractor/root-extractor.test.js
@@ -3,6 +3,7 @@ import fs from 'fs'
 import cheerio from 'cheerio'
 
 import RootExtractor from './root-extractor'
+import { select } from './root-extractor'
 import {
   cleanBySelectors,
   transformElements
@@ -125,11 +126,51 @@ describe('transformElements($content, $, { transforms })', () => {
   })
 })
 
-export function clean(string) {
+describe('select(opts)', () => {
+  it(`returns a node's text with a simple selector`, () => {
+    const html = `
+      <div><div class="author">Bob</div></div>
+    `
+    const $ = cheerio.load(html)
+    const opts = {
+      type: 'author',
+      $,
+      extractionOpts: {
+        selectors: ['.author']
+      }
+    }
+
+    const result = select(opts)
+    assert.equal(result, 'Bob')
+   })
+
+  it(`returns a node's attr with a attr selector`, () => {
+    const html = `
+      <div>
+        <time datetime="2016-09-07T05:07:59-04:00">
+          September 7, 2016
+        </time>
+      </div>
+    `
+    const $ = cheerio.load(html)
+    const opts = {
+      type: 'datePublished',
+      $,
+      extractionOpts: {
+        selectors: ['time[datetime]']
+      }
+    }
+
+    const result = select(opts)
+    assert.equal(result, '2016-09-07T09:07:59.000Z')
+   })
+})
+
+function clean(string) {
   return string.trim().replace(/\r?\n|\r/g, '').replace(/\s+/g, ' ')
 }
 
-export function assertClean(a, b) {
+function assertClean(a, b) {
   assert.equal(clean(a), clean(b))
 }