2023-02-21 19:55:31 -05:00
|
|
|
# frozen_string_literal: true
|
|
|
|
|
|
2021-11-05 18:23:05 -04:00
|
|
|
require 'rails_helper'
|
|
|
|
|
|
|
|
|
|
RSpec.describe LinkDetailsExtractor do
|
2023-08-03 09:41:51 -04:00
|
|
|
subject { described_class.new(original_url, html, nil) }
|
2023-02-19 23:24:14 -05:00
|
|
|
|
2023-08-03 09:41:51 -04:00
|
|
|
let(:original_url) { 'https://example.com/dog.html?tracking=123' }
|
2021-11-05 18:23:05 -04:00
|
|
|
|
|
|
|
|
describe '#canonical_url' do
|
2023-08-03 09:41:51 -04:00
|
|
|
let(:html) { "<!doctype html><link rel='canonical' href='#{url}'>" }
|
2021-11-05 18:23:05 -04:00
|
|
|
|
2023-08-03 09:41:51 -04:00
|
|
|
context 'when canonical URL points to the same host' do
|
|
|
|
|
let(:url) { 'https://example.com/dog.html' }
|
2021-11-05 18:23:05 -04:00
|
|
|
|
|
|
|
|
it 'ignores the canonical URLs' do
|
2023-08-03 09:41:51 -04:00
|
|
|
expect(subject.canonical_url).to eq 'https://example.com/dog.html'
|
2021-11-05 18:23:05 -04:00
|
|
|
end
|
|
|
|
|
end
|
|
|
|
|
|
2023-08-03 09:41:51 -04:00
|
|
|
context 'when canonical URL points to another host' do
|
|
|
|
|
let(:url) { 'https://different.example.net/dog.html' }
|
2021-11-05 18:23:05 -04:00
|
|
|
|
|
|
|
|
it 'ignores the canonical URLs' do
|
2023-08-03 09:41:51 -04:00
|
|
|
expect(subject.canonical_url).to eq original_url
|
2021-11-05 18:23:05 -04:00
|
|
|
end
|
|
|
|
|
end
|
2022-03-25 14:31:35 -04:00
|
|
|
|
|
|
|
|
context 'when canonical URL is set to "null"' do
|
2023-08-03 09:41:51 -04:00
|
|
|
let(:url) { 'null' }
|
2022-03-25 14:31:35 -04:00
|
|
|
|
|
|
|
|
it 'ignores the canonical URLs' do
|
|
|
|
|
expect(subject.canonical_url).to eq original_url
|
|
|
|
|
end
|
|
|
|
|
end
|
2021-11-05 18:23:05 -04:00
|
|
|
end
|
2022-02-07 12:16:31 -05:00
|
|
|
|
2023-08-03 09:41:51 -04:00
|
|
|
context 'when only basic metadata is present' do
|
|
|
|
|
let(:html) { <<~HTML }
|
|
|
|
|
<!doctype html>
|
|
|
|
|
<html lang="en">
|
|
|
|
|
<head>
|
|
|
|
|
<title>Man bites dog</title>
|
|
|
|
|
<meta name="description" content="A dog's tale">
|
|
|
|
|
</head>
|
|
|
|
|
</html>
|
|
|
|
|
HTML
|
2022-02-07 12:16:31 -05:00
|
|
|
|
2023-08-03 09:41:51 -04:00
|
|
|
describe '#title' do
|
|
|
|
|
it 'returns the title from title tag' do
|
|
|
|
|
expect(subject.title).to eq 'Man bites dog'
|
|
|
|
|
end
|
|
|
|
|
end
|
|
|
|
|
|
|
|
|
|
describe '#description' do
|
|
|
|
|
it 'returns the description from meta tag' do
|
|
|
|
|
expect(subject.description).to eq "A dog's tale"
|
|
|
|
|
end
|
|
|
|
|
end
|
|
|
|
|
|
|
|
|
|
describe '#language' do
|
|
|
|
|
it 'returns the language from lang attribute' do
|
|
|
|
|
expect(subject.language).to eq 'en'
|
|
|
|
|
end
|
|
|
|
|
end
|
|
|
|
|
end
|
|
|
|
|
|
|
|
|
|
context 'when structured data is present' do
|
|
|
|
|
let(:ld_json) do
|
|
|
|
|
{
|
|
|
|
|
'@context' => 'https://schema.org',
|
|
|
|
|
'@type' => 'NewsArticle',
|
|
|
|
|
'headline' => 'Man bites dog',
|
|
|
|
|
'description' => "A dog's tale",
|
|
|
|
|
'datePublished' => '2022-01-31T19:53:00+00:00',
|
|
|
|
|
'author' => {
|
|
|
|
|
'@type' => 'Organization',
|
|
|
|
|
'name' => 'Charlie Brown',
|
|
|
|
|
},
|
|
|
|
|
'publisher' => {
|
|
|
|
|
'@type' => 'NewsMediaOrganization',
|
|
|
|
|
'name' => 'Pet News',
|
|
|
|
|
'url' => 'https://example.com',
|
|
|
|
|
},
|
|
|
|
|
}.to_json
|
|
|
|
|
end
|
2022-02-07 12:16:31 -05:00
|
|
|
|
2023-08-03 09:41:51 -04:00
|
|
|
shared_examples 'structured data' do
|
2022-02-07 12:16:31 -05:00
|
|
|
describe '#title' do
|
|
|
|
|
it 'returns the title from structured data' do
|
2023-08-03 09:41:51 -04:00
|
|
|
expect(subject.title).to eq 'Man bites dog'
|
2022-02-07 12:16:31 -05:00
|
|
|
end
|
|
|
|
|
end
|
|
|
|
|
|
|
|
|
|
describe '#description' do
|
|
|
|
|
it 'returns the description from structured data' do
|
2023-08-03 09:41:51 -04:00
|
|
|
expect(subject.description).to eq "A dog's tale"
|
2022-02-07 12:16:31 -05:00
|
|
|
end
|
|
|
|
|
end
|
|
|
|
|
|
2023-08-03 09:41:51 -04:00
|
|
|
describe '#published_at' do
|
|
|
|
|
it 'returns the publicaton time from structured data' do
|
|
|
|
|
expect(subject.published_at).to eq '2022-01-31T19:53:00+00:00'
|
2022-02-07 12:16:31 -05:00
|
|
|
end
|
|
|
|
|
end
|
|
|
|
|
|
|
|
|
|
describe '#author_name' do
|
|
|
|
|
it 'returns the author name from structured data' do
|
2023-08-03 09:41:51 -04:00
|
|
|
expect(subject.author_name).to eq 'Charlie Brown'
|
|
|
|
|
end
|
|
|
|
|
end
|
|
|
|
|
|
|
|
|
|
describe '#provider_name' do
|
|
|
|
|
it 'returns the provider name from structured data' do
|
|
|
|
|
expect(subject.provider_name).to eq 'Pet News'
|
2022-02-07 12:16:31 -05:00
|
|
|
end
|
|
|
|
|
end
|
|
|
|
|
end
|
|
|
|
|
|
2023-08-03 09:41:51 -04:00
|
|
|
context 'when is wrapped in CDATA tags' do
|
|
|
|
|
let(:html) { <<~HTML }
|
|
|
|
|
<!doctype html>
|
|
|
|
|
<html>
|
|
|
|
|
<head>
|
|
|
|
|
<script type="application/ld+json">
|
|
|
|
|
//<![CDATA[
|
|
|
|
|
#{ld_json}
|
|
|
|
|
//]]>
|
|
|
|
|
</script>
|
|
|
|
|
</head>
|
|
|
|
|
</html>
|
|
|
|
|
HTML
|
|
|
|
|
|
|
|
|
|
include_examples 'structured data'
|
|
|
|
|
end
|
|
|
|
|
|
2023-05-03 23:49:08 -04:00
|
|
|
context 'with the first tag is invalid JSON' do
|
2023-02-17 16:56:20 -05:00
|
|
|
let(:html) { <<~HTML }
|
|
|
|
|
<!doctype html>
|
|
|
|
|
<html>
|
|
|
|
|
<body>
|
|
|
|
|
<script type="application/ld+json">
|
2023-08-03 09:41:51 -04:00
|
|
|
invalid LD+JSON
|
2023-02-17 16:56:20 -05:00
|
|
|
</script>
|
|
|
|
|
<script type="application/ld+json">
|
2023-08-03 09:41:51 -04:00
|
|
|
#{ld_json}
|
|
|
|
|
</script>
|
|
|
|
|
</body>
|
|
|
|
|
</html>
|
|
|
|
|
HTML
|
|
|
|
|
|
|
|
|
|
include_examples 'structured data'
|
|
|
|
|
end
|
|
|
|
|
|
|
|
|
|
context 'with preceding block of unsupported LD+JSON' do
|
|
|
|
|
let(:html) { <<~HTML }
|
|
|
|
|
<!doctype html>
|
|
|
|
|
<html>
|
|
|
|
|
<body>
|
|
|
|
|
<script type="application/ld+json">
|
|
|
|
|
[
|
|
|
|
|
{
|
|
|
|
|
"@context": "https://schema.org",
|
|
|
|
|
"@type": "ItemList",
|
|
|
|
|
"url": "https://example.com/cat.html",
|
|
|
|
|
"name": "Man bites cat",
|
|
|
|
|
"description": "A cat's tale"
|
2023-02-17 16:56:20 -05:00
|
|
|
},
|
2023-08-03 09:41:51 -04:00
|
|
|
{
|
|
|
|
|
"@context": "https://schema.org",
|
|
|
|
|
"@type": "BreadcrumbList",
|
|
|
|
|
"itemListElement":[
|
|
|
|
|
{
|
|
|
|
|
"@type": "ListItem",
|
|
|
|
|
"position": 1,
|
|
|
|
|
"item": {
|
|
|
|
|
"@id": "https://www.example.com",
|
|
|
|
|
"name": "Cat News"
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
]
|
2023-02-17 16:56:20 -05:00
|
|
|
}
|
2023-08-03 09:41:51 -04:00
|
|
|
]
|
|
|
|
|
</script>
|
|
|
|
|
<script type="application/ld+json">
|
|
|
|
|
#{ld_json}
|
2023-02-17 16:56:20 -05:00
|
|
|
</script>
|
|
|
|
|
</body>
|
|
|
|
|
</html>
|
2022-02-07 12:16:31 -05:00
|
|
|
HTML
|
|
|
|
|
|
2023-08-03 09:41:51 -04:00
|
|
|
include_examples 'structured data'
|
|
|
|
|
end
|
|
|
|
|
|
|
|
|
|
context 'with unsupported in same block LD+JSON' do
|
|
|
|
|
let(:html) { <<~HTML }
|
|
|
|
|
<!doctype html>
|
|
|
|
|
<html>
|
|
|
|
|
<body>
|
|
|
|
|
<script type="application/ld+json">
|
|
|
|
|
[
|
|
|
|
|
{
|
|
|
|
|
"@context": "https://schema.org",
|
|
|
|
|
"@type": "ItemList",
|
|
|
|
|
"url": "https://example.com/cat.html",
|
|
|
|
|
"name": "Man bites cat",
|
|
|
|
|
"description": "A cat's tale"
|
|
|
|
|
},
|
|
|
|
|
#{ld_json}
|
|
|
|
|
]
|
|
|
|
|
</script>
|
|
|
|
|
</body>
|
|
|
|
|
</html>
|
|
|
|
|
HTML
|
|
|
|
|
|
|
|
|
|
include_examples 'structured data'
|
|
|
|
|
end
|
|
|
|
|
end
|
|
|
|
|
|
|
|
|
|
context 'when Open Graph protocol data is present' do
|
|
|
|
|
let(:html) { <<~HTML }
|
|
|
|
|
<!doctype html>
|
|
|
|
|
<html>
|
|
|
|
|
<head>
|
|
|
|
|
<meta property="og:url" content="https://example.com/dog.html">
|
|
|
|
|
<meta property="og:title" content="Man bites dog">
|
|
|
|
|
<meta property="og:description" content="A dog's tale">
|
|
|
|
|
<meta property="article:published_time" content="2022-01-31T19:53:00+00:00">
|
|
|
|
|
<meta property="og:author" content="Charlie Brown">
|
|
|
|
|
<meta property="og:locale" content="en">
|
|
|
|
|
<meta property="og:image" content="https://example.com/snoopy.jpg">
|
|
|
|
|
<meta property="og:image:alt" content="A good boy">
|
|
|
|
|
<meta property="og:site_name" content="Pet News">
|
|
|
|
|
</head>
|
|
|
|
|
</html>
|
|
|
|
|
HTML
|
|
|
|
|
|
|
|
|
|
describe '#canonical_url' do
|
|
|
|
|
it 'returns the URL from Open Graph protocol data' do
|
|
|
|
|
expect(subject.canonical_url).to eq 'https://example.com/dog.html'
|
2022-02-07 12:16:31 -05:00
|
|
|
end
|
2023-08-03 09:41:51 -04:00
|
|
|
end
|
2022-02-07 12:16:31 -05:00
|
|
|
|
2023-08-03 09:41:51 -04:00
|
|
|
describe '#title' do
|
|
|
|
|
it 'returns the title from Open Graph protocol data' do
|
|
|
|
|
expect(subject.title).to eq 'Man bites dog'
|
2022-02-07 12:16:31 -05:00
|
|
|
end
|
2023-08-03 09:41:51 -04:00
|
|
|
end
|
2022-02-07 12:16:31 -05:00
|
|
|
|
2023-08-03 09:41:51 -04:00
|
|
|
describe '#description' do
|
|
|
|
|
it 'returns the description from Open Graph protocol data' do
|
|
|
|
|
expect(subject.description).to eq "A dog's tale"
|
2022-02-07 12:16:31 -05:00
|
|
|
end
|
2023-08-03 09:41:51 -04:00
|
|
|
end
|
2022-02-07 12:16:31 -05:00
|
|
|
|
2023-08-03 09:41:51 -04:00
|
|
|
describe '#published_at' do
|
|
|
|
|
it 'returns the publicaton time from Open Graph protocol data' do
|
|
|
|
|
expect(subject.published_at).to eq '2022-01-31T19:53:00+00:00'
|
|
|
|
|
end
|
|
|
|
|
end
|
|
|
|
|
|
|
|
|
|
describe '#author_name' do
|
|
|
|
|
it 'returns the author name from Open Graph protocol data' do
|
|
|
|
|
expect(subject.author_name).to eq 'Charlie Brown'
|
|
|
|
|
end
|
|
|
|
|
end
|
|
|
|
|
|
|
|
|
|
describe '#language' do
|
|
|
|
|
it 'returns the language from Open Graph protocol data' do
|
|
|
|
|
expect(subject.language).to eq 'en'
|
|
|
|
|
end
|
|
|
|
|
end
|
|
|
|
|
|
|
|
|
|
describe '#image' do
|
|
|
|
|
it 'returns the image from Open Graph protocol data' do
|
|
|
|
|
expect(subject.image).to eq 'https://example.com/snoopy.jpg'
|
|
|
|
|
end
|
|
|
|
|
end
|
|
|
|
|
|
|
|
|
|
describe '#image:alt' do
|
|
|
|
|
it 'returns the image description from Open Graph protocol data' do
|
|
|
|
|
expect(subject.image_alt).to eq 'A good boy'
|
|
|
|
|
end
|
|
|
|
|
end
|
|
|
|
|
|
|
|
|
|
describe '#provider_name' do
|
|
|
|
|
it 'returns the provider name from Open Graph protocol data' do
|
|
|
|
|
expect(subject.provider_name).to eq 'Pet News'
|
2022-02-07 12:16:31 -05:00
|
|
|
end
|
|
|
|
|
end
|
|
|
|
|
end
|
2021-11-05 18:23:05 -04:00
|
|
|
end
|
