nhkore 0.3.2 → 0.3.7
Sign up to get free protection for your applications and to get access to all the features.
- checksums.yaml +4 -4
- data/.yardopts +3 -0
- data/CHANGELOG.md +74 -6
- data/Gemfile.lock +86 -0
- data/README.md +49 -18
- data/Rakefile +48 -8
- data/lib/nhkore.rb +1 -22
- data/lib/nhkore/app.rb +3 -3
- data/lib/nhkore/article.rb +24 -7
- data/lib/nhkore/article_scraper.rb +21 -11
- data/lib/nhkore/cli/news_cmd.rb +2 -1
- data/lib/nhkore/cli/search_cmd.rb +2 -2
- data/lib/nhkore/cli/sift_cmd.rb +9 -112
- data/lib/nhkore/datetime_parser.rb +344 -0
- data/lib/nhkore/dict_scraper.rb +1 -1
- data/lib/nhkore/lib.rb +1 -0
- data/lib/nhkore/news.rb +13 -4
- data/lib/nhkore/scraper.rb +5 -5
- data/lib/nhkore/search_link.rb +39 -19
- data/lib/nhkore/search_scraper.rb +12 -4
- data/lib/nhkore/sifter.rb +91 -41
- data/lib/nhkore/util.rb +5 -19
- data/lib/nhkore/version.rb +1 -1
- data/lib/nhkore/word.rb +6 -2
- data/nhkore.gemspec +32 -17
- data/samples/looper.rb +82 -0
- metadata +39 -11
data/lib/nhkore/dict_scraper.rb
CHANGED
data/lib/nhkore/lib.rb
CHANGED
data/lib/nhkore/news.rb
CHANGED
@@ -49,7 +49,10 @@ module NHKore
|
|
49
49
|
end
|
50
50
|
|
51
51
|
def add_article(article,key: nil,overwrite: false)
|
52
|
-
|
52
|
+
url = article.url
|
53
|
+
url = url.to_s() unless url.nil?()
|
54
|
+
|
55
|
+
key = key.nil?() ? url : key.to_s()
|
53
56
|
|
54
57
|
if !overwrite
|
55
58
|
raise ArgumentError,"duplicate article[#{key}] in articles" if @articles.key?(key)
|
@@ -57,7 +60,7 @@ module NHKore
|
|
57
60
|
end
|
58
61
|
|
59
62
|
@articles[key] = article
|
60
|
-
@sha256s[article.sha256] =
|
63
|
+
@sha256s[article.sha256] = url
|
61
64
|
|
62
65
|
return self
|
63
66
|
end
|
@@ -91,16 +94,20 @@ module NHKore
|
|
91
94
|
end
|
92
95
|
|
93
96
|
def update_article(article,url)
|
97
|
+
url = url.to_s() unless url.nil?()
|
98
|
+
|
94
99
|
# Favor https.
|
95
|
-
return if article.url =~ FAVORED_URL
|
100
|
+
return if article.url.to_s() =~ FAVORED_URL
|
96
101
|
return if url !~ FAVORED_URL
|
97
102
|
|
98
|
-
@articles.delete(article.url)
|
103
|
+
@articles.delete(article.url) # Probably no to_s() here
|
99
104
|
@articles[url] = article
|
100
105
|
article.url = url
|
101
106
|
end
|
102
107
|
|
103
108
|
def article(key)
|
109
|
+
key = key.to_s() unless key.nil?()
|
110
|
+
|
104
111
|
return @articles[key]
|
105
112
|
end
|
106
113
|
|
@@ -119,6 +126,8 @@ module NHKore
|
|
119
126
|
end
|
120
127
|
|
121
128
|
def article?(key)
|
129
|
+
key = key.to_s() unless key.nil?()
|
130
|
+
|
122
131
|
return @articles.key?(key)
|
123
132
|
end
|
124
133
|
|
data/lib/nhkore/scraper.rb
CHANGED
@@ -21,6 +21,7 @@
|
|
21
21
|
#++
|
22
22
|
|
23
23
|
|
24
|
+
require 'attr_bool'
|
24
25
|
require 'nokogiri'
|
25
26
|
require 'open-uri'
|
26
27
|
|
@@ -34,14 +35,16 @@ module NHKore
|
|
34
35
|
# @since 0.2.0
|
35
36
|
###
|
36
37
|
class Scraper
|
38
|
+
extend AttrBool::Ext
|
39
|
+
|
37
40
|
DEFAULT_HEADER = {
|
38
41
|
'user-agent' => UserAgents.sample(),
|
39
42
|
'accept' => 'text/html,application/xhtml+xml,application/xml,application/rss+xml,text/xml;image/webp,image/apng,*/*;application/signed-exchange',
|
40
43
|
'dnt' => '1',
|
41
44
|
}
|
42
45
|
|
43
|
-
attr_accessor :eat_cookie
|
44
|
-
attr_accessor :is_file
|
46
|
+
attr_accessor? :eat_cookie
|
47
|
+
attr_accessor? :is_file
|
45
48
|
attr_reader :kargs
|
46
49
|
attr_accessor :max_redirects
|
47
50
|
attr_accessor :max_retries
|
@@ -49,9 +52,6 @@ module NHKore
|
|
49
52
|
attr_accessor :str_or_io
|
50
53
|
attr_accessor :url
|
51
54
|
|
52
|
-
alias_method :eat_cookie?,:eat_cookie
|
53
|
-
alias_method :is_file?,:is_file
|
54
|
-
|
55
55
|
# +max_redirects+ defaults to 3 for safety (infinite-loop attack).
|
56
56
|
#
|
57
57
|
# All URL options: https://ruby-doc.org/stdlib-2.7.0/libdoc/open-uri/rdoc/OpenURI/OpenRead.html
|
data/lib/nhkore/search_link.rb
CHANGED
@@ -21,6 +21,7 @@
|
|
21
21
|
#++
|
22
22
|
|
23
23
|
|
24
|
+
require 'attr_bool'
|
24
25
|
require 'time'
|
25
26
|
|
26
27
|
require 'nhkore/fileable'
|
@@ -33,14 +34,14 @@ module NHKore
|
|
33
34
|
# @since 0.2.0
|
34
35
|
###
|
35
36
|
class SearchLink
|
36
|
-
|
37
|
-
|
38
|
-
|
37
|
+
extend AttrBool::Ext
|
38
|
+
|
39
|
+
attr_reader :datetime
|
40
|
+
attr_reader :futsuurl
|
41
|
+
attr_accessor? :scraped
|
39
42
|
attr_accessor :sha256
|
40
43
|
attr_accessor :title
|
41
|
-
|
42
|
-
|
43
|
-
alias_method :scraped?,:scraped
|
44
|
+
attr_reader :url
|
44
45
|
|
45
46
|
def initialize(url,scraped: false)
|
46
47
|
super()
|
@@ -50,29 +51,27 @@ module NHKore
|
|
50
51
|
@scraped = scraped
|
51
52
|
@sha256 = sha256
|
52
53
|
@title = nil
|
53
|
-
|
54
|
+
self.url = url
|
54
55
|
end
|
55
56
|
|
56
57
|
def encode_with(coder)
|
57
58
|
# Order matters.
|
58
59
|
|
59
|
-
coder[:url] = @url
|
60
|
+
coder[:url] = @url.nil?() ? nil : @url.to_s()
|
60
61
|
coder[:scraped] = @scraped
|
61
|
-
coder[:datetime] = @datetime.nil?() ?
|
62
|
+
coder[:datetime] = @datetime.nil?() ? nil : @datetime.iso8601()
|
62
63
|
coder[:title] = @title
|
63
|
-
coder[:futsuurl] = @futsuurl
|
64
|
+
coder[:futsuurl] = @futsuurl.nil?() ? nil : @futsuurl.to_s()
|
64
65
|
coder[:sha256] = @sha256
|
65
66
|
end
|
66
67
|
|
67
68
|
def self.load_data(key,hash)
|
68
|
-
datetime = hash[:datetime]
|
69
|
-
|
70
69
|
slink = SearchLink.new(
|
71
70
|
hash[:url],
|
72
|
-
scraped: hash[:scraped]
|
71
|
+
scraped: hash[:scraped],
|
73
72
|
)
|
74
73
|
|
75
|
-
slink.datetime =
|
74
|
+
slink.datetime = hash[:datetime]
|
76
75
|
slink.futsuurl = hash[:futsuurl]
|
77
76
|
slink.sha256 = hash[:sha256]
|
78
77
|
slink.title = hash[:title]
|
@@ -83,13 +82,31 @@ module NHKore
|
|
83
82
|
def update_from_article(article)
|
84
83
|
# Don't update the url, as it may be different (e.g., http vs https).
|
85
84
|
|
86
|
-
|
87
|
-
|
85
|
+
self.datetime = article.datetime if @datetime.nil?()
|
86
|
+
self.futsuurl = article.futsuurl if Util.empty_web_str?(@futsuurl)
|
88
87
|
@scraped = true # If we have an article, it's been scraped
|
89
88
|
@sha256 = article.sha256 if Util.empty_web_str?(@sha256)
|
90
89
|
@title = article.title if Util.empty_web_str?(@title)
|
91
90
|
end
|
92
91
|
|
92
|
+
def datetime=(value)
|
93
|
+
if value.is_a?(Time)
|
94
|
+
@datetime = value
|
95
|
+
else
|
96
|
+
@datetime = Util.empty_web_str?(value) ? nil : Time.iso8601(value)
|
97
|
+
end
|
98
|
+
end
|
99
|
+
|
100
|
+
def futsuurl=(value)
|
101
|
+
# Don't store URI, store String.
|
102
|
+
@futsuurl = value.nil?() ? nil : value.to_s()
|
103
|
+
end
|
104
|
+
|
105
|
+
def url=(value)
|
106
|
+
# Don't store URI, store String.
|
107
|
+
@url = value.nil?() ? nil : value.to_s()
|
108
|
+
end
|
109
|
+
|
93
110
|
def to_s(mini: false)
|
94
111
|
s = ''.dup()
|
95
112
|
|
@@ -137,9 +154,11 @@ module NHKore
|
|
137
154
|
end
|
138
155
|
|
139
156
|
def add_link(link)
|
140
|
-
|
157
|
+
url = link.url.nil?() ? nil : link.url.to_s()
|
158
|
+
|
159
|
+
return self if @links.key?(url)
|
141
160
|
|
142
|
-
@links[
|
161
|
+
@links[url] = link
|
143
162
|
|
144
163
|
return self
|
145
164
|
end
|
@@ -163,7 +182,7 @@ module NHKore
|
|
163
182
|
|
164
183
|
if !links.nil?()
|
165
184
|
links.each() do |key,hash|
|
166
|
-
key = key.to_s()
|
185
|
+
key = key.to_s() unless key.nil?()
|
167
186
|
slinks.links[key] = SearchLink.load_data(key,hash)
|
168
187
|
end
|
169
188
|
end
|
@@ -173,6 +192,7 @@ module NHKore
|
|
173
192
|
|
174
193
|
def [](url)
|
175
194
|
url = url.url if url.respond_to?(:url)
|
195
|
+
url = url.to_s() unless url.nil?()
|
176
196
|
|
177
197
|
return @links[url]
|
178
198
|
end
|
@@ -45,6 +45,16 @@ module NHKore
|
|
45
45
|
# - https://www3.nhk.or.jp/news/easy/article/disaster_heat.html
|
46
46
|
YASASHII_REGEX = /\A[^\.]+\.#{Regexp.quote(YASASHII_SITE)}.+\.html?/i
|
47
47
|
|
48
|
+
IGNORE_LINK_REGEX = %r{
|
49
|
+
/about\.html? # https://www3.nhk.or.jp/news/easy/about.html
|
50
|
+
|/movieplayer\.html? # https://www3.nhk.or.jp/news/easy/movieplayer.html?id=k10038422811_1207251719_1207251728.mp4&teacuprbbs=4feb73432045dbb97c283d64d459f7cf
|
51
|
+
|/audio\.html? # https://www3.nhk.or.jp/news/easy/player/audio.html?id=k10011555691000
|
52
|
+
|/news/easy/index\.html? # http://www3.nhk.or.jp/news/easy/index.html
|
53
|
+
# https://cgi2.nhk.or.jp/news/easy/easy_enq/bin/form/enqform.html?id=k10011916321000&title=日本の会社が作った鉄道の車両「あずま」がイギリスで走る
|
54
|
+
# https://www3.nhk.or.jp/news/easy/easy_enq/bin/form/enqform.html?id=k10012689671000&title=「鬼滅の刃」の映画が台湾でも始まって大勢の人が見に行く
|
55
|
+
|/enqform\.html?
|
56
|
+
}x
|
57
|
+
|
48
58
|
# Search Engines are strict, so trigger using the default HTTP header fields
|
49
59
|
# with +header: {}+ and fetch/set the cookie using +eat_cookie: true+.
|
50
60
|
def initialize(url,eat_cookie: true,header: {},**kargs)
|
@@ -57,10 +67,8 @@ module NHKore
|
|
57
67
|
link = Util.unspace_web_str(link).downcase() unless cleaned
|
58
68
|
|
59
69
|
return true if link.empty?()
|
60
|
-
|
61
|
-
return true if
|
62
|
-
return true if link =~ /\/audio\.html?/ # https://www3.nhk.or.jp/news/easy/player/audio.html?id=k10011555691000
|
63
|
-
return true if link =~ /\/news\/easy\/index\.html?/ # http://www3.nhk.or.jp/news/easy/index.html
|
70
|
+
|
71
|
+
return true if IGNORE_LINK_REGEX.match?(link)
|
64
72
|
|
65
73
|
return false
|
66
74
|
end
|
data/lib/nhkore/sifter.rb
CHANGED
@@ -60,6 +60,40 @@ module NHKore
|
|
60
60
|
@output = nil
|
61
61
|
end
|
62
62
|
|
63
|
+
def build_header()
|
64
|
+
header = []
|
65
|
+
|
66
|
+
header << 'Frequency' unless @ignores[:freq]
|
67
|
+
header << 'Word' unless @ignores[:word]
|
68
|
+
header << 'Kana' unless @ignores[:kana]
|
69
|
+
header << 'English' unless @ignores[:eng]
|
70
|
+
header << 'Definition' unless @ignores[:defn]
|
71
|
+
|
72
|
+
return header
|
73
|
+
end
|
74
|
+
|
75
|
+
def build_rows(words)
|
76
|
+
rows = []
|
77
|
+
|
78
|
+
words.each() do |word|
|
79
|
+
rows << build_word_row(word)
|
80
|
+
end
|
81
|
+
|
82
|
+
return rows
|
83
|
+
end
|
84
|
+
|
85
|
+
def build_word_row(word)
|
86
|
+
row = []
|
87
|
+
|
88
|
+
row << word.freq unless @ignores[:freq]
|
89
|
+
row << word.word unless @ignores[:word]
|
90
|
+
row << word.kana unless @ignores[:kana]
|
91
|
+
row << word.eng unless @ignores[:eng]
|
92
|
+
row << word.defn unless @ignores[:defn]
|
93
|
+
|
94
|
+
return row
|
95
|
+
end
|
96
|
+
|
63
97
|
def filter?(article)
|
64
98
|
return false if @filters.empty?()
|
65
99
|
|
@@ -151,26 +185,10 @@ module NHKore
|
|
151
185
|
words = sift()
|
152
186
|
|
153
187
|
@output = CSV.generate(headers: :first_row,write_headers: true) do |csv|
|
154
|
-
|
155
|
-
|
156
|
-
row << 'Frequency' unless @ignores[:freq]
|
157
|
-
row << 'Word' unless @ignores[:word]
|
158
|
-
row << 'Kana' unless @ignores[:kana]
|
159
|
-
row << 'English' unless @ignores[:eng]
|
160
|
-
row << 'Definition' unless @ignores[:defn]
|
161
|
-
|
162
|
-
csv << row
|
188
|
+
csv << build_header()
|
163
189
|
|
164
190
|
words.each() do |word|
|
165
|
-
|
166
|
-
|
167
|
-
row << word.freq unless @ignores[:freq]
|
168
|
-
row << word.word unless @ignores[:word]
|
169
|
-
row << word.kana unless @ignores[:kana]
|
170
|
-
row << word.eng unless @ignores[:eng]
|
171
|
-
row << word.defn unless @ignores[:defn]
|
172
|
-
|
173
|
-
csv << row
|
191
|
+
csv << build_word_row(word)
|
174
192
|
end
|
175
193
|
end
|
176
194
|
|
@@ -232,7 +250,7 @@ module NHKore
|
|
232
250
|
<h2>#{@caption}</h2>
|
233
251
|
<table>
|
234
252
|
EOH
|
235
|
-
#"
|
253
|
+
#"
|
236
254
|
|
237
255
|
# If have too few or too many '<col>', invalid HTML.
|
238
256
|
@output << %Q{<col style="width:6em;">\n} unless @ignores[:freq]
|
@@ -242,20 +260,20 @@ module NHKore
|
|
242
260
|
@output << "<col>\n" unless @ignores[:defn] # No width for defn, fills rest of page
|
243
261
|
|
244
262
|
@output << '<tr>'
|
245
|
-
|
246
|
-
|
247
|
-
|
248
|
-
|
249
|
-
|
263
|
+
|
264
|
+
build_header().each() do |h|
|
265
|
+
@output << "<th>#{h}</th>"
|
266
|
+
end
|
267
|
+
|
250
268
|
@output << "</tr>\n"
|
251
269
|
|
252
270
|
words.each() do |word|
|
253
271
|
@output << '<tr>'
|
254
|
-
|
255
|
-
|
256
|
-
|
257
|
-
|
258
|
-
|
272
|
+
|
273
|
+
build_word_row(word).each() do |w|
|
274
|
+
@output << "<td>#{Util.escape_html(w.to_s())}</td>"
|
275
|
+
end
|
276
|
+
|
259
277
|
@output << "</tr>\n"
|
260
278
|
end
|
261
279
|
|
@@ -264,31 +282,63 @@ module NHKore
|
|
264
282
|
</body>
|
265
283
|
</html>
|
266
284
|
EOH
|
267
|
-
#/
|
285
|
+
#/
|
268
286
|
|
269
287
|
return @output
|
270
288
|
end
|
271
289
|
|
272
|
-
def
|
290
|
+
def put_json!()
|
291
|
+
require 'json'
|
292
|
+
|
273
293
|
words = sift()
|
274
294
|
|
275
|
-
|
276
|
-
|
277
|
-
|
278
|
-
|
279
|
-
|
280
|
-
|
281
|
-
|
295
|
+
@output = ''.dup()
|
296
|
+
|
297
|
+
@output << <<~EOJ
|
298
|
+
{
|
299
|
+
"caption": #{JSON.generate(@caption)},
|
300
|
+
"header": #{JSON.generate(build_header())},
|
301
|
+
"words": [
|
302
|
+
EOJ
|
303
|
+
|
304
|
+
if !words.empty?()
|
305
|
+
0.upto(words.length - 2) do |i|
|
306
|
+
@output << " #{JSON.generate(build_word_row(words[i]))},\n"
|
282
307
|
end
|
308
|
+
|
309
|
+
@output << " #{JSON.generate(build_word_row(words[-1]))}\n"
|
283
310
|
end
|
284
311
|
|
312
|
+
@output << "]\n}\n"
|
313
|
+
|
314
|
+
return @output
|
315
|
+
end
|
316
|
+
|
317
|
+
def put_yaml!()
|
318
|
+
require 'psychgus'
|
319
|
+
|
320
|
+
words = sift()
|
321
|
+
|
285
322
|
yaml = {
|
286
323
|
caption: @caption,
|
287
|
-
|
324
|
+
header: build_header(),
|
325
|
+
words: build_rows(words),
|
288
326
|
}
|
289
327
|
|
328
|
+
header_styler = Class.new() do
|
329
|
+
include Psychgus::Styler
|
330
|
+
|
331
|
+
def style_sequence(sniffer,node)
|
332
|
+
parent = sniffer.parent
|
333
|
+
|
334
|
+
if !parent.nil?() && parent.node.respond_to?(:value) && parent.value == 'header'
|
335
|
+
node.style = Psychgus::SEQUENCE_FLOW
|
336
|
+
end
|
337
|
+
end
|
338
|
+
end
|
339
|
+
|
290
340
|
# Put each Word on one line (flow/inline style).
|
291
|
-
@output = Util.dump_yaml(yaml,flow_level: 4)
|
341
|
+
@output = Util.dump_yaml(yaml,flow_level: 4,stylers: header_styler.new())
|
292
342
|
|
293
343
|
return @output
|
294
344
|
end
|
@@ -306,7 +356,7 @@ module NHKore
|
|
306
356
|
|
307
357
|
words = master_article.words.values()
|
308
358
|
|
309
|
-
words
|
359
|
+
words.sort!() do |word1,word2|
|
310
360
|
# Order by freq DESC (most frequent words to top).
|
311
361
|
i = (word2.freq <=> word1.freq)
|
312
362
|
|