OSDN Git Service

c8c3f6aa338afe8794b1a45ca3ff1b6f53728743
[opentween/open-tween.git] / OpenTween / TweetExtractor.cs
1 // OpenTween - Client of Twitter
2 // Copyright (c) 2015 kim_upsilon (@kim_upsilon) <https://upsilo.net/~upsilon/>
3 // Copyright (c) 2015 takke (@takke) <http://takke.jp/>
4 // All rights reserved.
5 //
6 // This file is part of OpenTween.
7 //
8 // This program is free software; you can redistribute it and/or modify it
9 // under the terms of the GNU General Public License as published by the Free
10 // Software Foundation; either version 3 of the License, or (at your option)
11 // any later version.
12 //
13 // This program is distributed in the hope that it will be useful, but
14 // WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
15 // or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
16 // for more details.
17 //
18 // You should have received a copy of the GNU General Public License along
19 // with this program. If not, see <http://www.gnu.org/licenses/>, or write to
20 // the Free Software Foundation, Inc., 51 Franklin Street - Fifth Floor,
21 // Boston, MA 02110-1301, USA.
22
23 using System;
24 using System.Collections.Generic;
25 using System.Linq;
26 using System.Text;
27 using System.Text.RegularExpressions;
28 using System.Threading.Tasks;
29 using OpenTween.Api.DataModel;
30
31 namespace OpenTween
32 {
33     public static class TweetExtractor
34     {
35         public static readonly Regex EmojiPattern = new Regex(@"/(?:\ud83d\udc68\ud83c\udffc\u200d\ud83e\udd1d\u200d\ud83d\udc68\ud83c\udffb|\ud83d\udc68\ud83c\udffd\u200d\ud83e\udd1d\u200d\ud83d\udc68\ud83c[\udffb\udffc]|\ud83d\udc68\ud83c\udffe\u200d\ud83e\udd1d\u200d\ud83d\udc68\ud83c[\udffb-\udffd]|\ud83d\udc68\ud83c\udfff\u200d\ud83e\udd1d\u200d\ud83d\udc68\ud83c[\udffb-\udffe]|\ud83d\udc69\ud83c\udffb\u200d\ud83e\udd1d\u200d\ud83d\udc68\ud83c[\udffc-\udfff]|\ud83d\udc69\ud83c\udffc\u200d\ud83e\udd1d\u200d\ud83d\udc68\ud83c[\udffb\udffd-\udfff]|\ud83d\udc69\ud83c\udffc\u200d\ud83e\udd1d\u200d\ud83d\udc69\ud83c\udffb|\ud83d\udc69\ud83c\udffd\u200d\ud83e\udd1d\u200d\ud83d\udc68\ud83c[\udffb\udffc\udffe\udfff]|\ud83d\udc69\ud83c\udffd\u200d\ud83e\udd1d\u200d\ud83d\udc69\ud83c[\udffb\udffc]|\ud83d\udc69\ud83c\udffe\u200d\ud83e\udd1d\u200d\ud83d\udc68\ud83c[\udffb-\udffd\udfff]|\ud83d\udc69\ud83c\udffe\u200d\ud83e\udd1d\u200d\ud83d\udc69\ud83c[\udffb-\udffd]|\ud83d\udc69\ud83c\udfff\u200d\ud83e\udd1d\u200d\ud83d\udc68\ud83c[\udffb-\udffe]|\ud83d\udc69\ud83c\udfff\u200d\ud83e\udd1d\u200d\ud83d\udc69\ud83c[\udffb-\udffe]|\ud83e\uddd1\ud83c\udffb\u200d\ud83e\udd1d\u200d\ud83e\uddd1\ud83c\udffb|\ud83e\uddd1\ud83c\udffc\u200d\ud83e\udd1d\u200d\ud83e\uddd1\ud83c[\udffb\udffc]|\ud83e\uddd1\ud83c\udffd\u200d\ud83e\udd1d\u200d\ud83e\uddd1\ud83c[\udffb-\udffd]|\ud83e\uddd1\ud83c\udffe\u200d\ud83e\udd1d\u200d\ud83e\uddd1\ud83c[\udffb-\udffe]|\ud83e\uddd1\ud83c\udfff\u200d\ud83e\udd1d\u200d\ud83e\uddd1\ud83c[\udffb-\udfff]|\ud83e\uddd1\u200d\ud83e\udd1d\u200d\ud83e\uddd1|\ud83d\udc6b\ud83c[\udffb-\udfff]|\ud83d\udc6c\ud83c[\udffb-\udfff]|\ud83d\udc6d\ud83c[\udffb-\udfff]|\ud83d[\udc6b-\udc6d])|(?:\ud83d[\udc68\udc69])(?:\ud83c[\udffb-\udfff])?\u200d(?:\u2695\ufe0f|\u2696\ufe0f|\u2708\ufe0f|\ud83c[\udf3e\udf73\udf93\udfa4\udfa8\udfeb\udfed]|\ud83d[\udcbb\udcbc\udd27\udd2c\ude80\ude92]|\ud83e[\uddaf-\uddb3\uddbc\uddbd])|(?:\ud83c[\udfcb\udfcc]|\ud83d[\udd74\udd75]|\u26f9)((?:\ud83c[\udffb-\udfff]|\ufe0f)\u200d[\u2640\u2642]\ufe0f)|(?:\ud83c[\udfc3\udfc4\udfca]|\ud83d[\udc6e\udc71\udc73\udc77\udc81\udc82\udc86\udc87\ude45-\ude47\ude4b\ude4d\ude4e\udea3\udeb4-\udeb6]|\ud83e[\udd26\udd35\udd37-\udd39\udd3d\udd3e\uddb8\uddb9\uddcd-\uddcf\uddd6-\udddd])(?:\ud83c[\udffb-\udfff])?\u200d[\u2640\u2642]\ufe0f|(?:\ud83d\udc68\u200d\u2764\ufe0f\u200d\ud83d\udc8b\u200d\ud83d\udc68|\ud83d\udc68\u200d\ud83d\udc68\u200d\ud83d\udc66\u200d\ud83d\udc66|\ud83d\udc68\u200d\ud83d\udc68\u200d\ud83d\udc67\u200d\ud83d[\udc66\udc67]|\ud83d\udc68\u200d\ud83d\udc69\u200d\ud83d\udc66\u200d\ud83d\udc66|\ud83d\udc68\u200d\ud83d\udc69\u200d\ud83d\udc67\u200d\ud83d[\udc66\udc67]|\ud83d\udc69\u200d\u2764\ufe0f\u200d\ud83d\udc8b\u200d\ud83d[\udc68\udc69]|\ud83d\udc69\u200d\ud83d\udc69\u200d\ud83d\udc66\u200d\ud83d\udc66|\ud83d\udc69\u200d\ud83d\udc69\u200d\ud83d\udc67\u200d\ud83d[\udc66\udc67]|\ud83d\udc68\u200d\u2764\ufe0f\u200d\ud83d\udc68|\ud83d\udc68\u200d\ud83d\udc66\u200d\ud83d\udc66|\ud83d\udc68\u200d\ud83d\udc67\u200d\ud83d[\udc66\udc67]|\ud83d\udc68\u200d\ud83d\udc68\u200d\ud83d[\udc66\udc67]|\ud83d\udc68\u200d\ud83d\udc69\u200d\ud83d[\udc66\udc67]|\ud83d\udc69\u200d\u2764\ufe0f\u200d\ud83d[\udc68\udc69]|\ud83d\udc69\u200d\ud83d\udc66\u200d\ud83d\udc66|\ud83d\udc69\u200d\ud83d\udc67\u200d\ud83d[\udc66\udc67]|\ud83d\udc69\u200d\ud83d\udc69\u200d\ud83d[\udc66\udc67]|\ud83c\udff3\ufe0f\u200d\ud83c\udf08|\ud83c\udff4\u200d\u2620\ufe0f|\ud83d\udc15\u200d\ud83e\uddba|\ud83d\udc41\u200d\ud83d\udde8|\ud83d\udc68\u200d\ud83d[\udc66\udc67]|\ud83d\udc69\u200d\ud83d[\udc66\udc67]|\ud83d\udc6f\u200d\u2640\ufe0f|\ud83d\udc6f\u200d\u2642\ufe0f|\ud83e\udd3c\u200d\u2640\ufe0f|\ud83e\udd3c\u200d\u2642\ufe0f|\ud83e\uddde\u200d\u2640\ufe0f|\ud83e\uddde\u200d\u2642\ufe0f|\ud83e\udddf\u200d\u2640\ufe0f|\ud83e\udddf\u200d\u2642\ufe0f)|[#*0-9]\ufe0f?\u20e3|(?:[©®\u2122\u265f]\ufe0f)|(?:\ud83c[\udc04\udd70\udd71\udd7e\udd7f\ude02\ude1a\ude2f\ude37\udf21\udf24-\udf2c\udf36\udf7d\udf96\udf97\udf99-\udf9b\udf9e\udf9f\udfcd\udfce\udfd4-\udfdf\udff3\udff5\udff7]|\ud83d[\udc3f\udc41\udcfd\udd49\udd4a\udd6f\udd70\udd73\udd76-\udd79\udd87\udd8a-\udd8d\udda5\udda8\uddb1\uddb2\uddbc\uddc2-\uddc4\uddd1-\uddd3\udddc-\uddde\udde1\udde3\udde8\uddef\uddf3\uddfa\udecb\udecd-\udecf\udee0-\udee5\udee9\udef0\udef3]|[\u203c\u2049\u2139\u2194-\u2199\u21a9\u21aa\u231a\u231b\u2328\u23cf\u23ed-\u23ef\u23f1\u23f2\u23f8-\u23fa\u24c2\u25aa\u25ab\u25b6\u25c0\u25fb-\u25fe\u2600-\u2604\u260e\u2611\u2614\u2615\u2618\u2620\u2622\u2623\u2626\u262a\u262e\u262f\u2638-\u263a\u2640\u2642\u2648-\u2653\u2660\u2663\u2665\u2666\u2668\u267b\u267f\u2692-\u2697\u2699\u269b\u269c\u26a0\u26a1\u26aa\u26ab\u26b0\u26b1\u26bd\u26be\u26c4\u26c5\u26c8\u26cf\u26d1\u26d3\u26d4\u26e9\u26ea\u26f0-\u26f5\u26f8\u26fa\u26fd\u2702\u2708\u2709\u270f\u2712\u2714\u2716\u271d\u2721\u2733\u2734\u2744\u2747\u2757\u2763\u2764\u27a1\u2934\u2935\u2b05-\u2b07\u2b1b\u2b1c\u2b50\u2b55\u3030\u303d\u3297\u3299])(?:\ufe0f|(?!\ufe0e))|(?:(?:\ud83c[\udfcb\udfcc]|\ud83d[\udd74\udd75\udd90]|[\u261d\u26f7\u26f9\u270c\u270d])(?:\ufe0f|(?!\ufe0e))|(?:\ud83c[\udf85\udfc2-\udfc4\udfc7\udfca]|\ud83d[\udc42\udc43\udc46-\udc50\udc66-\udc69\udc6e\udc70-\udc78\udc7c\udc81-\udc83\udc85-\udc87\udcaa\udd7a\udd95\udd96\ude45-\ude47\ude4b-\ude4f\udea3\udeb4-\udeb6\udec0\udecc]|\ud83e[\udd0f\udd18-\udd1c\udd1e\udd1f\udd26\udd30-\udd39\udd3d\udd3e\uddb5\uddb6\uddb8\uddb9\uddbb\uddcd-\uddcf\uddd1-\udddd]|[\u270a\u270b]))(?:\ud83c[\udffb-\udfff])?|(?:\ud83c\udff4\udb40\udc67\udb40\udc62\udb40\udc65\udb40\udc6e\udb40\udc67\udb40\udc7f|\ud83c\udff4\udb40\udc67\udb40\udc62\udb40\udc73\udb40\udc63\udb40\udc74\udb40\udc7f|\ud83c\udff4\udb40\udc67\udb40\udc62\udb40\udc77\udb40\udc6c\udb40\udc73\udb40\udc7f|\ud83c\udde6\ud83c[\udde8-\uddec\uddee\uddf1\uddf2\uddf4\uddf6-\uddfa\uddfc\uddfd\uddff]|\ud83c\udde7\ud83c[\udde6\udde7\udde9-\uddef\uddf1-\uddf4\uddf6-\uddf9\uddfb\uddfc\uddfe\uddff]|\ud83c\udde8\ud83c[\udde6\udde8\udde9\uddeb-\uddee\uddf0-\uddf5\uddf7\uddfa-\uddff]|\ud83c\udde9\ud83c[\uddea\uddec\uddef\uddf0\uddf2\uddf4\uddff]|\ud83c\uddea\ud83c[\udde6\udde8\uddea\uddec\udded\uddf7-\uddfa]|\ud83c\uddeb\ud83c[\uddee-\uddf0\uddf2\uddf4\uddf7]|\ud83c\uddec\ud83c[\udde6\udde7\udde9-\uddee\uddf1-\uddf3\uddf5-\uddfa\uddfc\uddfe]|\ud83c\udded\ud83c[\uddf0\uddf2\uddf3\uddf7\uddf9\uddfa]|\ud83c\uddee\ud83c[\udde8-\uddea\uddf1-\uddf4\uddf6-\uddf9]|\ud83c\uddef\ud83c[\uddea\uddf2\uddf4\uddf5]|\ud83c\uddf0\ud83c[\uddea\uddec-\uddee\uddf2\uddf3\uddf5\uddf7\uddfc\uddfe\uddff]|\ud83c\uddf1\ud83c[\udde6-\udde8\uddee\uddf0\uddf7-\uddfb\uddfe]|\ud83c\uddf2\ud83c[\udde6\udde8-\udded\uddf0-\uddff]|\ud83c\uddf3\ud83c[\udde6\udde8\uddea-\uddec\uddee\uddf1\uddf4\uddf5\uddf7\uddfa\uddff]|\ud83c\uddf4\ud83c\uddf2|\ud83c\uddf5\ud83c[\udde6\uddea-\udded\uddf0-\uddf3\uddf7-\uddf9\uddfc\uddfe]|\ud83c\uddf6\ud83c\udde6|\ud83c\uddf7\ud83c[\uddea\uddf4\uddf8\uddfa\uddfc]|\ud83c\uddf8\ud83c[\udde6-\uddea\uddec-\uddf4\uddf7-\uddf9\uddfb\uddfd-\uddff]|\ud83c\uddf9\ud83c[\udde6\udde8\udde9\uddeb-\udded\uddef-\uddf4\uddf7\uddf9\uddfb\uddfc\uddff]|\ud83c\uddfa\ud83c[\udde6\uddec\uddf2\uddf3\uddf8\uddfe\uddff]|\ud83c\uddfb\ud83c[\udde6\udde8\uddea\uddec\uddee\uddf3\uddfa]|\ud83c\uddfc\ud83c[\uddeb\uddf8]|\ud83c\uddfd\ud83c\uddf0|\ud83c\uddfe\ud83c[\uddea\uddf9]|\ud83c\uddff\ud83c[\udde6\uddf2\uddfc]|\ud83c[\udccf\udd8e\udd91-\udd9a\udde6-\uddff\ude01\ude32-\ude36\ude38-\ude3a\ude50\ude51\udf00-\udf20\udf2d-\udf35\udf37-\udf7c\udf7e-\udf84\udf86-\udf93\udfa0-\udfc1\udfc5\udfc6\udfc8\udfc9\udfcf-\udfd3\udfe0-\udff0\udff4\udff8-\udfff]|\ud83d[\udc00-\udc3e\udc40\udc44\udc45\udc51-\udc65\udc6a-\udc6d\udc6f\udc79-\udc7b\udc7d-\udc80\udc84\udc88-\udca9\udcab-\udcfc\udcff-\udd3d\udd4b-\udd4e\udd50-\udd67\udda4\uddfb-\ude44\ude48-\ude4a\ude80-\udea2\udea4-\udeb3\udeb7-\udebf\udec1-\udec5\uded0-\uded2\uded5\udeeb\udeec\udef4-\udefa\udfe0-\udfeb]|\ud83e[\udd0d\udd0e\udd10-\udd17\udd1d\udd20-\udd25\udd27-\udd2f\udd3a\udd3c\udd3f-\udd45\udd47-\udd71\udd73-\udd76\udd7a-\udda2\udda5-\uddaa\uddae-\uddb4\uddb7\uddba\uddbc-\uddca\uddd0\uddde-\uddff\ude70-\ude73\ude78-\ude7a\ude80-\ude82\ude90-\ude95]|[\u23e9-\u23ec\u23f0\u23f3\u267e\u26ce\u2705\u2728\u274c\u274e\u2753-\u2755\u2795-\u2797\u27b0\u27bf\ue50a])|\ufe0f");
36
37         /// <summary>
38         /// テキストから URL を抽出して返します
39         /// </summary>
40         public static IEnumerable<string> ExtractUrls(string text)
41             => ExtractUrlEntities(text).Select(x => x.Url);
42
43         /// <summary>
44         /// テキストから URL を抽出してエンティティとして返します
45         /// </summary>
46         public static IEnumerable<TwitterEntityUrl> ExtractUrlEntities(string text)
47         {
48             var urlMatches = Regex.Matches(text, Twitter.rgUrl, RegexOptions.IgnoreCase).Cast<Match>();
49             foreach (var m in urlMatches)
50             {
51                 var before = m.Groups["before"].Value;
52                 var url = m.Groups["url"].Value;
53                 var protocol = m.Groups["protocol"].Value;
54                 var domain = m.Groups["domain"].Value;
55                 var path = m.Groups["path"].Value;
56
57                 var validUrl = false;
58                 if (protocol.Length == 0)
59                 {
60                     if (Regex.IsMatch(before, Twitter.url_invalid_without_protocol_preceding_chars))
61                         continue;
62
63                     string lasturl = null;
64
65                     var last_url_invalid_match = false;
66                     var domainMatches = Regex.Matches(domain, Twitter.url_valid_ascii_domain, RegexOptions.IgnoreCase).Cast<Match>();
67                     foreach (var mm in domainMatches)
68                     {
69                         lasturl = mm.Value;
70                         last_url_invalid_match = Regex.IsMatch(lasturl, Twitter.url_invalid_short_domain, RegexOptions.IgnoreCase);
71                         if (!last_url_invalid_match)
72                         {
73                             validUrl = true;
74                         }
75                     }
76
77                     if (last_url_invalid_match && path.Length != 0)
78                     {
79                         validUrl = true;
80                     }
81                 }
82                 else
83                 {
84                     validUrl = true;
85                 }
86
87                 if (validUrl)
88                 {
89                     var urlGroup = m.Groups["url"];
90                     var startPos = text.GetCodepointCount(0, urlGroup.Index);
91                     var endPos = startPos + text.GetCodepointCount(urlGroup.Index, urlGroup.Index + urlGroup.Length);
92
93                     yield return new TwitterEntityUrl
94                     {
95                         Indices = new[] { startPos, endPos },
96                         Url = url,
97                         ExpandedUrl = url,
98                         DisplayUrl = url,
99                     };
100                 }
101             }
102         }
103
104         /// <summary>
105         /// テキストからメンションを抽出してエンティティとして返します
106         /// </summary>
107         public static IEnumerable<TwitterEntityMention> ExtractMentionEntities(string text)
108         {
109             // リスト
110             var matchesAtList = Regex.Matches(text, @"(?<=^|[^a-zA-Z0-9_/])([@@][a-zA-Z0-9_]{1,20}/[a-zA-Z][a-zA-Z0-9\p{IsLatin-1Supplement}\-]{0,79})");
111             foreach (var match in matchesAtList.Cast<Match>())
112             {
113                 var startPos = text.GetCodepointCount(0, match.Index);
114                 var endPos = startPos + text.GetCodepointCount(match.Index, match.Index + match.Length);
115
116                 yield return new TwitterEntityMention
117                 {
118                     Indices = new[] { startPos, endPos },
119                     ScreenName = match.Value.Substring(1), // 先頭の「@」は取り除く
120                 };
121             }
122
123             // 通常のメンション
124             var matchesAtUser = Regex.Matches(text, "(?<=^|[^a-zA-Z0-9_/])([@@][a-zA-Z0-9_]{1,20})(?=[^a-zA-Z0-9_/]|$)");
125             foreach (var match in matchesAtUser.Cast<Match>())
126             {
127                 var startPos = text.GetCodepointCount(0, match.Index);
128                 var endPos = startPos + text.GetCodepointCount(match.Index, match.Index + match.Length);
129
130                 yield return new TwitterEntityMention
131                 {
132                     Indices = new[] { startPos, endPos },
133                     ScreenName = match.Value.Substring(1), // 先頭の「@」は取り除く
134                 };
135             }
136         }
137
138         /// <summary>
139         /// テキストからハッシュタグを抽出してエンティティとして返します
140         /// </summary>
141         public static IEnumerable<TwitterEntityHashtag> ExtractHashtagEntities(string text)
142         {
143             var matches = Regex.Matches(text, Twitter.HASHTAG);
144             foreach (var match in matches.Cast<Match>())
145             {
146                 var groupHashtagSharp = match.Groups[2];
147                 var groupHashtagText = match.Groups[3];
148                 var startPos = text.GetCodepointCount(0, groupHashtagSharp.Index);
149                 var endPos = startPos + text.GetCodepointCount(groupHashtagSharp.Index, groupHashtagSharp.Index + groupHashtagSharp.Length + groupHashtagText.Length);
150
151                 yield return new TwitterEntityHashtag
152                 {
153                     Indices = new[] { startPos, endPos },
154                     Text = groupHashtagText.Value,
155                 };
156             }
157         }
158
159         /// <summary>
160         /// テキストから絵文字を抽出して <see cref="TwitterEntityEmoji"/> として返します
161         /// </summary>
162         public static IEnumerable<TwitterEntityEmoji> ExtractEmojiEntities(string text)
163         {
164             var matches = TweetExtractor.EmojiPattern.Matches(text);
165
166             foreach (var match in matches.Cast<Match>())
167             {
168                 var input = match.Value;
169
170                 // 異字体セレクタ U+FE0F (emoji style) は除去する (ZWJ を含まない場合のみ)
171                 if (!input.Contains('\u200D'))
172                     input = input.Replace("\uFE0F", "");
173
174                 var codepointHex = new List<string>();
175
176                 for (var i = 0; i < input.Length; i += char.IsSurrogatePair(input, i) ? 2 : 1)
177                 {
178                     var codepoint = char.ConvertToUtf32(input, i);
179                     codepointHex.Add($"{codepoint:x}");
180                 }
181
182                 var startPos = text.GetCodepointCount(0, match.Index);
183                 var endPos = startPos + text.GetCodepointCount(match.Index, match.Index + match.Length);
184
185                 TwitterEntityEmoji entity;
186                 if (codepointHex.Count >= 1)
187                 {
188                     entity = new TwitterEntityEmoji
189                     {
190                         Indices = new[] { startPos, endPos },
191                         Text = input,
192                         Url = "https://twemoji.maxcdn.com/2/72x72/" + string.Join("-", codepointHex) + ".png",
193                     };
194                 }
195                 else
196                 {
197                     entity = new TwitterEntityEmoji
198                     {
199                         Indices = new[] { startPos, endPos },
200                         Text = input,
201                         Url = "",
202                     };
203                 }
204
205                 yield return entity;
206             }
207         }
208     }
209 }