src/subrip_reader.cc

   1 /*
   2     Copyright (C) 2014-2020 Carl Hetherington <cth@carlh.net>
   3
   4     This program is free software; you can redistribute it and/or modify
   5     it under the terms of the GNU General Public License as published by
   6     the Free Software Foundation; either version 2 of the License, or
   7     (at your option) any later version.
   8
   9     This program is distributed in the hope that it will be useful,
  10     but WITHOUT ANY WARRANTY; without even the implied warranty of
  11     MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  12     GNU General Public License for more details.
  13
  14     You should have received a copy of the GNU General Public License
  15     along with this program; if not, write to the Free Software
  16     Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
  17
  18 */
  19
  20 /** @file  src/subrip_reader.cc
  21  *  @brief SubripReader class.
  22  */
  23
  24 #include "subrip_reader.h"
  25 #include "exceptions.h"
  26 #include "util.h"
  27 #include "sub_assert.h"
  28 #include "raw_convert.h"
  29 #include "ssa_reader.h"
  30 #include <boost/algorithm/string.hpp>
  31 #include <boost/algorithm/string_regex.hpp>
  32 #include <boost/lexical_cast.hpp>
  33 #include <boost/regex.hpp>
  34 #include <boost/bind.hpp>
  35 #include <cstdio>
  36 #include <vector>
  37 #include <iostream>
  38
  39 using std::string;
  40 using std::vector;
  41 using std::cout;
  42 using std::hex;
  43 using boost::lexical_cast;
  44 using boost::to_upper;
  45 using boost::optional;
  46 using boost::function;
  47 using boost::algorithm::replace_all;
  48 using namespace sub;
  49
  50 /** @param s Subtitle string encoded in UTF-8 */
  51 SubripReader::SubripReader (string s)
  52 {
  53         this->read (boost::bind(&get_line_string, &s));
  54 }
  55
  56 /** @param f Subtitle file encoded in UTF-8 */
  57 SubripReader::SubripReader (FILE* f)
  58 {
  59         this->read (boost::bind (&get_line_file, f));
  60 }
  61
  62 void
  63 SubripReader::read (function<optional<string> ()> get_line)
  64 {
  65         enum {
  66                 COUNTER,
  67                 METADATA,
  68                 CONTENT
  69         } state = COUNTER;
  70
  71         RawSubtitle rs;
  72
  73         rs.vertical_position.line = 0;
  74         rs.vertical_position.reference = TOP_OF_SUBTITLE;
  75
  76         while (true) {
  77                 auto line = get_line ();
  78                 if (!line) {
  79                         break;
  80                 }
  81
  82                 trim_right_if (*line, boost::is_any_of ("\n\r"));
  83                 remove_unicode_bom (line);
  84
  85                 /* Keep some history in case there is an error to report */
  86                 _context.push_back (*line);
  87                 if (_context.size() > 5) {
  88                         _context.pop_front ();
  89                 }
  90
  91                 switch (state) {
  92                 case COUNTER:
  93                 {
  94                         if (line->empty ()) {
  95                                 /* a blank line at the start is ok */
  96                                 break;
  97                         }
  98
  99                         state = METADATA;
 100
 101                         /* Reset stuff that should not persist across separate subtitles */
 102                         rs.bold = false;
 103                         rs.italic = false;
 104                         rs.underline = false;
 105                         rs.vertical_position.line = 0;
 106                         rs.vertical_position.reference = TOP_OF_SUBTITLE;
 107                 }
 108                 break;
 109                 case METADATA:
 110                 {
 111                         vector<string> p;
 112
 113                         /* Further trim this line, removing spaces from the end */
 114                         trim_right_if (*line, boost::is_any_of (" "));
 115
 116                         boost::algorithm::split (p, *line, boost::algorithm::is_any_of (" "), boost::token_compress_on);
 117                         if (p.size() != 3 && p.size() != 7) {
 118                                 for (int i = 0; i < 2; ++i) {
 119                                         optional<string> ex = get_line ();
 120                                         if (ex) {
 121                                                 _context.push_back (*ex);
 122                                         }
 123                                 }
 124                                 throw SubripError (*line, "a time/position line", _context);
 125                         }
 126
 127                         string expected;
 128                         auto from = convert_time(p[0], &expected);
 129                         if (!from) {
 130                                 throw SubripError(p[0], expected, _context);
 131                         }
 132                         rs.from = *from;
 133
 134                         auto to = convert_time(p[2], &expected);
 135                         if (!to) {
 136                                 throw SubripError(p[2], expected, _context);
 137                         }
 138                         rs.to = *to;
 139
 140                         /* XXX: should not ignore coordinate specifications */
 141
 142                         state = CONTENT;
 143                         break;
 144                 }
 145                 case CONTENT:
 146                         if (line->empty ()) {
 147                                 state = COUNTER;
 148                         } else {
 149                                 vector<string> sub_lines;
 150                                 /* Split up this line on unicode "LINE SEPARATOR".  This feels hacky but also
 151                                  * the least unpleasant place to do it.
 152                                  */
 153                                 boost::algorithm::split_regex(sub_lines, *line, boost::regex("\xe2\x80\xa8"));
 154                                 for (auto sub_line: sub_lines) {
 155                                         convert_line(sub_line, rs);
 156                                         rs.vertical_position.line = rs.vertical_position.line.get() + 1;
 157                                         rs.text.clear();
 158                                 }
 159                         }
 160                         break;
 161                 }
 162         }
 163 }
 164
 165 optional<Time>
 166 SubripReader::convert_time(string t, string* expected)
 167 {
 168         auto report_expected = [expected](string const& s) {
 169                 if (expected) {
 170                         *expected = s;
 171                 }
 172         };
 173
 174         vector<string> a;
 175         boost::algorithm::split (a, t, boost::is_any_of (":"));
 176         if (a.size() != 3) {
 177                 report_expected("time in the format h:m:s,ms");
 178                 return {};
 179         }
 180
 181         vector<string> b;
 182         boost::algorithm::split (b, a[2], boost::is_any_of (","));
 183         if (b.size() != 2) {
 184                 report_expected("time in the format h:m:s,ms");
 185                 return {};
 186         }
 187
 188         int h, m, s, ms;
 189
 190         try {
 191                 h = lexical_cast<int>(a[0]);
 192         } catch (boost::bad_lexical_cast &) {
 193                 report_expected("integer hour value");
 194                 return {};
 195         }
 196
 197         try {
 198                 m = lexical_cast<int>(a[1]);
 199         } catch (boost::bad_lexical_cast &) {
 200                 report_expected("integer minute value");
 201                 return {};
 202         }
 203
 204         try {
 205                 s = lexical_cast<int>(b[0]);
 206         } catch (boost::bad_lexical_cast &) {
 207                 report_expected("integer second value");
 208                 return {};
 209         }
 210
 211         try {
 212                 ms = lexical_cast<int>(b[1]);
 213         } catch (boost::bad_lexical_cast &) {
 214                 report_expected("integer millisecond value");
 215                 return {};
 216         }
 217
 218         return Time::from_hms (h, m, s, ms);
 219 }
 220
 221 void
 222 SubripReader::convert_line (string t, RawSubtitle& p)
 223 {
 224         vector<Colour> colours;
 225         colours.push_back (Colour (1, 1, 1));
 226
 227         auto has_next = [](string line, size_t& index, string s) {
 228                 boost::to_lower(s);
 229                 auto next = line.substr(index, s.size());
 230                 boost::to_lower(next);
 231                 if (next != s) {
 232                         return false;
 233                 }
 234
 235                 index += s.size();
 236                 return true;
 237         };
 238
 239         size_t i = 0;
 240         while (i < t.size()) {
 241                 if (has_next(t, i, "<b>") || has_next(t, i, "{b}")) {
 242                         maybe_content (p);
 243                         p.bold = true;
 244                 } else if (has_next(t, i, "</b>") || has_next(t, i, "{/b}")) {
 245                         maybe_content (p);
 246                         p.bold = false;
 247                 } else if (has_next(t, i, "<i>") || has_next(t, i, "{i}")) {
 248                         maybe_content (p);
 249                         p.italic = true;
 250                 } else if (has_next(t, i, "</i>") || has_next(t, i, "{/i}")) {
 251                         maybe_content (p);
 252                         p.italic = false;
 253                 } else if (has_next(t, i, "<u>") || has_next(t, i, "{u}")) {
 254                         maybe_content (p);
 255                         p.underline = true;
 256                 } else if (has_next(t, i, "</u>") || has_next(t, i, "{/u}")) {
 257                         maybe_content (p);
 258                         p.underline = false;
 259                 } else if (has_next(t, i, "<font") || has_next(t, i, "<Font")) {
 260                         maybe_content (p);
 261                         boost::regex re (".*color=\"?#([[:xdigit:]]+)\"?");
 262                         boost::smatch match;
 263                         string tag;
 264                         while (i < t.size() && t[i] != '>') {
 265                                 tag += t[i];
 266                                 ++i;
 267                         }
 268                         ++i;
 269                         if (boost::regex_search (tag, match, re) && string (match[1]).size() == 6) {
 270                                 p.colour = Colour::from_rgb_hex (match[1]);
 271                                 colours.push_back (p.colour);
 272                         } else {
 273                                 re = boost::regex (
 274                                         ".*color=\"rgba\\("
 275                                         "[[:space:]]*([[:digit:]]+)[[:space:]]*,"
 276                                         "[[:space:]]*([[:digit:]]+)[[:space:]]*,"
 277                                         "[[:space:]]*([[:digit:]]+)[[:space:]]*,"
 278                                         "[[:space:]]*([[:digit:]]+)[[:space:]]*"
 279                                         "\\)\""
 280                                         );
 281                                 if (boost::regex_search (tag, match, re) && match.size() == 5) {
 282                                         p.colour.r = raw_convert<int>(string(match[1])) / 255.0;
 283                                         p.colour.g = raw_convert<int>(string(match[2])) / 255.0;
 284                                         p.colour.b = raw_convert<int>(string(match[3])) / 255.0;
 285                                         colours.push_back (p.colour);
 286                                 } else {
 287                                         throw SubripError (tag, "a colour in the format #rrggbb or rgba(rr,gg,bb,aa)", _context);
 288                                 }
 289                         }
 290                 } else if (has_next(t, i, "</font>")) {
 291                         maybe_content (p);
 292                         SUB_ASSERT (!colours.empty());
 293                         colours.pop_back ();
 294                         p.colour = colours.back ();
 295                 } else if (has_next(t, i, "{\\")) {
 296                         string ssa = "\\";
 297                         while (i < t.size() && t[i] != '}') {
 298                                 ssa += t[i];
 299                                 ++i;
 300                         }
 301                         ++i;
 302                         SSAReader::parse_style (p, ssa, 288, 288, Colour(1, 1, 1));
 303                 } else {
 304                         p.text += t[i];
 305                         ++i;
 306                 }
 307         }
 308
 309         /* Strip Unicode U+202B (right-to-left embedding) as sometimes it is rendered
 310            as a missing character.  This may be a hack.
 311         */
 312         replace_all (p.text, "\xe2\x80\xab", "");
 313
 314         maybe_content (p);
 315 }
 316
 317 /* Push p into _subs if it has some text, and clear the text out of p */
 318 void
 319 SubripReader::maybe_content (RawSubtitle& p)
 320 {
 321         if (!p.text.empty ()) {
 322                 _subs.push_back (p);
 323                 p.text.clear ();
 324         }
 325 }