☰

Extract Special Characters

Extract special character data from ocr-gt-tools wiki

  1. // ==UserScript==
  2. // @name        Extract Special Characters
  3. // @namespace   http://github.com/kba/
  4. // @include     https://github.com/UB-Mannheim/ocr-gt-tools/wiki/Special-Characters
  5. // @include     https://github.com/UB-Mannheim/ocr-gt-tools/wiki/Error-Tags
  6. // @description Extract special character data from ocr-gt-tools wiki
  7. // @version     1
  8. // @require     https://code.jquery.com/jquery-2.2.3.min.js
  9. // @require     https://cdnjs.cloudflare.com/ajax/libs/z-schema/3.17.0/ZSchema-browser.js
  10. // @grant       GM_addStyle
  11. // @grant       GM_setClipboard
  12. // ==/UserScript==
  13. /*globals GM_addStyle */
  14. /*globals ZSchema */
  15.  
  16. var CSS = `
  17. pre.schema-error
  18. {
  19.     background: #a00;
  20.     color: white;
  21.     white-space: pre-wrap;
  22. }
  23. div#glyph-bar
  24. {
  25.     font-size: x-large;
  26.     position:fixed;
  27.     bottom: 0;
  28.     height: 48px;
  29.     border: 2px solid black;
  30.     background: white;
  31.     width: 100%;
  32. }
  33. div#glyph-bar .left * { float: left; }
  34. div#glyph-bar .right * { float: right; }
  35. div#glyph-bar *
  36. {
  37.     height: 100%;
  38.     font-size: x-large;
  39. }
  40. div#glyph-bar input[type='text']
  41. {
  42.     font-family: "Garamond", "Bookman", serif;
  43. }
  44. div#schema-bar
  45. {
  46.     position: fixed;
  47.     z-index: 3000;
  48.     top: 0;
  49.     background: #900;
  50.     color: white !important;
  51.     width: 100%;
  52.     font-size: x-large;
  53.     height: 48px;
  54.     border: 2px solid black;
  55. }
  56. div#schema-invalid
  57. {
  58.     display: none;
  59. }
  60. div#schema-invalid a
  61. {
  62.     display: inline-block;
  63.     color: white !important;
  64.     float: none;
  65.     margin: 0 2px;
  66. }
  67. `;
  68.  
  69. var SCHEMAS = {
  70.     'Special-Characters': {
  71.         'type': 'object',
  72.         "additionalProperties": false,
  73.         'properties': {
  74.             'id': {
  75.                 'type': 'string',
  76.                 'pattern': '^[a-z0-9-]+$',
  77.             },
  78.             'sample': {
  79.                 'type': 'array',
  80.                 'items': {
  81.                     'type': 'string',
  82.                     'pattern': '^<a.*<img.*',
  83.                 }
  84.             },
  85.             'recognition': {
  86.                 'type': 'string'
  87.             },
  88.             'baseLetter': {
  89.                 'type': 'array'
  90.             },
  91.             'name': {
  92.                 'type': 'object'
  93.             },
  94.             'notes': {
  95.                 'type': 'object'
  96.             },
  97.             'shortcutLinux': {
  98.                 'type': 'string',
  99.                 'pattern': '^<kbd',
  100.             },
  101.             'shortcutWindows': {
  102.                 'type': 'string',
  103.                 'pattern': '^<kbd',
  104.             },
  105.         },
  106.         'required': ['sample', 'recognition', 'baseLetter'],
  107.     },
  108.     'Error-Tags': {
  109.         'type': 'object',
  110.         "additionalProperties": false,
  111.         'properties': {
  112.             'id': {
  113.                 'type': 'string',
  114.                 'pattern': '^[a-z0-9-]+$',
  115.             },
  116.             'name': {
  117.                 'type': 'object',
  118.                 'properties': {
  119.                     'de': {
  120.                         'type': 'string',
  121.                     },
  122.                     'en': {
  123.                         'type': 'string',
  124.                     },
  125.                 },
  126.                 'required': ['de']
  127.             },
  128.             'frequencyAvg': {
  129.                 'type': 'number',
  130.                 // 'format': 'float',
  131.             },
  132.             'total': {
  133.                 'type': 'number',
  134.                 // 'format': 'integer',
  135.             },
  136.             'comment': {
  137.                 'type': 'object',
  138.                 'properties': {
  139.                     'de': {
  140.                         'type': 'string',
  141.                     },
  142.                     'en': {
  143.                         'type': 'string',
  144.                     },
  145.                 },
  146.                 'required': ['de']
  147.             },
  148.         },
  149.         'required': ['name'],
  150.     }
  151. };
  152.  
  153. // var log = {
  154. //     'debug': console.log.bind(console),
  155. //     'info': console.info.bind(console),
  156. //     'error': console.error.bind(console),
  157. // };
  158.  
  159. var ON_LOAD = {
  160.     'Special-Characters': function(scraped) {
  161.         $("body").append(`
  162.             <div id="glyph-bar">
  163.                 <div class="left">
  164.                     <label for="glyph-input" style="font-family: monospace; font-size: 30px">TRY&gt;</label>
  165.                     <input id="glyph-input" type="text"/>
  166.                     <div id="glyph-propose">&nbsp;</div>
  167.                 </div>
  168.             </div>
  169.         `);
  170.         $("#glyph-input").on('keyup', function(e) {
  171.             var $input = $("#glyph-input");
  172.             var from = $input[0].selectionStart;
  173.             var to = $input[0].selectionEnd;
  174.             if (from == to) {
  175.                 from -= 1;
  176.             }
  177.             $('#glyph-propose').empty();
  178.             var $propose = $('#glyph-propose');
  179.             var val = $input.val();
  180.             var chosen = val.substring(from, to);
  181.             console.log(chosen, from, to);
  182.             $.each(scraped, function() {
  183.                 var glyphDesc = this;
  184.                 if (glyphDesc.baseLetter.indexOf(chosen) === -1) {
  185.                     return;
  186.                 }
  187.                 $.each(glyphDesc.sample, function(i, sample) {
  188.                     $propose.append($(sample)
  189.                         .on('click', function(e) {
  190.                             e.preventDefault();
  191.                             $input.val(val.substr(0, from) + glyphDesc.recognition + val.substr(to));
  192.                         }));
  193.                 });
  194.             });
  195.         });
  196.     },
  197.     'Error-Tags': function(scraped) {
  198.         window.alert('Not Implemented');
  199.     }
  200. };
  201.  
  202. function scrapeJsonFromWikiPage(schema) {
  203.     var parsed = {};
  204.     var validator = new ZSchema();
  205.     var h2s = $(".markdown-body h2").get();
  206.     for (var i = 0; i < h2s.length; i++) {
  207.         var $h2 = $(h2s[i]);
  208.         var thingDesc = {};
  209.         var thingId = $h2.text().trim();
  210.         parsed[thingId] = thingDesc;
  211.         var lis = $h2.next('ul').find('li').get();
  212.         for (var j = 0; j < lis.length; j++) {
  213.             var liHtml = $(lis[j]).html();
  214.             var colonIndex = liHtml.indexOf(':');
  215.             var varName = liHtml.substring(0, colonIndex)
  216.                 .toLowerCase()
  217.                 .replace(/[^a-z0-9]+/g, '_')
  218.                 .replace(/_([a-z])/g, function(orig, ch) {
  219.                     return ch.toUpperCase();
  220.                 })
  221.                 .replace(/^_|_$/, '');
  222.             // console.log(`Parsing '${varName}'`);
  223.             var rawValue = liHtml.substring(colonIndex + 1).trim();
  224.             if (schema.properties[varName] && schema.properties[varName].type === 'array') {
  225.                 thingDesc[varName] = rawValue.split(/\s*;\s*/);
  226.             } else if (schema.properties[varName] && schema.properties[varName].type === 'number') {
  227.                 thingDesc[varName] = parseFloat(rawValue);
  228.             } else if (/[A-Z][a-z]$/.test(varName)) {
  229.                 var lang = varName.substr(-2).toLowerCase();
  230.                 varName = varName.substring(0, varName.length - 2);
  231.                 thingDesc[varName] = thingDesc[varName] || {};
  232.                 thingDesc[varName][lang] = rawValue;
  233.             } else {
  234.                 thingDesc[varName] = rawValue;
  235.             }
  236.         }
  237.         console.log([thingDesc, schema]);
  238.         if (!validator.validate(thingId, schema.properties.id)) {
  239.             showError(thingId, validator.getLastErrors());
  240.         }
  241.         if (!validator.validate(thingDesc, schema)) {
  242.             showError(thingId, validator.getLastErrors());
  243.         }
  244.     }
  245.     return parsed;
  246. };
  247.  
  248. function escapeHTML(str) {
  249.     var entityMap = {
  250.         "&": "&amp;",
  251.         "<": "&lt;",
  252.         ">": "&gt;",
  253.         '"': '&quot;',
  254.         "'": '&#39;',
  255.         "/": '&#x2F;'
  256.     };
  257.     return String(str).replace(/[&<>"'\/]/g, function(s) {
  258.         return entityMap[s];
  259.     });
  260. }
  261.  
  262. function showError(faultyId, err) {
  263.     $(`h2:contains('${faultyId}')`).append(
  264.         `<pre class='schema-error'>${escapeHTML(JSON.stringify(err, null, 2))}</pre>`);
  265.     $("#schema-invalid").show().append(
  266.         `<a href="#${faultyId}">[${ $("#schema-invalid a").length + 1}]</a>`);
  267. }
  268.  
  269. $(function() {
  270.     GM_addStyle(CSS);
  271.     $("body").prepend(
  272.     `
  273. <div id="schema-bar">
  274.     <div id="schema-invalid">!! INVALID </div>
  275.     <div class="right">
  276.         <button id="copy-schema">Copy Schema</button>
  277.         <button id="copy-json">Copy Data</button>
  278.     </div>
  279. </div>
  280.     `);
  281.     var wikiPage = window.location.href.replace(/.*\//, '').replace(/#.*$/, '');
  282.     var schema = SCHEMAS[wikiPage];
  283.     var scraped = scrapeJsonFromWikiPage(schema);
  284.     ON_LOAD[wikiPage](scraped);
  285.     $("#copy-schema").on('click', function() {
  286.         GM_setClipboard(JSON.stringify(SCHEMAS[schema], null, 2));
  287.         window.alert("Copied JSON schema to clipboard");
  288.     });
  289.     $("#copy-json").on('click', function() {
  290.         GM_setClipboard(JSON.stringify(scraped, null, 2));
  291.         window.alert("Copied JSON schema to clipboard");
  292.     });
  293. });