const stringPadding = require('string-padding'); const stringToStream = require('string-to-stream') const BLOCK_OPEN = "[[BLOCK_OPEN]]"; const BLOCK_CLOSE = "[[BLOCK_CLOSE]]"; const NEWLINE = "[[NEWLINE]]"; const NEWLINE_MERGED = "[[MERGED]]"; const SPACE = "[[SPACE]]"; function processMdArrayNewLines(md) { while (md.length && md[0] == BLOCK_OPEN) { md.shift(); } while (md.length && md[md.length - 1] == BLOCK_CLOSE) { md.pop(); } let temp = []; let last = ''; for (let i = 0; i < md.length; i++) { let v = md[i]; if (isNewLineBlock(last) && isNewLineBlock(v) && last == v) { // Skip it } else { temp.push(v); } last = v; } md = temp; temp = []; last = ""; for (let i = 0; i < md.length; i++) { let v = md[i]; if (last == BLOCK_CLOSE && v == BLOCK_OPEN) { temp.pop(); temp.push(NEWLINE_MERGED); } else { temp.push(v); } last = v; } md = temp; temp = []; last = ""; for (let i = 0; i < md.length; i++) { let v = md[i]; if (last == NEWLINE && (v == NEWLINE_MERGED || v == BLOCK_CLOSE)) { // Skip it } else { temp.push(v); } last = v; } md = temp; // NEW!!! temp = []; last = ""; for (let i = 0; i < md.length; i++) { let v = md[i]; if (last == NEWLINE && (v == NEWLINE_MERGED || v == BLOCK_OPEN)) { // Skip it } else { temp.push(v); } last = v; } md = temp; if (md.length > 2) { if (md[md.length - 2] == NEWLINE_MERGED && md[md.length - 1] == NEWLINE) { md.pop(); } } let output = ''; let previous = ''; let start = true; for (let i = 0; i < md.length; i++) { let v = md[i]; let add = ''; if (v == BLOCK_CLOSE || v == BLOCK_OPEN || v == NEWLINE || v == NEWLINE_MERGED) { add = "\n"; } else if (v == SPACE) { if (previous == SPACE || previous == "\n" || start) { continue; // skip } else { add = " "; } } else { add = v; } start = false; output += add; previous = add; } if (!output.trim().length) return ''; // To simplify the result, we only allow up to one empty line between blocks of text const mergeMultipleNewLines = function(lines) { let output = []; let newlineCount = 0; for (let i = 0; i < lines.length; i++) { const line = lines[i]; if (!line.trim()) { newlineCount++; } else { newlineCount = 0; } if (newlineCount >= 2) continue; output.push(line); } return output; } let lines = output.replace(/\\r/g, '').split('\n'); lines = formatMdLayout(lines) lines = mergeMultipleNewLines(lines); return lines.join('\n'); } // While the processMdArrayNewLines() function adds newlines in a way that's technically correct, the resulting Markdown can look messy. // This is because while a "block" element should be surrounded by newlines, in practice, some should be surrounded by TWO new lines, while // others by only ONE. // // For instance, this: // //
Some long paragraph
And another one
And the last paragraph
// // should result in this: // // Some long paragraph // // And another one // // And the last paragraph // // So in one case, one newline between tags, and in another two newlines. In HTML this would be done via CSS, but in Markdown we need // to add new lines. It's also important to get these newlines right because two blocks of text next to each others might be renderered // differently than if there's a newlines between them. So the function below parses the almost final MD and add new lines depending // on various rules. const isHeading = function(line) { return !!line.match(/^#+\s/); } const isListItem = function(line) { return line && line.trim().indexOf('- ') === 0; } const isCodeLine = function(line) { return line && line.indexOf('\t') === 0; } const isTableLine = function(line) { return line.indexOf('| ') === 0; } const isPlainParagraph = function(line) { // Note: if a line is no longer than 80 characters, we don't consider it's a paragraph, which // means no newlines will be added before or after. This is to handle text that has been // written with "hard" new lines. if (!line || line.length < 80) return false; if (isListItem(line)) return false; if (isHeading(line)) return false; if (isCodeLine(line)) return false; if (isTableLine(line)) return false; return true; } function formatMdLayout(lines) { let previous = ''; let newLines = []; for (let i = 0; i < lines.length; i++) { const line = lines[i]; // Add a new line at the end of a list of items if (isListItem(previous) && line && !isListItem(line)) { newLines.push(''); // Add a new line at the beginning of a list of items } else if (isListItem(line) && previous && !isListItem(previous)) { newLines.push(''); // Add a new line before a heading } else if (isHeading(line) && previous) { newLines.push(''); // Add a new line after a heading } else if (isHeading(previous) && line) { newLines.push(''); } else if (isCodeLine(line) && !isCodeLine(previous)) { newLines.push(''); } else if (!isCodeLine(line) && isCodeLine(previous)) { newLines.push(''); } else if (isTableLine(line) && !isTableLine(previous)) { newLines.push(''); } else if (!isTableLine(line) && isTableLine(previous)) { newLines.push(''); // Add a new line at beginning of paragraph } else if (isPlainParagraph(line) && previous) { newLines.push(''); // Add a new line at end of paragraph } else if (isPlainParagraph(previous) && line) { newLines.push(''); } newLines.push(line); previous = newLines[newLines.length - 1]; } return newLines; } function isWhiteSpace(c) { return c == '\n' || c == '\r' || c == '\v' || c == '\f' || c == '\t' || c == ' '; } // Like QString::simpified(), except that it preserves non-breaking spaces (which // Evernote uses for identation, etc.) function simplifyString(s) { let output = ''; let previousWhite = false; for (let i = 0; i < s.length; i++) { let c = s[i]; let isWhite = isWhiteSpace(c); if (previousWhite && isWhite) { // skip } else { output += c; } previousWhite = isWhite; } while (output.length && isWhiteSpace(output[0])) output = output.substr(1); while (output.length && isWhiteSpace(output[output.length - 1])) output = output.substr(0, output.length - 1); return output; } function collapseWhiteSpaceAndAppend(lines, state, text) { if (state.inCode.length) { lines.push(text); } else { // Remove all \n and \r from the left and right of the text while (text.length && (text[0] == "\n" || text[0] == "\r")) text = text.substr(1); while (text.length && (text[text.length - 1] == "\n" || text[text.length - 1] == "\r")) text = text.substr(0, text.length - 1); // Collapse all white spaces to just one. If there are spaces to the left and right of the string // also collapse them to just one space. let spaceLeft = text.length && text[0] == ' '; let spaceRight = text.length && text[text.length - 1] == ' '; text = simplifyString(text); if (!spaceLeft && !spaceRight && text == "") return lines; if (state.inQuote) { // Add a ">" at the beginning of the block then at the beginning of each lines. So it turns this: // "my quote\nsecond line" into this => "> my quote\n> second line" lines.push('> '); if (lines.indexOf('\r') >= 0) { text = text.replace(/\n\r/g, '\n\r> '); } else { text = text.replace(/\n/g, '\n> '); } } if (spaceLeft) lines.push(SPACE); lines.push(text); if (spaceRight) lines.push(SPACE); } return lines; } const imageMimeTypes = ["image/cgm", "image/fits", "image/g3fax", "image/gif", "image/ief", "image/jp2", "image/jpeg", "image/jpm", "image/jpx", "image/naplps", "image/png", "image/prs.btif", "image/prs.pti", "image/t38", "image/tiff", "image/tiff-fx", "image/vnd.adobe.photoshop", "image/vnd.cns.inf2", "image/vnd.djvu", "image/vnd.dwg", "image/vnd.dxf", "image/vnd.fastbidsheet", "image/vnd.fpx", "image/vnd.fst", "image/vnd.fujixerox.edmics-mmr", "image/vnd.fujixerox.edmics-rlc", "image/vnd.globalgraphics.pgb", "image/vnd.microsoft.icon", "image/vnd.mix", "image/vnd.ms-modi", "image/vnd.net-fpx", "image/vnd.sealed.png", "image/vnd.sealedmedia.softseal.gif", "image/vnd.sealedmedia.softseal.jpg", "image/vnd.svf", "image/vnd.wap.wbmp", "image/vnd.xiff"]; function isImageMimeType(m) { return imageMimeTypes.indexOf(m) >= 0; } function tagAttributeToMdText(attr) { // HTML attributes may contain newlines so remove them. // https://github.com/laurent22/joplin/issues/1583 if (!attr) return ''; attr = attr.replace(/[\n\r]+/g, ' '); attr = attr.replace(/\]/g, '\\]'); return attr; } function addResourceTag(lines, resource, alt = "") { // Note: refactor to use Resource.markdownTag let tagAlt = alt == "" ? resource.alt : alt; if (!tagAlt) tagAlt = ''; tagAlt = tagAttributeToMdText(tagAlt); if (isImageMimeType(resource.mime)) { lines.push("!["); lines.push(tagAlt); lines.push("](:/" + resource.id + ")"); } else { lines.push("["); lines.push(tagAlt); lines.push("](:/" + resource.id + ")"); } return lines; } function isBlockTag(n) { return ["div", "p", "dl", "dd", 'dt', "center", 'address'].indexOf(n) >= 0; } function isStrongTag(n) { return n == "strong" || n == "b" || n == 'big'; } function isStrikeTag(n) { return n == "strike" || n == "s" || n == 'del'; } function isEmTag(n) { return n == "em" || n == "i" || n == "u"; } function isAnchor(n) { return n == "a"; } function isIgnoredEndTag(n) { return ["en-note", "en-todo", "span", "body", "html", "font", "br", 'hr', 'tbody', 'sup', 'img', 'abbr', 'cite', 'thead', 'small', 'tt', 'sub', 'colgroup', 'col', 'ins', 'caption', 'var', 'map', 'area'].indexOf(n) >= 0; } function isListTag(n) { return n == "ol" || n == "ul"; } // Elements that don't require any special treatment beside adding a newline character function isNewLineOnlyEndTag(n) { return ["div", "p", "li", "h1", "h2", "h3", "h4", "h5", 'h6', "dl", "dd", 'dt', "center", 'address'].indexOf(n) >= 0; } function isInlineCodeTag(n) { return ['samp', 'kbd'].indexOf(n) >= 0; } function isNewLineBlock(s) { return s == BLOCK_OPEN || s == BLOCK_CLOSE; } function xmlNodeText(xmlNode) { if (!xmlNode || !xmlNode.length) return ''; return xmlNode[0]; } function attributeToLowerCase(node) { if (!node.attributes) return {}; let output = {}; for (let n in node.attributes) { if (!node.attributes.hasOwnProperty(n)) continue; output[n.toLowerCase()] = node.attributes[n]; } return output; } function enexXmlToMdArray(stream, resources) { let remainingResources = resources.slice(); const removeRemainingResource = (id) => { for (let i = 0; i < remainingResources.length; i++) { const r = remainingResources[i]; if (r.id === id) { remainingResources.splice(i, 1); } } } return new Promise((resolve, reject) => { let state = { inCode: [], inPre: false, inQuote: false, lists: [], anchorAttributes: [], }; let options = {}; let strict = false; var saxStream = require('sax').createStream(strict, options) let section = { type: 'text', lines: [], parent: null, }; saxStream.on('error', function(e) { console.warn(e); //reject(e); }) const unwrapInnerText = text => { const lines = text.split('\n'); let output = ''; for (let i = 0; i < lines.length; i++) { const line = lines[i]; const nextLine = i < lines.length - 1 ? lines[i+1] : ''; if (!line) { output += '\n'; continue; } if (nextLine) { output += line + ' '; } else { output += line; } } return output; } saxStream.on('text', function(text) { if (['table', 'tr', 'tbody'].indexOf(section.type) >= 0) return; text = !state.inPre ? unwrapInnerText(text) : text; section.lines = collapseWhiteSpaceAndAppend(section.lines, state, text); }) saxStream.on('opentag', function(node) { const nodeAttributes = attributeToLowerCase(node); let n = node.name.toLowerCase(); const currentList = state.lists && state.lists.length ? state.lists[state.lists.length - 1] : null; // Kind of a hack: If we are inside a list, at the beginning of an item (when a "- " or "1. " has been added // but no other text yet), if the current tag is eg. a
// elements by ) and only the inner ones, those that don't contain any other tables, are rendered as actual tables. This is generally
// the required behaviour since the outer tables are usually for layout and the inner ones are the content.
function drawTable(table) {
// | First Header | Second Header |
// | ------------- | ------------- |
// | Content Cell | Content Cell |
// | Content Cell | Content Cell |
// There must be at least 3 dashes separating each header cell.
// https://gist.github.com/IanWang/28965e13cdafdef4e11dc91f578d160d#tables
const flatRender = tableHasSubTables(table); // Render the table has regular text
let lines = [];
lines.push(BLOCK_OPEN);
let headerDone = false;
for (let trIndex = 0; trIndex < table.lines.length; trIndex++) {
const tr = table.lines[trIndex];
const isHeader = tr.isHeader;
let line = [];
let headerLine = [];
let emptyHeader = null;
for (let tdIndex = 0; tdIndex < tr.lines.length; tdIndex++) {
const td = tr.lines[tdIndex];
if (flatRender) {
line.push(BLOCK_OPEN);
let currentCells = [];
const renderCurrentCells = () => {
if (!currentCells.length) return;
const cellText = processMdArrayNewLines(currentCells);
line.push(cellText);
currentCells = [];
}
// In here, recursively render the tables
for (let i = 0; i < td.lines.length; i++) {
const c = td.lines[i];
if (typeof c === 'object' && ['table', 'td', 'tr', 'th'].indexOf(c.type) >= 0) { // This is a table
renderCurrentCells();
currentCells = currentCells.concat(drawTable(c));
} else { // This is plain text
currentCells.push(c);
}
}
renderCurrentCells();
line.push(BLOCK_CLOSE);
} else { // Regular table rendering
// A cell in a Markdown table cannot have actual new lines so replace
// them with , which are supported by the markdown renderers. let cellText = processMdArrayNewLines(td.lines, true) let lines = cellText.split('\n'); lines = postProcessMarkdown(lines); cellText = lines.join('\n').replace(/\n+/g, " "); // Inside tables cells, "|" needs to be escaped cellText = cellText.replace(/\|/g, "\\|"); // Previously the width of the cell was as big as the content since it looks nicer, however that often doesn't work // since the content can be very long, resulting in unreadable markdown. So no solution is perfect but making it a // width of 3 is a bit better. Note that 3 is the minimum width of a cell - below this, it won't be rendered by // markdown parsers. const width = 3; line.push(stringPadding(cellText, width, ' ', stringPadding.RIGHT)); if (!headerDone) { if (!isHeader) { if (!emptyHeader) emptyHeader = []; let h = stringPadding(' ', width, ' ', stringPadding.RIGHT); emptyHeader.push(h); } headerLine.push('-'.repeat(width)); } } } if (flatRender) { headerDone = true; lines.push(BLOCK_OPEN); lines = lines.concat(line); lines.push(BLOCK_CLOSE); } else { if (emptyHeader) { lines.push('| ' + emptyHeader.join(' | ') + ' |'); lines.push('| ' + headerLine.join(' | ') + ' |'); headerDone = true; } lines.push('| ' + line.join(' | ') + ' |'); if (!headerDone) { lines.push('| ' + headerLine.join(' | ') + ' |'); headerDone = true; } } } lines.push(BLOCK_CLOSE); return flatRender ? lines : lines.join('<<<<:D>>>>' + NEWLINE + '<<<<:D>>>>').split('<<<<:D>>>>'); } function postProcessMarkdown(lines) { // After importing HTML, the resulting Markdown often has empty lines at the beginning and end due to // block start/end or elements that were ignored, etc. If these white spaces were intended it's not really // possible to detect it, so simply trim them all so that the result is more deterministic and can be // easily unit tested. const trimEmptyLines = function(lines) { while (lines.length) { if (!lines[0].trim()) { lines.splice(0, 1); } else { break; } } while (lines.length) { if (!lines[lines.length - 1].trim()) { lines.pop(); } else { break; } } return lines; } function cleanUpSpaces(lines) { const output = []; for (let i = 0; i < lines.length; i++) { let line = lines[i]; if (line.length) { // eg. " - Some list item" => " - Some list item" // Note that spaces before the "-" are preserved line = line.replace(/^(\s+|)-\s+/, '$1- ') // eg "Some text " => "Some text" line = line.replace(/^(.*?)\s+$/, '$1') } output.push(line); } return output; } lines = trimEmptyLines(lines) lines = cleanUpSpaces(lines) return lines; } async function enexXmlToMd(xmlString, resources, options = {}) { const stream = stringToStream(xmlString); let result = await enexXmlToMdArray(stream, resources, options); let mdLines = []; for (let i = 0; i < result.content.lines.length; i++) { let line = result.content.lines[i]; if (typeof line === 'object' && line.type === 'table') { // A table const table = line; const tableLines = drawTable(table); mdLines = mdLines.concat(tableLines); } else if (typeof line === 'object' && line.type === 'code') { mdLines = mdLines.concat(line.lines); } else if (typeof line === 'object') { console.warn('Unhandled object type:', line); mdLines = mdLines.concat(line.lines); } else { // an actual line mdLines.push(line); } } let firstAttachment = true; for (let i = 0; i < result.resources.length; i++) { let r = result.resources[i]; if (firstAttachment) mdLines.push(NEWLINE); mdLines.push(NEWLINE); mdLines = addResourceTag(mdLines, r, r.filename); firstAttachment = false; } let output = processMdArrayNewLines(mdLines).split('\n') output = postProcessMarkdown(output); return output.join('\n'); } module.exports = { enexXmlToMd, processMdArrayNewLines, NEWLINE, addResourceTag }; |