(function (global, factory) { typeof exports === 'object' && typeof module !== 'undefined' ? module.exports = factory() : typeof define === 'function' && define.amd ? define(factory) : (global = typeof globalThis !== 'undefined' ? globalThis : global || self, global.doc = factory()); }(this, (function () { 'use strict'; /** * @param {Doc[]} parts * @returns Doc */ function concat(parts) { // access the internals of a document directly. // if(parts.length === 1) { // // If it's a single document, no need to concat it. // return parts[0]; // } return { type: "concat", parts }; } /** * @param {Doc} contents * @returns Doc */ function indent(contents) { return { type: "indent", contents }; } /** * @param {number | string} n * @param {Doc} contents * @returns Doc */ function align(n, contents) { return { type: "align", contents, n }; } /** * @param {Doc} contents * @param {object} [opts] - TBD ??? * @returns Doc */ function group(contents, opts) { opts = opts || {}; return { type: "group", id: opts.id, contents, break: !!opts.shouldBreak, expandedStates: opts.expandedStates }; } /** * @param {Doc} contents * @returns Doc */ function dedentToRoot(contents) { return align(-Infinity, contents); } /** * @param {Doc} contents * @returns Doc */ function markAsRoot(contents) { // @ts-ignore - TBD ???: return align({ type: "root" }, contents); } /** * @param {Doc} contents * @returns Doc */ function dedent(contents) { return align(-1, contents); } /** * @param {Doc[]} states * @param {object} [opts] - TBD ??? * @returns Doc */ function conditionalGroup(states, opts) { return group(states[0], Object.assign({}, opts, { expandedStates: states })); } /** * @param {Doc[]} parts * @returns Doc */ function fill(parts) { return { type: "fill", parts }; } /** * @param {Doc} [breakContents] * @param {Doc} [flatContents] * @param {object} [opts] - TBD ??? * @returns Doc */ function ifBreak(breakContents, flatContents, opts) { opts = opts || {}; return { type: "if-break", breakContents, flatContents, groupId: opts.groupId }; } /** * @param {Doc} contents * @returns Doc */ function lineSuffix(contents) { return { type: "line-suffix", contents }; } const lineSuffixBoundary = { type: "line-suffix-boundary" }; const breakParent = { type: "break-parent" }; const trim = { type: "trim" }; const line = { type: "line" }; const softline = { type: "line", soft: true }; const hardline = concat([{ type: "line", hard: true }, breakParent]); const literalline = concat([{ type: "line", hard: true, literal: true }, breakParent]); const cursor = { type: "cursor", placeholder: Symbol("cursor") }; /** * @param {Doc} sep * @param {Doc[]} arr * @returns Doc */ function join(sep, arr) { const res = []; for (let i = 0; i < arr.length; i++) { if (i !== 0) { res.push(sep); } res.push(arr[i]); } return concat(res); } /** * @param {Doc} doc * @param {number} size * @param {number} tabWidth */ function addAlignmentToDoc(doc, size, tabWidth) { let aligned = doc; if (size > 0) { // Use indent to add tabs for all the levels of tabs we need for (let i = 0; i < Math.floor(size / tabWidth); ++i) { aligned = indent(aligned); } // Use align for all the spaces that are needed aligned = align(size % tabWidth, aligned); // size is absolute from 0 and not relative to the current // indentation, so we use -Infinity to reset the indentation to 0 aligned = align(-Infinity, aligned); } return aligned; } var docBuilders = { concat, join, line, softline, hardline, literalline, group, conditionalGroup, fill, lineSuffix, lineSuffixBoundary, cursor, breakParent, ifBreak, trim, indent, align, addAlignmentToDoc, markAsRoot, dedentToRoot, dedent }; var ansiRegex = ({ onlyFirst = false } = {}) => { const pattern = ['[\\u001B\\u009B][[\\]()#;?]*(?:(?:(?:[a-zA-Z\\d]*(?:;[-a-zA-Z\\d\\/#&.:=?%@~_]*)*)?\\u0007)', '(?:(?:\\d{1,4}(?:;\\d{0,4})*)?[\\dA-PR-TZcf-ntqry=><~]))'].join('|'); return new RegExp(pattern, onlyFirst ? undefined : 'g'); }; var stripAnsi = string => typeof string === 'string' ? string.replace(ansiRegex(), '') : string; /* eslint-disable yoda */ const isFullwidthCodePoint = codePoint => { if (Number.isNaN(codePoint)) { return false; } // Code points are derived from: // http://www.unix.org/Public/UNIDATA/EastAsianWidth.txt if (codePoint >= 0x1100 && (codePoint <= 0x115F || // Hangul Jamo codePoint === 0x2329 || // LEFT-POINTING ANGLE BRACKET codePoint === 0x232A || // RIGHT-POINTING ANGLE BRACKET // CJK Radicals Supplement .. Enclosed CJK Letters and Months 0x2E80 <= codePoint && codePoint <= 0x3247 && codePoint !== 0x303F || // Enclosed CJK Letters and Months .. CJK Unified Ideographs Extension A 0x3250 <= codePoint && codePoint <= 0x4DBF || // CJK Unified Ideographs .. Yi Radicals 0x4E00 <= codePoint && codePoint <= 0xA4C6 || // Hangul Jamo Extended-A 0xA960 <= codePoint && codePoint <= 0xA97C || // Hangul Syllables 0xAC00 <= codePoint && codePoint <= 0xD7A3 || // CJK Compatibility Ideographs 0xF900 <= codePoint && codePoint <= 0xFAFF || // Vertical Forms 0xFE10 <= codePoint && codePoint <= 0xFE19 || // CJK Compatibility Forms .. Small Form Variants 0xFE30 <= codePoint && codePoint <= 0xFE6B || // Halfwidth and Fullwidth Forms 0xFF01 <= codePoint && codePoint <= 0xFF60 || 0xFFE0 <= codePoint && codePoint <= 0xFFE6 || // Kana Supplement 0x1B000 <= codePoint && codePoint <= 0x1B001 || // Enclosed Ideographic Supplement 0x1F200 <= codePoint && codePoint <= 0x1F251 || // CJK Unified Ideographs Extension B .. Tertiary Ideographic Plane 0x20000 <= codePoint && codePoint <= 0x3FFFD)) { return true; } return false; }; var isFullwidthCodePoint_1 = isFullwidthCodePoint; var _default = isFullwidthCodePoint; isFullwidthCodePoint_1.default = _default; var emojiRegex = function emojiRegex() { // https://mths.be/emoji return /\uD83C\uDFF4\uDB40\uDC67\uDB40\uDC62(?:\uDB40\uDC65\uDB40\uDC6E\uDB40\uDC67|\uDB40\uDC73\uDB40\uDC63\uDB40\uDC74|\uDB40\uDC77\uDB40\uDC6C\uDB40\uDC73)\uDB40\uDC7F|\uD83D\uDC68(?:\uD83C\uDFFC\u200D(?:\uD83E\uDD1D\u200D\uD83D\uDC68\uD83C\uDFFB|\uD83C[\uDF3E\uDF73\uDF93\uDFA4\uDFA8\uDFEB\uDFED]|\uD83D[\uDCBB\uDCBC\uDD27\uDD2C\uDE80\uDE92]|\uD83E[\uDDAF-\uDDB3\uDDBC\uDDBD])|\uD83C\uDFFF\u200D(?:\uD83E\uDD1D\u200D\uD83D\uDC68(?:\uD83C[\uDFFB-\uDFFE])|\uD83C[\uDF3E\uDF73\uDF93\uDFA4\uDFA8\uDFEB\uDFED]|\uD83D[\uDCBB\uDCBC\uDD27\uDD2C\uDE80\uDE92]|\uD83E[\uDDAF-\uDDB3\uDDBC\uDDBD])|\uD83C\uDFFE\u200D(?:\uD83E\uDD1D\u200D\uD83D\uDC68(?:\uD83C[\uDFFB-\uDFFD])|\uD83C[\uDF3E\uDF73\uDF93\uDFA4\uDFA8\uDFEB\uDFED]|\uD83D[\uDCBB\uDCBC\uDD27\uDD2C\uDE80\uDE92]|\uD83E[\uDDAF-\uDDB3\uDDBC\uDDBD])|\uD83C\uDFFD\u200D(?:\uD83E\uDD1D\u200D\uD83D\uDC68(?:\uD83C[\uDFFB\uDFFC])|\uD83C[\uDF3E\uDF73\uDF93\uDFA4\uDFA8\uDFEB\uDFED]|\uD83D[\uDCBB\uDCBC\uDD27\uDD2C\uDE80\uDE92]|\uD83E[\uDDAF-\uDDB3\uDDBC\uDDBD])|\u200D(?:\u2764\uFE0F\u200D(?:\uD83D\uDC8B\u200D)?\uD83D\uDC68|(?:\uD83D[\uDC68\uDC69])\u200D(?:\uD83D\uDC66\u200D\uD83D\uDC66|\uD83D\uDC67\u200D(?:\uD83D[\uDC66\uDC67]))|\uD83D\uDC66\u200D\uD83D\uDC66|\uD83D\uDC67\u200D(?:\uD83D[\uDC66\uDC67])|(?:\uD83D[\uDC68\uDC69])\u200D(?:\uD83D[\uDC66\uDC67])|[\u2695\u2696\u2708]\uFE0F|\uD83D[\uDC66\uDC67]|\uD83C[\uDF3E\uDF73\uDF93\uDFA4\uDFA8\uDFEB\uDFED]|\uD83D[\uDCBB\uDCBC\uDD27\uDD2C\uDE80\uDE92]|\uD83E[\uDDAF-\uDDB3\uDDBC\uDDBD])|(?:\uD83C\uDFFB\u200D[\u2695\u2696\u2708]|\uD83C\uDFFF\u200D[\u2695\u2696\u2708]|\uD83C\uDFFE\u200D[\u2695\u2696\u2708]|\uD83C\uDFFD\u200D[\u2695\u2696\u2708]|\uD83C\uDFFC\u200D[\u2695\u2696\u2708])\uFE0F|\uD83C\uDFFB\u200D(?:\uD83C[\uDF3E\uDF73\uDF93\uDFA4\uDFA8\uDFEB\uDFED]|\uD83D[\uDCBB\uDCBC\uDD27\uDD2C\uDE80\uDE92]|\uD83E[\uDDAF-\uDDB3\uDDBC\uDDBD])|\uD83C[\uDFFB-\uDFFF])|(?:\uD83E\uDDD1\uD83C\uDFFB\u200D\uD83E\uDD1D\u200D\uD83E\uDDD1|\uD83D\uDC69\uD83C\uDFFC\u200D\uD83E\uDD1D\u200D\uD83D\uDC69)\uD83C\uDFFB|\uD83E\uDDD1(?:\uD83C\uDFFF\u200D\uD83E\uDD1D\u200D\uD83E\uDDD1(?:\uD83C[\uDFFB-\uDFFF])|\u200D\uD83E\uDD1D\u200D\uD83E\uDDD1)|(?:\uD83E\uDDD1\uD83C\uDFFE\u200D\uD83E\uDD1D\u200D\uD83E\uDDD1|\uD83D\uDC69\uD83C\uDFFF\u200D\uD83E\uDD1D\u200D(?:\uD83D[\uDC68\uDC69]))(?:\uD83C[\uDFFB-\uDFFE])|(?:\uD83E\uDDD1\uD83C\uDFFC\u200D\uD83E\uDD1D\u200D\uD83E\uDDD1|\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])|\uD83C[\uDF3E\uDF73\uDF93\uDFA4\uDFA8\uDFEB\uDFED]|\uD83D[\uDCBB\uDCBC\uDD27\uDD2C\uDE80\uDE92]|\uD83E[\uDDAF-\uDDB3\uDDBC\uDDBD])|\uD83C\uDFFC\u200D(?:\uD83E\uDD1D\u200D\uD83D\uDC68(?:\uD83C[\uDFFB\uDFFD-\uDFFF])|\uD83C[\uDF3E\uDF73\uDF93\uDFA4\uDFA8\uDFEB\uDFED]|\uD83D[\uDCBB\uDCBC\uDD27\uDD2C\uDE80\uDE92]|\uD83E[\uDDAF-\uDDB3\uDDBC\uDDBD])|\uD83C\uDFFB\u200D(?:\uD83E\uDD1D\u200D\uD83D\uDC68(?:\uD83C[\uDFFC-\uDFFF])|\uD83C[\uDF3E\uDF73\uDF93\uDFA4\uDFA8\uDFEB\uDFED]|\uD83D[\uDCBB\uDCBC\uDD27\uDD2C\uDE80\uDE92]|\uD83E[\uDDAF-\uDDB3\uDDBC\uDDBD])|\uD83C\uDFFD\u200D(?:\uD83E\uDD1D\u200D\uD83D\uDC68(?:\uD83C[\uDFFB\uDFFC\uDFFE\uDFFF])|\uD83C[\uDF3E\uDF73\uDF93\uDFA4\uDFA8\uDFEB\uDFED]|\uD83D[\uDCBB\uDCBC\uDD27\uDD2C\uDE80\uDE92]|\uD83E[\uDDAF-\uDDB3\uDDBC\uDDBD])|\u200D(?:\u2764\uFE0F\u200D(?:\uD83D\uDC8B\u200D(?:\uD83D[\uDC68\uDC69])|\uD83D[\uDC68\uDC69])|\uD83C[\uDF3E\uDF73\uDF93\uDFA4\uDFA8\uDFEB\uDFED]|\uD83D[\uDCBB\uDCBC\uDD27\uDD2C\uDE80\uDE92]|\uD83E[\uDDAF-\uDDB3\uDDBC\uDDBD])|\uD83C\uDFFF\u200D(?:\uD83C[\uDF3E\uDF73\uDF93\uDFA4\uDFA8\uDFEB\uDFED]|\uD83D[\uDCBB\uDCBC\uDD27\uDD2C\uDE80\uDE92]|\uD83E[\uDDAF-\uDDB3\uDDBC\uDDBD]))|\uD83D\uDC69\u200D\uD83D\uDC69\u200D(?:\uD83D\uDC66\u200D\uD83D\uDC66|\uD83D\uDC67\u200D(?:\uD83D[\uDC66\uDC67]))|(?:\uD83E\uDDD1\uD83C\uDFFD\u200D\uD83E\uDD1D\u200D\uD83E\uDDD1|\uD83D\uDC69\uD83C\uDFFE\u200D\uD83E\uDD1D\u200D\uD83D\uDC69)(?:\uD83C[\uDFFB-\uDFFD])|\uD83D\uDC69\u200D\uD83D\uDC66\u200D\uD83D\uDC66|\uD83D\uDC69\u200D\uD83D\uDC69\u200D(?:\uD83D[\uDC66\uDC67])|(?:\uD83D\uDC41\uFE0F\u200D\uD83D\uDDE8|\uD83D\uDC69(?:\uD83C\uDFFF\u200D[\u2695\u2696\u2708]|\uD83C\uDFFE\u200D[\u2695\u2696\u2708]|\uD83C\uDFFC\u200D[\u2695\u2696\u2708]|\uD83C\uDFFB\u200D[\u2695\u2696\u2708]|\uD83C\uDFFD\u200D[\u2695\u2696\u2708]|\u200D[\u2695\u2696\u2708])|(?:(?:\u26F9|\uD83C[\uDFCB\uDFCC]|\uD83D\uDD75)\uFE0F|\uD83D\uDC6F|\uD83E[\uDD3C\uDDDE\uDDDF])\u200D[\u2640\u2642]|(?:\u26F9|\uD83C[\uDFCB\uDFCC]|\uD83D\uDD75)(?:\uD83C[\uDFFB-\uDFFF])\u200D[\u2640\u2642]|(?:\uD83C[\uDFC3\uDFC4\uDFCA]|\uD83D[\uDC6E\uDC71\uDC73\uDC77\uDC81\uDC82\uDC86\uDC87\uDE45-\uDE47\uDE4B\uDE4D\uDE4E\uDEA3\uDEB4-\uDEB6]|\uD83E[\uDD26\uDD37-\uDD39\uDD3D\uDD3E\uDDB8\uDDB9\uDDCD-\uDDCF\uDDD6-\uDDDD])(?:(?:\uD83C[\uDFFB-\uDFFF])\u200D[\u2640\u2642]|\u200D[\u2640\u2642])|\uD83C\uDFF4\u200D\u2620)\uFE0F|\uD83D\uDC69\u200D\uD83D\uDC67\u200D(?:\uD83D[\uDC66\uDC67])|\uD83C\uDFF3\uFE0F\u200D\uD83C\uDF08|\uD83D\uDC15\u200D\uD83E\uDDBA|\uD83D\uDC69\u200D\uD83D\uDC66|\uD83D\uDC69\u200D\uD83D\uDC67|\uD83C\uDDFD\uD83C\uDDF0|\uD83C\uDDF4\uD83C\uDDF2|\uD83C\uDDF6\uD83C\uDDE6|[#\*0-9]\uFE0F\u20E3|\uD83C\uDDE7(?:\uD83C[\uDDE6\uDDE7\uDDE9-\uDDEF\uDDF1-\uDDF4\uDDF6-\uDDF9\uDDFB\uDDFC\uDDFE\uDDFF])|\uD83C\uDDF9(?:\uD83C[\uDDE6\uDDE8\uDDE9\uDDEB-\uDDED\uDDEF-\uDDF4\uDDF7\uDDF9\uDDFB\uDDFC\uDDFF])|\uD83C\uDDEA(?:\uD83C[\uDDE6\uDDE8\uDDEA\uDDEC\uDDED\uDDF7-\uDDFA])|\uD83E\uDDD1(?:\uD83C[\uDFFB-\uDFFF])|\uD83C\uDDF7(?:\uD83C[\uDDEA\uDDF4\uDDF8\uDDFA\uDDFC])|\uD83D\uDC69(?:\uD83C[\uDFFB-\uDFFF])|\uD83C\uDDF2(?:\uD83C[\uDDE6\uDDE8-\uDDED\uDDF0-\uDDFF])|\uD83C\uDDE6(?:\uD83C[\uDDE8-\uDDEC\uDDEE\uDDF1\uDDF2\uDDF4\uDDF6-\uDDFA\uDDFC\uDDFD\uDDFF])|\uD83C\uDDF0(?:\uD83C[\uDDEA\uDDEC-\uDDEE\uDDF2\uDDF3\uDDF5\uDDF7\uDDFC\uDDFE\uDDFF])|\uD83C\uDDED(?:\uD83C[\uDDF0\uDDF2\uDDF3\uDDF7\uDDF9\uDDFA])|\uD83C\uDDE9(?:\uD83C[\uDDEA\uDDEC\uDDEF\uDDF0\uDDF2\uDDF4\uDDFF])|\uD83C\uDDFE(?:\uD83C[\uDDEA\uDDF9])|\uD83C\uDDEC(?:\uD83C[\uDDE6\uDDE7\uDDE9-\uDDEE\uDDF1-\uDDF3\uDDF5-\uDDFA\uDDFC\uDDFE])|\uD83C\uDDF8(?:\uD83C[\uDDE6-\uDDEA\uDDEC-\uDDF4\uDDF7-\uDDF9\uDDFB\uDDFD-\uDDFF])|\uD83C\uDDEB(?:\uD83C[\uDDEE-\uDDF0\uDDF2\uDDF4\uDDF7])|\uD83C\uDDF5(?:\uD83C[\uDDE6\uDDEA-\uDDED\uDDF0-\uDDF3\uDDF7-\uDDF9\uDDFC\uDDFE])|\uD83C\uDDFB(?:\uD83C[\uDDE6\uDDE8\uDDEA\uDDEC\uDDEE\uDDF3\uDDFA])|\uD83C\uDDF3(?:\uD83C[\uDDE6\uDDE8\uDDEA-\uDDEC\uDDEE\uDDF1\uDDF4\uDDF5\uDDF7\uDDFA\uDDFF])|\uD83C\uDDE8(?:\uD83C[\uDDE6\uDDE8\uDDE9\uDDEB-\uDDEE\uDDF0-\uDDF5\uDDF7\uDDFA-\uDDFF])|\uD83C\uDDF1(?:\uD83C[\uDDE6-\uDDE8\uDDEE\uDDF0\uDDF7-\uDDFB\uDDFE])|\uD83C\uDDFF(?:\uD83C[\uDDE6\uDDF2\uDDFC])|\uD83C\uDDFC(?:\uD83C[\uDDEB\uDDF8])|\uD83C\uDDFA(?:\uD83C[\uDDE6\uDDEC\uDDF2\uDDF3\uDDF8\uDDFE\uDDFF])|\uD83C\uDDEE(?:\uD83C[\uDDE8-\uDDEA\uDDF1-\uDDF4\uDDF6-\uDDF9])|\uD83C\uDDEF(?:\uD83C[\uDDEA\uDDF2\uDDF4\uDDF5])|(?:\uD83C[\uDFC3\uDFC4\uDFCA]|\uD83D[\uDC6E\uDC71\uDC73\uDC77\uDC81\uDC82\uDC86\uDC87\uDE45-\uDE47\uDE4B\uDE4D\uDE4E\uDEA3\uDEB4-\uDEB6]|\uD83E[\uDD26\uDD37-\uDD39\uDD3D\uDD3E\uDDB8\uDDB9\uDDCD-\uDDCF\uDDD6-\uDDDD])(?:\uD83C[\uDFFB-\uDFFF])|(?:\u26F9|\uD83C[\uDFCB\uDFCC]|\uD83D\uDD75)(?:\uD83C[\uDFFB-\uDFFF])|(?:[\u261D\u270A-\u270D]|\uD83C[\uDF85\uDFC2\uDFC7]|\uD83D[\uDC42\uDC43\uDC46-\uDC50\uDC66\uDC67\uDC6B-\uDC6D\uDC70\uDC72\uDC74-\uDC76\uDC78\uDC7C\uDC83\uDC85\uDCAA\uDD74\uDD7A\uDD90\uDD95\uDD96\uDE4C\uDE4F\uDEC0\uDECC]|\uD83E[\uDD0F\uDD18-\uDD1C\uDD1E\uDD1F\uDD30-\uDD36\uDDB5\uDDB6\uDDBB\uDDD2-\uDDD5])(?:\uD83C[\uDFFB-\uDFFF])|(?:[\u231A\u231B\u23E9-\u23EC\u23F0\u23F3\u25FD\u25FE\u2614\u2615\u2648-\u2653\u267F\u2693\u26A1\u26AA\u26AB\u26BD\u26BE\u26C4\u26C5\u26CE\u26D4\u26EA\u26F2\u26F3\u26F5\u26FA\u26FD\u2705\u270A\u270B\u2728\u274C\u274E\u2753-\u2755\u2757\u2795-\u2797\u27B0\u27BF\u2B1B\u2B1C\u2B50\u2B55]|\uD83C[\uDC04\uDCCF\uDD8E\uDD91-\uDD9A\uDDE6-\uDDFF\uDE01\uDE1A\uDE2F\uDE32-\uDE36\uDE38-\uDE3A\uDE50\uDE51\uDF00-\uDF20\uDF2D-\uDF35\uDF37-\uDF7C\uDF7E-\uDF93\uDFA0-\uDFCA\uDFCF-\uDFD3\uDFE0-\uDFF0\uDFF4\uDFF8-\uDFFF]|\uD83D[\uDC00-\uDC3E\uDC40\uDC42-\uDCFC\uDCFF-\uDD3D\uDD4B-\uDD4E\uDD50-\uDD67\uDD7A\uDD95\uDD96\uDDA4\uDDFB-\uDE4F\uDE80-\uDEC5\uDECC\uDED0-\uDED2\uDED5\uDEEB\uDEEC\uDEF4-\uDEFA\uDFE0-\uDFEB]|\uD83E[\uDD0D-\uDD3A\uDD3C-\uDD45\uDD47-\uDD71\uDD73-\uDD76\uDD7A-\uDDA2\uDDA5-\uDDAA\uDDAE-\uDDCA\uDDCD-\uDDFF\uDE70-\uDE73\uDE78-\uDE7A\uDE80-\uDE82\uDE90-\uDE95])|(?:[#\*0-9\xA9\xAE\u203C\u2049\u2122\u2139\u2194-\u2199\u21A9\u21AA\u231A\u231B\u2328\u23CF\u23E9-\u23F3\u23F8-\u23FA\u24C2\u25AA\u25AB\u25B6\u25C0\u25FB-\u25FE\u2600-\u2604\u260E\u2611\u2614\u2615\u2618\u261D\u2620\u2622\u2623\u2626\u262A\u262E\u262F\u2638-\u263A\u2640\u2642\u2648-\u2653\u265F\u2660\u2663\u2665\u2666\u2668\u267B\u267E\u267F\u2692-\u2697\u2699\u269B\u269C\u26A0\u26A1\u26AA\u26AB\u26B0\u26B1\u26BD\u26BE\u26C4\u26C5\u26C8\u26CE\u26CF\u26D1\u26D3\u26D4\u26E9\u26EA\u26F0-\u26F5\u26F7-\u26FA\u26FD\u2702\u2705\u2708-\u270D\u270F\u2712\u2714\u2716\u271D\u2721\u2728\u2733\u2734\u2744\u2747\u274C\u274E\u2753-\u2755\u2757\u2763\u2764\u2795-\u2797\u27A1\u27B0\u27BF\u2934\u2935\u2B05-\u2B07\u2B1B\u2B1C\u2B50\u2B55\u3030\u303D\u3297\u3299]|\uD83C[\uDC04\uDCCF\uDD70\uDD71\uDD7E\uDD7F\uDD8E\uDD91-\uDD9A\uDDE6-\uDDFF\uDE01\uDE02\uDE1A\uDE2F\uDE32-\uDE3A\uDE50\uDE51\uDF00-\uDF21\uDF24-\uDF93\uDF96\uDF97\uDF99-\uDF9B\uDF9E-\uDFF0\uDFF3-\uDFF5\uDFF7-\uDFFF]|\uD83D[\uDC00-\uDCFD\uDCFF-\uDD3D\uDD49-\uDD4E\uDD50-\uDD67\uDD6F\uDD70\uDD73-\uDD7A\uDD87\uDD8A-\uDD8D\uDD90\uDD95\uDD96\uDDA4\uDDA5\uDDA8\uDDB1\uDDB2\uDDBC\uDDC2-\uDDC4\uDDD1-\uDDD3\uDDDC-\uDDDE\uDDE1\uDDE3\uDDE8\uDDEF\uDDF3\uDDFA-\uDE4F\uDE80-\uDEC5\uDECB-\uDED2\uDED5\uDEE0-\uDEE5\uDEE9\uDEEB\uDEEC\uDEF0\uDEF3-\uDEFA\uDFE0-\uDFEB]|\uD83E[\uDD0D-\uDD3A\uDD3C-\uDD45\uDD47-\uDD71\uDD73-\uDD76\uDD7A-\uDDA2\uDDA5-\uDDAA\uDDAE-\uDDCA\uDDCD-\uDDFF\uDE70-\uDE73\uDE78-\uDE7A\uDE80-\uDE82\uDE90-\uDE95])\uFE0F|(?:[\u261D\u26F9\u270A-\u270D]|\uD83C[\uDF85\uDFC2-\uDFC4\uDFC7\uDFCA-\uDFCC]|\uD83D[\uDC42\uDC43\uDC46-\uDC50\uDC66-\uDC78\uDC7C\uDC81-\uDC83\uDC85-\uDC87\uDC8F\uDC91\uDCAA\uDD74\uDD75\uDD7A\uDD90\uDD95\uDD96\uDE45-\uDE47\uDE4B-\uDE4F\uDEA3\uDEB4-\uDEB6\uDEC0\uDECC]|\uD83E[\uDD0F\uDD18-\uDD1F\uDD26\uDD30-\uDD39\uDD3C-\uDD3E\uDDB5\uDDB6\uDDB8\uDDB9\uDDBB\uDDCD-\uDDCF\uDDD1-\uDDDD])/g; }; const stringWidth = string => { string = string.replace(emojiRegex(), ' '); if (typeof string !== 'string' || string.length === 0) { return 0; } string = stripAnsi(string); let width = 0; for (let i = 0; i < string.length; i++) { const code = string.codePointAt(i); // Ignore control characters if (code <= 0x1F || code >= 0x7F && code <= 0x9F) { continue; } // Ignore combining characters if (code >= 0x300 && code <= 0x36F) { continue; } // Surrogates if (code > 0xFFFF) { i++; } width += isFullwidthCodePoint_1(code) ? 2 : 1; } return width; }; var stringWidth_1 = stringWidth; // TODO: remove this in the next major version var _default$1 = stringWidth; stringWidth_1.default = _default$1; var escapeStringRegexp = string => { if (typeof string !== 'string') { throw new TypeError('Expected a string'); } // Escape characters with special meaning either inside or outside character sets. // Use a simple backslash escape when it’s always valid, and a \unnnn escape when the simpler form would be disallowed by Unicode patterns’ stricter grammar. return string.replace(/[|\\{}()[\]^$+*?.]/g, '\\$&').replace(/-/g, '\\x2d'); }; var getLast = arr => arr[arr.length - 1]; function _objectWithoutPropertiesLoose(source, excluded) { if (source == null) return {}; var target = {}; var sourceKeys = Object.keys(source); var key, i; for (i = 0; i < sourceKeys.length; i++) { key = sourceKeys[i]; if (excluded.indexOf(key) >= 0) continue; target[key] = source[key]; } return target; } function _taggedTemplateLiteral(strings, raw) { if (!raw) { raw = strings.slice(0); } return Object.freeze(Object.defineProperties(strings, { raw: { value: Object.freeze(raw) } })); } var global$1 = typeof global !== "undefined" ? global : typeof self !== "undefined" ? self : typeof window !== "undefined" ? window : {}; // based off https://github.com/defunctzombie/node-process/blob/master/browser.js function defaultSetTimout() { throw new Error('setTimeout has not been defined'); } function defaultClearTimeout() { throw new Error('clearTimeout has not been defined'); } var cachedSetTimeout = defaultSetTimout; var cachedClearTimeout = defaultClearTimeout; if (typeof global$1.setTimeout === 'function') { cachedSetTimeout = setTimeout; } if (typeof global$1.clearTimeout === 'function') { cachedClearTimeout = clearTimeout; } function runTimeout(fun) { if (cachedSetTimeout === setTimeout) { //normal enviroments in sane situations return setTimeout(fun, 0); } // if setTimeout wasn't available but was latter defined if ((cachedSetTimeout === defaultSetTimout || !cachedSetTimeout) && setTimeout) { cachedSetTimeout = setTimeout; return setTimeout(fun, 0); } try { // when when somebody has screwed with setTimeout but no I.E. maddness return cachedSetTimeout(fun, 0); } catch (e) { try { // When we are in I.E. but the script has been evaled so I.E. doesn't trust the global object when called normally return cachedSetTimeout.call(null, fun, 0); } catch (e) { // same as above but when it's a version of I.E. that must have the global object for 'this', hopfully our context correct otherwise it will throw a global error return cachedSetTimeout.call(this, fun, 0); } } } function runClearTimeout(marker) { if (cachedClearTimeout === clearTimeout) { //normal enviroments in sane situations return clearTimeout(marker); } // if clearTimeout wasn't available but was latter defined if ((cachedClearTimeout === defaultClearTimeout || !cachedClearTimeout) && clearTimeout) { cachedClearTimeout = clearTimeout; return clearTimeout(marker); } try { // when when somebody has screwed with setTimeout but no I.E. maddness return cachedClearTimeout(marker); } catch (e) { try { // When we are in I.E. but the script has been evaled so I.E. doesn't trust the global object when called normally return cachedClearTimeout.call(null, marker); } catch (e) { // same as above but when it's a version of I.E. that must have the global object for 'this', hopfully our context correct otherwise it will throw a global error. // Some versions of I.E. have different rules for clearTimeout vs setTimeout return cachedClearTimeout.call(this, marker); } } } var queue = []; var draining = false; var currentQueue; var queueIndex = -1; function cleanUpNextTick() { if (!draining || !currentQueue) { return; } draining = false; if (currentQueue.length) { queue = currentQueue.concat(queue); } else { queueIndex = -1; } if (queue.length) { drainQueue(); } } function drainQueue() { if (draining) { return; } var timeout = runTimeout(cleanUpNextTick); draining = true; var len = queue.length; while (len) { currentQueue = queue; queue = []; while (++queueIndex < len) { if (currentQueue) { currentQueue[queueIndex].run(); } } queueIndex = -1; len = queue.length; } currentQueue = null; draining = false; runClearTimeout(timeout); } function nextTick(fun) { var args = new Array(arguments.length - 1); if (arguments.length > 1) { for (var i = 1; i < arguments.length; i++) { args[i - 1] = arguments[i]; } } queue.push(new Item(fun, args)); if (queue.length === 1 && !draining) { runTimeout(drainQueue); } } // v8 likes predictible objects function Item(fun, array) { this.fun = fun; this.array = array; } Item.prototype.run = function () { this.fun.apply(null, this.array); }; var title = 'browser'; var platform = 'browser'; var browser = true; var env = {}; var argv = []; var version = ''; // empty string to avoid regexp issues var versions = {}; var release = {}; var config = {}; function noop() {} var on = noop; var addListener = noop; var once = noop; var off = noop; var removeListener = noop; var removeAllListeners = noop; var emit = noop; function binding(name) { throw new Error('process.binding is not supported'); } function cwd() { return '/'; } function chdir(dir) { throw new Error('process.chdir is not supported'); } function umask() { return 0; } // from https://github.com/kumavis/browser-process-hrtime/blob/master/index.js var performance = global$1.performance || {}; var performanceNow = performance.now || performance.mozNow || performance.msNow || performance.oNow || performance.webkitNow || function () { return new Date().getTime(); }; // generate timestamp or delta // see http://nodejs.org/api/process.html#process_process_hrtime function hrtime(previousTimestamp) { var clocktime = performanceNow.call(performance) * 1e-3; var seconds = Math.floor(clocktime); var nanoseconds = Math.floor(clocktime % 1 * 1e9); if (previousTimestamp) { seconds = seconds - previousTimestamp[0]; nanoseconds = nanoseconds - previousTimestamp[1]; if (nanoseconds < 0) { seconds--; nanoseconds += 1e9; } } return [seconds, nanoseconds]; } var startTime = new Date(); function uptime() { var currentTime = new Date(); var dif = currentTime - startTime; return dif / 1000; } var process = { nextTick: nextTick, title: title, browser: browser, env: env, argv: argv, version: version, versions: versions, on: on, addListener: addListener, once: once, off: off, removeListener: removeListener, removeAllListeners: removeAllListeners, emit: emit, binding: binding, cwd: cwd, chdir: chdir, umask: umask, hrtime: hrtime, platform: platform, release: release, config: config, uptime: uptime }; const debug = typeof process === 'object' && process.env && process.env.NODE_DEBUG && /\bsemver\b/i.test(process.env.NODE_DEBUG) ? (...args) => console.error('SEMVER', ...args) : () => {}; var debug_1 = debug; // Note: this is the semver.org version of the spec that it implements // Not necessarily the package version of this code. const SEMVER_SPEC_VERSION = '2.0.0'; const MAX_LENGTH = 256; const MAX_SAFE_INTEGER = Number.MAX_SAFE_INTEGER || /* istanbul ignore next */ 9007199254740991; // Max safe segment length for coercion. const MAX_SAFE_COMPONENT_LENGTH = 16; var constants = { SEMVER_SPEC_VERSION, MAX_LENGTH, MAX_SAFE_INTEGER, MAX_SAFE_COMPONENT_LENGTH }; function createCommonjsModule(fn, basedir, module) { return module = { path: basedir, exports: {}, require: function (path, base) { return commonjsRequire(path, (base === undefined || base === null) ? module.path : base); } }, fn(module, module.exports), module.exports; } function getCjsExportFromNamespace (n) { return n && n['default'] || n; } function commonjsRequire () { throw new Error('Dynamic requires are not currently supported by @rollup/plugin-commonjs'); } var re_1 = createCommonjsModule(function (module, exports) { const { MAX_SAFE_COMPONENT_LENGTH } = constants; exports = module.exports = {}; // The actual regexps go on exports.re const re = exports.re = []; const src = exports.src = []; const t = exports.t = {}; let R = 0; const createToken = (name, value, isGlobal) => { const index = R++; debug_1(index, value); t[name] = index; src[index] = value; re[index] = new RegExp(value, isGlobal ? 'g' : undefined); }; // The following Regular Expressions can be used for tokenizing, // validating, and parsing SemVer version strings. // ## Numeric Identifier // A single `0`, or a non-zero digit followed by zero or more digits. createToken('NUMERICIDENTIFIER', '0|[1-9]\\d*'); createToken('NUMERICIDENTIFIERLOOSE', '[0-9]+'); // ## Non-numeric Identifier // Zero or more digits, followed by a letter or hyphen, and then zero or // more letters, digits, or hyphens. createToken('NONNUMERICIDENTIFIER', '\\d*[a-zA-Z-][a-zA-Z0-9-]*'); // ## Main Version // Three dot-separated numeric identifiers. createToken('MAINVERSION', "(".concat(src[t.NUMERICIDENTIFIER], ")\\.") + "(".concat(src[t.NUMERICIDENTIFIER], ")\\.") + "(".concat(src[t.NUMERICIDENTIFIER], ")")); createToken('MAINVERSIONLOOSE', "(".concat(src[t.NUMERICIDENTIFIERLOOSE], ")\\.") + "(".concat(src[t.NUMERICIDENTIFIERLOOSE], ")\\.") + "(".concat(src[t.NUMERICIDENTIFIERLOOSE], ")")); // ## Pre-release Version Identifier // A numeric identifier, or a non-numeric identifier. createToken('PRERELEASEIDENTIFIER', "(?:".concat(src[t.NUMERICIDENTIFIER], "|").concat(src[t.NONNUMERICIDENTIFIER], ")")); createToken('PRERELEASEIDENTIFIERLOOSE', "(?:".concat(src[t.NUMERICIDENTIFIERLOOSE], "|").concat(src[t.NONNUMERICIDENTIFIER], ")")); // ## Pre-release Version // Hyphen, followed by one or more dot-separated pre-release version // identifiers. createToken('PRERELEASE', "(?:-(".concat(src[t.PRERELEASEIDENTIFIER], "(?:\\.").concat(src[t.PRERELEASEIDENTIFIER], ")*))")); createToken('PRERELEASELOOSE', "(?:-?(".concat(src[t.PRERELEASEIDENTIFIERLOOSE], "(?:\\.").concat(src[t.PRERELEASEIDENTIFIERLOOSE], ")*))")); // ## Build Metadata Identifier // Any combination of digits, letters, or hyphens. createToken('BUILDIDENTIFIER', '[0-9A-Za-z-]+'); // ## Build Metadata // Plus sign, followed by one or more period-separated build metadata // identifiers. createToken('BUILD', "(?:\\+(".concat(src[t.BUILDIDENTIFIER], "(?:\\.").concat(src[t.BUILDIDENTIFIER], ")*))")); // ## Full Version String // A main version, followed optionally by a pre-release version and // build metadata. // Note that the only major, minor, patch, and pre-release sections of // the version string are capturing groups. The build metadata is not a // capturing group, because it should not ever be used in version // comparison. createToken('FULLPLAIN', "v?".concat(src[t.MAINVERSION]).concat(src[t.PRERELEASE], "?").concat(src[t.BUILD], "?")); createToken('FULL', "^".concat(src[t.FULLPLAIN], "$")); // like full, but allows v1.2.3 and =1.2.3, which people do sometimes. // also, 1.0.0alpha1 (prerelease without the hyphen) which is pretty // common in the npm registry. createToken('LOOSEPLAIN', "[v=\\s]*".concat(src[t.MAINVERSIONLOOSE]).concat(src[t.PRERELEASELOOSE], "?").concat(src[t.BUILD], "?")); createToken('LOOSE', "^".concat(src[t.LOOSEPLAIN], "$")); createToken('GTLT', '((?:<|>)?=?)'); // Something like "2.*" or "1.2.x". // Note that "x.x" is a valid xRange identifer, meaning "any version" // Only the first item is strictly required. createToken('XRANGEIDENTIFIERLOOSE', "".concat(src[t.NUMERICIDENTIFIERLOOSE], "|x|X|\\*")); createToken('XRANGEIDENTIFIER', "".concat(src[t.NUMERICIDENTIFIER], "|x|X|\\*")); createToken('XRANGEPLAIN', "[v=\\s]*(".concat(src[t.XRANGEIDENTIFIER], ")") + "(?:\\.(".concat(src[t.XRANGEIDENTIFIER], ")") + "(?:\\.(".concat(src[t.XRANGEIDENTIFIER], ")") + "(?:".concat(src[t.PRERELEASE], ")?").concat(src[t.BUILD], "?") + ")?)?"); createToken('XRANGEPLAINLOOSE', "[v=\\s]*(".concat(src[t.XRANGEIDENTIFIERLOOSE], ")") + "(?:\\.(".concat(src[t.XRANGEIDENTIFIERLOOSE], ")") + "(?:\\.(".concat(src[t.XRANGEIDENTIFIERLOOSE], ")") + "(?:".concat(src[t.PRERELEASELOOSE], ")?").concat(src[t.BUILD], "?") + ")?)?"); createToken('XRANGE', "^".concat(src[t.GTLT], "\\s*").concat(src[t.XRANGEPLAIN], "$")); createToken('XRANGELOOSE', "^".concat(src[t.GTLT], "\\s*").concat(src[t.XRANGEPLAINLOOSE], "$")); // Coercion. // Extract anything that could conceivably be a part of a valid semver createToken('COERCE', "".concat('(^|[^\\d])' + '(\\d{1,').concat(MAX_SAFE_COMPONENT_LENGTH, "})") + "(?:\\.(\\d{1,".concat(MAX_SAFE_COMPONENT_LENGTH, "}))?") + "(?:\\.(\\d{1,".concat(MAX_SAFE_COMPONENT_LENGTH, "}))?") + "(?:$|[^\\d])"); createToken('COERCERTL', src[t.COERCE], true); // Tilde ranges. // Meaning is "reasonably at or greater than" createToken('LONETILDE', '(?:~>?)'); createToken('TILDETRIM', "(\\s*)".concat(src[t.LONETILDE], "\\s+"), true); exports.tildeTrimReplace = '$1~'; createToken('TILDE', "^".concat(src[t.LONETILDE]).concat(src[t.XRANGEPLAIN], "$")); createToken('TILDELOOSE', "^".concat(src[t.LONETILDE]).concat(src[t.XRANGEPLAINLOOSE], "$")); // Caret ranges. // Meaning is "at least and backwards compatible with" createToken('LONECARET', '(?:\\^)'); createToken('CARETTRIM', "(\\s*)".concat(src[t.LONECARET], "\\s+"), true); exports.caretTrimReplace = '$1^'; createToken('CARET', "^".concat(src[t.LONECARET]).concat(src[t.XRANGEPLAIN], "$")); createToken('CARETLOOSE', "^".concat(src[t.LONECARET]).concat(src[t.XRANGEPLAINLOOSE], "$")); // A simple gt/lt/eq thing, or just "" to indicate "any version" createToken('COMPARATORLOOSE', "^".concat(src[t.GTLT], "\\s*(").concat(src[t.LOOSEPLAIN], ")$|^$")); createToken('COMPARATOR', "^".concat(src[t.GTLT], "\\s*(").concat(src[t.FULLPLAIN], ")$|^$")); // An expression to strip any whitespace between the gtlt and the thing // it modifies, so that `> 1.2.3` ==> `>1.2.3` createToken('COMPARATORTRIM', "(\\s*)".concat(src[t.GTLT], "\\s*(").concat(src[t.LOOSEPLAIN], "|").concat(src[t.XRANGEPLAIN], ")"), true); exports.comparatorTrimReplace = '$1$2$3'; // Something like `1.2.3 - 1.2.4` // Note that these all use the loose form, because they'll be // checked against either the strict or loose comparator form // later. createToken('HYPHENRANGE', "^\\s*(".concat(src[t.XRANGEPLAIN], ")") + "\\s+-\\s+" + "(".concat(src[t.XRANGEPLAIN], ")") + "\\s*$"); createToken('HYPHENRANGELOOSE', "^\\s*(".concat(src[t.XRANGEPLAINLOOSE], ")") + "\\s+-\\s+" + "(".concat(src[t.XRANGEPLAINLOOSE], ")") + "\\s*$"); // Star ranges basically just allow anything at all. createToken('STAR', '(<|>)?=?\\s*\\*'); // >=0.0.0 is like a star createToken('GTE0', '^\\s*>=\\s*0\.0\.0\\s*$'); createToken('GTE0PRE', '^\\s*>=\\s*0\.0\.0-0\\s*$'); }); const numeric = /^[0-9]+$/; const compareIdentifiers = (a, b) => { const anum = numeric.test(a); const bnum = numeric.test(b); if (anum && bnum) { a = +a; b = +b; } return a === b ? 0 : anum && !bnum ? -1 : bnum && !anum ? 1 : a < b ? -1 : 1; }; const rcompareIdentifiers = (a, b) => compareIdentifiers(b, a); var identifiers = { compareIdentifiers, rcompareIdentifiers }; const { MAX_LENGTH: MAX_LENGTH$1, MAX_SAFE_INTEGER: MAX_SAFE_INTEGER$1 } = constants; const { re, t } = re_1; const { compareIdentifiers: compareIdentifiers$1 } = identifiers; class SemVer { constructor(version, options) { if (!options || typeof options !== 'object') { options = { loose: !!options, includePrerelease: false }; } if (version instanceof SemVer) { if (version.loose === !!options.loose && version.includePrerelease === !!options.includePrerelease) { return version; } else { version = version.version; } } else if (typeof version !== 'string') { throw new TypeError("Invalid Version: ".concat(version)); } if (version.length > MAX_LENGTH$1) { throw new TypeError("version is longer than ".concat(MAX_LENGTH$1, " characters")); } debug_1('SemVer', version, options); this.options = options; this.loose = !!options.loose; // this isn't actually relevant for versions, but keep it so that we // don't run into trouble passing this.options around. this.includePrerelease = !!options.includePrerelease; const m = version.trim().match(options.loose ? re[t.LOOSE] : re[t.FULL]); if (!m) { throw new TypeError("Invalid Version: ".concat(version)); } this.raw = version; // these are actually numbers this.major = +m[1]; this.minor = +m[2]; this.patch = +m[3]; if (this.major > MAX_SAFE_INTEGER$1 || this.major < 0) { throw new TypeError('Invalid major version'); } if (this.minor > MAX_SAFE_INTEGER$1 || this.minor < 0) { throw new TypeError('Invalid minor version'); } if (this.patch > MAX_SAFE_INTEGER$1 || this.patch < 0) { throw new TypeError('Invalid patch version'); } // numberify any prerelease numeric ids if (!m[4]) { this.prerelease = []; } else { this.prerelease = m[4].split('.').map(id => { if (/^[0-9]+$/.test(id)) { const num = +id; if (num >= 0 && num < MAX_SAFE_INTEGER$1) { return num; } } return id; }); } this.build = m[5] ? m[5].split('.') : []; this.format(); } format() { this.version = "".concat(this.major, ".").concat(this.minor, ".").concat(this.patch); if (this.prerelease.length) { this.version += "-".concat(this.prerelease.join('.')); } return this.version; } toString() { return this.version; } compare(other) { debug_1('SemVer.compare', this.version, this.options, other); if (!(other instanceof SemVer)) { if (typeof other === 'string' && other === this.version) { return 0; } other = new SemVer(other, this.options); } if (other.version === this.version) { return 0; } return this.compareMain(other) || this.comparePre(other); } compareMain(other) { if (!(other instanceof SemVer)) { other = new SemVer(other, this.options); } return compareIdentifiers$1(this.major, other.major) || compareIdentifiers$1(this.minor, other.minor) || compareIdentifiers$1(this.patch, other.patch); } comparePre(other) { if (!(other instanceof SemVer)) { other = new SemVer(other, this.options); } // NOT having a prerelease is > having one if (this.prerelease.length && !other.prerelease.length) { return -1; } else if (!this.prerelease.length && other.prerelease.length) { return 1; } else if (!this.prerelease.length && !other.prerelease.length) { return 0; } let i = 0; do { const a = this.prerelease[i]; const b = other.prerelease[i]; debug_1('prerelease compare', i, a, b); if (a === undefined && b === undefined) { return 0; } else if (b === undefined) { return 1; } else if (a === undefined) { return -1; } else if (a === b) { continue; } else { return compareIdentifiers$1(a, b); } } while (++i); } compareBuild(other) { if (!(other instanceof SemVer)) { other = new SemVer(other, this.options); } let i = 0; do { const a = this.build[i]; const b = other.build[i]; debug_1('prerelease compare', i, a, b); if (a === undefined && b === undefined) { return 0; } else if (b === undefined) { return 1; } else if (a === undefined) { return -1; } else if (a === b) { continue; } else { return compareIdentifiers$1(a, b); } } while (++i); } // preminor will bump the version up to the next minor release, and immediately // down to pre-release. premajor and prepatch work the same way. inc(release, identifier) { switch (release) { case 'premajor': this.prerelease.length = 0; this.patch = 0; this.minor = 0; this.major++; this.inc('pre', identifier); break; case 'preminor': this.prerelease.length = 0; this.patch = 0; this.minor++; this.inc('pre', identifier); break; case 'prepatch': // If this is already a prerelease, it will bump to the next version // drop any prereleases that might already exist, since they are not // relevant at this point. this.prerelease.length = 0; this.inc('patch', identifier); this.inc('pre', identifier); break; // If the input is a non-prerelease version, this acts the same as // prepatch. case 'prerelease': if (this.prerelease.length === 0) { this.inc('patch', identifier); } this.inc('pre', identifier); break; case 'major': // If this is a pre-major version, bump up to the same major version. // Otherwise increment major. // 1.0.0-5 bumps to 1.0.0 // 1.1.0 bumps to 2.0.0 if (this.minor !== 0 || this.patch !== 0 || this.prerelease.length === 0) { this.major++; } this.minor = 0; this.patch = 0; this.prerelease = []; break; case 'minor': // If this is a pre-minor version, bump up to the same minor version. // Otherwise increment minor. // 1.2.0-5 bumps to 1.2.0 // 1.2.1 bumps to 1.3.0 if (this.patch !== 0 || this.prerelease.length === 0) { this.minor++; } this.patch = 0; this.prerelease = []; break; case 'patch': // If this is not a pre-release version, it will increment the patch. // If it is a pre-release it will bump up to the same patch version. // 1.2.0-5 patches to 1.2.0 // 1.2.0 patches to 1.2.1 if (this.prerelease.length === 0) { this.patch++; } this.prerelease = []; break; // This probably shouldn't be used publicly. // 1.0.0 'pre' would become 1.0.0-0 which is the wrong direction. case 'pre': if (this.prerelease.length === 0) { this.prerelease = [0]; } else { let i = this.prerelease.length; while (--i >= 0) { if (typeof this.prerelease[i] === 'number') { this.prerelease[i]++; i = -2; } } if (i === -1) { // didn't increment anything this.prerelease.push(0); } } if (identifier) { // 1.2.0-beta.1 bumps to 1.2.0-beta.2, // 1.2.0-beta.fooblz or 1.2.0-beta bumps to 1.2.0-beta.0 if (this.prerelease[0] === identifier) { if (isNaN(this.prerelease[1])) { this.prerelease = [identifier, 0]; } } else { this.prerelease = [identifier, 0]; } } break; default: throw new Error("invalid increment argument: ".concat(release)); } this.format(); this.raw = this.version; return this; } } var semver = SemVer; const compare = (a, b, loose) => new semver(a, loose).compare(new semver(b, loose)); var compare_1 = compare; const lt = (a, b, loose) => compare_1(a, b, loose) < 0; var lt_1 = lt; const gte = (a, b, loose) => compare_1(a, b, loose) >= 0; var gte_1 = gte; var arrayify = (object, keyName) => Object.entries(object).map(([key, value]) => Object.assign({ [keyName]: key }, value)); var name = "prettier"; var version$1 = "2.1.1"; var description = "Prettier is an opinionated code formatter"; var bin = "./bin/prettier.js"; var repository = "prettier/prettier"; var homepage = "https://prettier.io"; var author = "James Long"; var license = "MIT"; var main = "./index.js"; var browser$1 = "./standalone.js"; var unpkg = "./standalone.js"; var engines = { node: ">=10.13.0" }; var files = [ "index.js", "standalone.js", "src", "bin" ]; var dependencies = { "@angular/compiler": "10.0.12", "@babel/code-frame": "7.10.4", "@babel/parser": "7.11.2", "@glimmer/syntax": "0.59.0", "@iarna/toml": "2.2.5", "@typescript-eslint/typescript-estree": "3.10.0", "angular-estree-parser": "2.2.0", "angular-html-parser": "1.7.1", camelcase: "6.0.0", chalk: "4.1.0", "ci-info": "watson/ci-info#f43f6a1cefff47fb361c88cf4b943fdbcaafe540", "cjk-regex": "2.0.0", cosmiconfig: "7.0.0", dashify: "2.0.0", diff: "4.0.2", editorconfig: "0.15.3", "editorconfig-to-prettier": "0.1.1", "escape-string-regexp": "4.0.0", esutils: "2.0.3", "fast-glob": "3.2.4", "fast-json-stable-stringify": "2.1.0", "find-parent-dir": "0.3.0", "flow-parser": "0.132.0", "get-stream": "6.0.0", globby: "11.0.1", graphql: "15.3.0", "html-element-attributes": "2.2.1", "html-styles": "1.0.0", "html-tag-names": "1.1.5", "html-void-elements": "1.0.5", ignore: "4.0.6", "jest-docblock": "26.0.0", json5: "2.1.3", leven: "3.1.0", "lines-and-columns": "1.1.6", "linguist-languages": "7.10.0", lodash: "4.17.20", mem: "6.1.0", minimatch: "3.0.4", minimist: "1.2.5", "n-readlines": "1.0.0", outdent: "0.7.1", "parse-srcset": "ikatyang/parse-srcset#54eb9c1cb21db5c62b4d0e275d7249516df6f0ee", "please-upgrade-node": "3.2.0", "postcss-less": "3.1.4", "postcss-media-query-parser": "0.2.3", "postcss-scss": "2.1.1", "postcss-selector-parser": "2.2.3", "postcss-values-parser": "2.0.1", "regexp-util": "1.2.2", "remark-footnotes": "2.0.0", "remark-math": "1.0.6", "remark-parse": "8.0.3", resolve: "1.17.0", semver: "7.3.2", "string-width": "4.2.0", typescript: "4.0.2", "unicode-regex": "3.0.0", unified: "9.2.0", vnopts: "1.0.2", "yaml-unist-parser": "1.3.0" }; var devDependencies = { "@babel/core": "7.11.4", "@babel/preset-env": "7.11.0", "@babel/types": "7.11.0", "@glimmer/reference": "0.59.0", "@rollup/plugin-alias": "3.1.1", "@rollup/plugin-babel": "5.2.0", "@rollup/plugin-commonjs": "14.0.0", "@rollup/plugin-json": "4.1.0", "@rollup/plugin-node-resolve": "9.0.0", "@rollup/plugin-replace": "2.3.3", "@types/estree": "0.0.45", "@types/node": "14.6.0", "@typescript-eslint/types": "3.10.0", "babel-loader": "8.1.0", benchmark: "2.1.4", "builtin-modules": "3.1.0", "cross-env": "7.0.2", cspell: "4.1.0", eslint: "7.7.0", "eslint-config-prettier": "6.11.0", "eslint-formatter-friendly": "7.0.0", "eslint-plugin-import": "2.22.0", "eslint-plugin-jest": "23.20.0", "eslint-plugin-prettier-internal-rules": "file:scripts/tools/eslint-plugin-prettier-internal-rules", "eslint-plugin-react": "7.20.6", "eslint-plugin-unicorn": "21.0.0", execa: "4.0.3", jest: "26.4.2", "jest-snapshot-serializer-ansi": "1.0.0", "jest-snapshot-serializer-raw": "1.1.0", "jest-watch-typeahead": "0.6.0", "npm-run-all": "4.1.5", prettier: "2.1.0", rimraf: "3.0.2", rollup: "2.26.5", "rollup-plugin-node-globals": "1.4.0", "rollup-plugin-terser": "7.0.0", shelljs: "0.8.4", "snapshot-diff": "0.8.1", "strip-ansi": "6.0.0", "synchronous-promise": "2.0.13", tempy: "0.6.0", "terser-webpack-plugin": "4.1.0", webpack: "4.44.1" }; var scripts = { prepublishOnly: "echo \"Error: must publish from dist/\" && exit 1", "prepare-release": "yarn && yarn build && yarn test:dist", test: "jest", "test:dev-package": "cross-env INSTALL_PACKAGE=1 jest", "test:dist": "cross-env NODE_ENV=production jest", "test:dist-standalone": "cross-env NODE_ENV=production TEST_STANDALONE=1 jest", "test:integration": "jest tests_integration", "perf:repeat": "yarn && yarn build && cross-env NODE_ENV=production node ./dist/bin-prettier.js --debug-repeat ${PERF_REPEAT:-1000} --loglevel debug ${PERF_FILE:-./index.js} > /dev/null", "perf:repeat-inspect": "yarn && yarn build && cross-env NODE_ENV=production node --inspect-brk ./dist/bin-prettier.js --debug-repeat ${PERF_REPEAT:-1000} --loglevel debug ${PERF_FILE:-./index.js} > /dev/null", "perf:benchmark": "yarn && yarn build && cross-env NODE_ENV=production node ./dist/bin-prettier.js --debug-benchmark --loglevel debug ${PERF_FILE:-./index.js} > /dev/null", lint: "run-p lint:*", "lint:typecheck": "tsc", "lint:eslint": "cross-env EFF_NO_LINK_RULES=true eslint . --format friendly", "lint:changelog": "node ./scripts/lint-changelog.js", "lint:prettier": "prettier . \"!test*\" --check", "lint:dist": "eslint --no-eslintrc --no-ignore --env=es6,browser --parser-options=ecmaVersion:2016 \"dist/!(bin-prettier|index|third-party).js\"", "lint:spellcheck": "cspell \"**/*\" \".github/**/*\"", "lint:deps": "node ./scripts/check-deps.js", fix: "run-s fix:eslint fix:prettier", "fix:eslint": "yarn lint:eslint --fix", "fix:prettier": "yarn lint:prettier --write", build: "node ./scripts/build/build.js", "build-docs": "node ./scripts/build-docs.js" }; var _package = { name: name, version: version$1, description: description, bin: bin, repository: repository, homepage: homepage, author: author, license: license, main: main, browser: browser$1, unpkg: unpkg, engines: engines, files: files, dependencies: dependencies, devDependencies: devDependencies, scripts: scripts }; var _package$1 = /*#__PURE__*/Object.freeze({ __proto__: null, name: name, version: version$1, description: description, bin: bin, repository: repository, homepage: homepage, author: author, license: license, main: main, browser: browser$1, unpkg: unpkg, engines: engines, files: files, dependencies: dependencies, devDependencies: devDependencies, scripts: scripts, 'default': _package }); var lib = createCommonjsModule(function (module, exports) { Object.defineProperty(exports, "__esModule", { value: true }); // In the absence of a WeakSet or WeakMap implementation, don't break, but don't cache either. function noop() { var args = []; for (var _i = 0; _i < arguments.length; _i++) { args[_i] = arguments[_i]; } } function createWeakMap() { if (typeof WeakMap !== 'undefined') { return new WeakMap(); } else { return fakeSetOrMap(); } } /** * Creates and returns a no-op implementation of a WeakMap / WeakSet that never stores anything. */ function fakeSetOrMap() { return { add: noop, delete: noop, get: noop, set: noop, has: function has(k) { return false; } }; } // Safe hasOwnProperty var hop = Object.prototype.hasOwnProperty; var has = function has(obj, prop) { return hop.call(obj, prop); }; // Copy all own enumerable properties from source to target function extend(target, source) { for (var prop in source) { if (has(source, prop)) { target[prop] = source[prop]; } } return target; } var reLeadingNewline = /^[ \t]*(?:\r\n|\r|\n)/; var reTrailingNewline = /(?:\r\n|\r|\n)[ \t]*$/; var reStartsWithNewlineOrIsEmpty = /^(?:[\r\n]|$)/; var reDetectIndentation = /(?:\r\n|\r|\n)([ \t]*)(?:[^ \t\r\n]|$)/; var reOnlyWhitespaceWithAtLeastOneNewline = /^[ \t]*[\r\n][ \t\r\n]*$/; function _outdentArray(strings, firstInterpolatedValueSetsIndentationLevel, options) { // If first interpolated value is a reference to outdent, // determine indentation level from the indentation of the interpolated value. var indentationLevel = 0; var match = strings[0].match(reDetectIndentation); if (match) { indentationLevel = match[1].length; } var reSource = "(\\r\\n|\\r|\\n).{0," + indentationLevel + "}"; var reMatchIndent = new RegExp(reSource, 'g'); if (firstInterpolatedValueSetsIndentationLevel) { strings = strings.slice(1); } var newline = options.newline, trimLeadingNewline = options.trimLeadingNewline, trimTrailingNewline = options.trimTrailingNewline; var normalizeNewlines = typeof newline === 'string'; var l = strings.length; var outdentedStrings = strings.map(function (v, i) { // Remove leading indentation from all lines v = v.replace(reMatchIndent, '$1'); // Trim a leading newline from the first string if (i === 0 && trimLeadingNewline) { v = v.replace(reLeadingNewline, ''); } // Trim a trailing newline from the last string if (i === l - 1 && trimTrailingNewline) { v = v.replace(reTrailingNewline, ''); } // Normalize newlines if (normalizeNewlines) { v = v.replace(/\r\n|\n|\r/g, function (_) { return newline; }); } return v; }); return outdentedStrings; } function concatStringsAndValues(strings, values) { var ret = ''; for (var i = 0, l = strings.length; i < l; i++) { ret += strings[i]; if (i < l - 1) { ret += values[i]; } } return ret; } function isTemplateStringsArray(v) { return has(v, 'raw') && has(v, 'length'); } /** * It is assumed that opts will not change. If this is a problem, clone your options object and pass the clone to * makeInstance * @param options * @return {outdent} */ function createInstance(options) { /** Cache of pre-processed template literal arrays */ var arrayAutoIndentCache = createWeakMap(); /** * Cache of pre-processed template literal arrays, where first interpolated value is a reference to outdent, * before interpolated values are injected. */ var arrayFirstInterpSetsIndentCache = createWeakMap(); function outdent(stringsOrOptions) { var values = []; for (var _i = 1; _i < arguments.length; _i++) { values[_i - 1] = arguments[_i]; } /* tslint:enable:no-shadowed-variable */ if (isTemplateStringsArray(stringsOrOptions)) { var strings = stringsOrOptions; // Is first interpolated value a reference to outdent, alone on its own line, without any preceding non-whitespace? var firstInterpolatedValueSetsIndentationLevel = (values[0] === outdent || values[0] === defaultOutdent) && reOnlyWhitespaceWithAtLeastOneNewline.test(strings[0]) && reStartsWithNewlineOrIsEmpty.test(strings[1]); // Perform outdentation var cache = firstInterpolatedValueSetsIndentationLevel ? arrayFirstInterpSetsIndentCache : arrayAutoIndentCache; var renderedArray = cache.get(strings); if (!renderedArray) { renderedArray = _outdentArray(strings, firstInterpolatedValueSetsIndentationLevel, options); cache.set(strings, renderedArray); } /** If no interpolated values, skip concatenation step */ if (values.length === 0) { return renderedArray[0]; } /** Concatenate string literals with interpolated values */ var rendered = concatStringsAndValues(renderedArray, firstInterpolatedValueSetsIndentationLevel ? values.slice(1) : values); return rendered; } else { // Create and return a new instance of outdent with the given options return createInstance(extend(extend({}, options), stringsOrOptions || {})); } } var fullOutdent = extend(outdent, { string: function string(str) { return _outdentArray([str], false, options)[0]; } }); return fullOutdent; } var defaultOutdent = createInstance({ trimLeadingNewline: true, trimTrailingNewline: true }); exports.outdent = defaultOutdent; // Named exports. Simple and preferred. // import outdent from 'outdent'; exports.default = defaultOutdent; { // In webpack harmony-modules environments, module.exports is read-only, // so we fail gracefully. try { module.exports = defaultOutdent; Object.defineProperty(defaultOutdent, '__esModule', { value: true }); defaultOutdent.default = defaultOutdent; defaultOutdent.outdent = defaultOutdent; } catch (e) {} } }); function _templateObject6() { const data = _taggedTemplateLiteral(["\n Require either '@prettier' or '@format' to be present in the file's first docblock comment\n in order for it to be formatted.\n "]); _templateObject6 = function _templateObject6() { return data; }; return data; } function _templateObject5() { const data = _taggedTemplateLiteral(["\n Format code starting at a given character offset.\n The range will extend backwards to the start of the first line containing the selected statement.\n This option cannot be used with --cursor-offset.\n "]); _templateObject5 = function _templateObject5() { return data; }; return data; } function _templateObject4() { const data = _taggedTemplateLiteral(["\n Format code ending at a given character offset (exclusive).\n The range will extend forwards to the end of the selected statement.\n This option cannot be used with --cursor-offset.\n "]); _templateObject4 = function _templateObject4() { return data; }; return data; } function _templateObject3() { const data = _taggedTemplateLiteral(["\n Custom directory that contains prettier plugins in node_modules subdirectory.\n Overrides default behavior when plugins are searched relatively to the location of Prettier.\n Multiple values are accepted.\n "]); _templateObject3 = function _templateObject3() { return data; }; return data; } function _templateObject2() { const data = _taggedTemplateLiteral(["\n Maintain existing\n (mixed values within one file are normalised by looking at what's used after the first line)\n "]); _templateObject2 = function _templateObject2() { return data; }; return data; } function _templateObject() { const data = _taggedTemplateLiteral(["\n Print (to stderr) where a cursor at the given position would move to after formatting.\n This option cannot be used with --range-start and --range-end.\n "]); _templateObject = function _templateObject() { return data; }; return data; } const { outdent } = lib; const CATEGORY_CONFIG = "Config"; const CATEGORY_EDITOR = "Editor"; const CATEGORY_FORMAT = "Format"; const CATEGORY_OTHER = "Other"; const CATEGORY_OUTPUT = "Output"; const CATEGORY_GLOBAL = "Global"; const CATEGORY_SPECIAL = "Special"; /** * @typedef {Object} OptionInfo * @property {string} [since] - available since version * @property {string} category * @property {'int' | 'boolean' | 'choice' | 'path'} type * @property {boolean} [array] - indicate it's an array of the specified type * @property {OptionValueInfo} [default] * @property {OptionRangeInfo} [range] - for type int * @property {string} description * @property {string} [deprecated] - deprecated since version * @property {OptionRedirectInfo} [redirect] - redirect deprecated option * @property {(value: any) => boolean} [exception] * @property {OptionChoiceInfo[]} [choices] - for type choice * @property {string} [cliName] * @property {string} [cliCategory] * @property {string} [cliDescription] * * @typedef {number | boolean | string} OptionValue * @typedef {OptionValue | [{ value: OptionValue[] }] | Array<{ since: string, value: OptionValue}>} OptionValueInfo * * @typedef {Object} OptionRedirectInfo * @property {string} option * @property {OptionValue} value * * @typedef {Object} OptionRangeInfo * @property {number} start - recommended range start * @property {number} end - recommended range end * @property {number} step - recommended range step * * @typedef {Object} OptionChoiceInfo * @property {boolean | string} value - boolean for the option that is originally boolean type * @property {string} description * @property {string} [since] - undefined if available since the first version of the option * @property {string} [deprecated] - deprecated since version * @property {OptionValueInfo} [redirect] - redirect deprecated value */ /** @type {{ [name: string]: OptionInfo }} */ const options = { cursorOffset: { since: "1.4.0", category: CATEGORY_SPECIAL, type: "int", default: -1, range: { start: -1, end: Infinity, step: 1 }, description: outdent(_templateObject()), cliCategory: CATEGORY_EDITOR }, endOfLine: { since: "1.15.0", category: CATEGORY_GLOBAL, type: "choice", default: [{ since: "1.15.0", value: "auto" }, { since: "2.0.0", value: "lf" }], description: "Which end of line characters to apply.", choices: [{ value: "lf", description: "Line Feed only (\\n), common on Linux and macOS as well as inside git repos" }, { value: "crlf", description: "Carriage Return + Line Feed characters (\\r\\n), common on Windows" }, { value: "cr", description: "Carriage Return character only (\\r), used very rarely" }, { value: "auto", description: outdent(_templateObject2()) }] }, filepath: { since: "1.4.0", category: CATEGORY_SPECIAL, type: "path", description: "Specify the input filepath. This will be used to do parser inference.", cliName: "stdin-filepath", cliCategory: CATEGORY_OTHER, cliDescription: "Path to the file to pretend that stdin comes from." }, insertPragma: { since: "1.8.0", category: CATEGORY_SPECIAL, type: "boolean", default: false, description: "Insert @format pragma into file's first docblock comment.", cliCategory: CATEGORY_OTHER }, parser: { since: "0.0.10", category: CATEGORY_GLOBAL, type: "choice", default: [{ since: "0.0.10", value: "babylon" }, { since: "1.13.0", value: undefined }], description: "Which parser to use.", exception: value => typeof value === "string" || typeof value === "function", choices: [{ value: "flow", description: "Flow" }, { value: "babel", since: "1.16.0", description: "JavaScript" }, { value: "babel-flow", since: "1.16.0", description: "Flow" }, { value: "babel-ts", since: "2.0.0", description: "TypeScript" }, { value: "typescript", since: "1.4.0", description: "TypeScript" }, { value: "css", since: "1.7.1", description: "CSS" }, { value: "less", since: "1.7.1", description: "Less" }, { value: "scss", since: "1.7.1", description: "SCSS" }, { value: "json", since: "1.5.0", description: "JSON" }, { value: "json5", since: "1.13.0", description: "JSON5" }, { value: "json-stringify", since: "1.13.0", description: "JSON.stringify" }, { value: "graphql", since: "1.5.0", description: "GraphQL" }, { value: "markdown", since: "1.8.0", description: "Markdown" }, { value: "mdx", since: "1.15.0", description: "MDX" }, { value: "vue", since: "1.10.0", description: "Vue" }, { value: "yaml", since: "1.14.0", description: "YAML" }, { value: "glimmer", since: null, description: "Handlebars" }, { value: "html", since: "1.15.0", description: "HTML" }, { value: "angular", since: "1.15.0", description: "Angular" }, { value: "lwc", since: "1.17.0", description: "Lightning Web Components" }] }, plugins: { since: "1.10.0", type: "path", array: true, default: [{ value: [] }], category: CATEGORY_GLOBAL, description: "Add a plugin. Multiple plugins can be passed as separate `--plugin`s.", exception: value => typeof value === "string" || typeof value === "object", cliName: "plugin", cliCategory: CATEGORY_CONFIG }, pluginSearchDirs: { since: "1.13.0", type: "path", array: true, default: [{ value: [] }], category: CATEGORY_GLOBAL, description: outdent(_templateObject3()), exception: value => typeof value === "string" || typeof value === "object", cliName: "plugin-search-dir", cliCategory: CATEGORY_CONFIG }, printWidth: { since: "0.0.0", category: CATEGORY_GLOBAL, type: "int", default: 80, description: "The line length where Prettier will try wrap.", range: { start: 0, end: Infinity, step: 1 } }, rangeEnd: { since: "1.4.0", category: CATEGORY_SPECIAL, type: "int", default: Infinity, range: { start: 0, end: Infinity, step: 1 }, description: outdent(_templateObject4()), cliCategory: CATEGORY_EDITOR }, rangeStart: { since: "1.4.0", category: CATEGORY_SPECIAL, type: "int", default: 0, range: { start: 0, end: Infinity, step: 1 }, description: outdent(_templateObject5()), cliCategory: CATEGORY_EDITOR }, requirePragma: { since: "1.7.0", category: CATEGORY_SPECIAL, type: "boolean", default: false, description: outdent(_templateObject6()), cliCategory: CATEGORY_OTHER }, tabWidth: { type: "int", category: CATEGORY_GLOBAL, default: 2, description: "Number of spaces per indentation level.", range: { start: 0, end: Infinity, step: 1 } }, useTabs: { since: "1.0.0", category: CATEGORY_GLOBAL, type: "boolean", default: false, description: "Indent with tabs instead of spaces." }, embeddedLanguageFormatting: { since: "2.1.0", category: CATEGORY_GLOBAL, type: "choice", default: [{ since: "2.1.0", value: "auto" }], description: "Control how Prettier formats quoted code embedded in the file.", choices: [{ value: "auto", description: "Format embedded code if Prettier can automatically identify it." }, { value: "off", description: "Never automatically format embedded code." }] } }; var coreOptions = { CATEGORY_CONFIG, CATEGORY_EDITOR, CATEGORY_FORMAT, CATEGORY_OTHER, CATEGORY_OUTPUT, CATEGORY_GLOBAL, CATEGORY_SPECIAL, options }; var require$$3 = getCjsExportFromNamespace(_package$1); const semver$1 = { compare: compare_1, lt: lt_1, gte: gte_1 }; const currentVersion = require$$3.version; const coreOptions$1 = coreOptions.options; /** * Strings in `plugins` and `pluginSearchDirs` are handled by a wrapped version * of this function created by `withPlugins`. Don't pass them here directly. * @param {object} param0 * @param {(string | object)[]=} param0.plugins Strings are resolved by `withPlugins`. * @param {string[]=} param0.pluginSearchDirs Added by `withPlugins`. * @param {boolean=} param0.showUnreleased * @param {boolean=} param0.showDeprecated * @param {boolean=} param0.showInternal */ function getSupportInfo({ plugins = [], showUnreleased = false, showDeprecated = false, showInternal = false } = {}) { // pre-release version is smaller than the normal version in semver, // we need to treat it as the normal one so as to test new features. const version = currentVersion.split("-", 1)[0]; const languages = plugins.reduce((all, plugin) => all.concat(plugin.languages || []), []).filter(filterSince); const options = arrayify(Object.assign({}, ...plugins.map(({ options }) => options), coreOptions$1), "name").filter(option => filterSince(option) && filterDeprecated(option)).sort((a, b) => a.name === b.name ? 0 : a.name < b.name ? -1 : 1).map(mapInternal).map(option => { option = Object.assign({}, option); if (Array.isArray(option.default)) { option.default = option.default.length === 1 ? option.default[0].value : option.default.filter(filterSince).sort((info1, info2) => semver$1.compare(info2.since, info1.since))[0].value; } if (Array.isArray(option.choices)) { option.choices = option.choices.filter(option => filterSince(option) && filterDeprecated(option)); if (option.name === "parser") { collectParsersFromLanguages(option, languages, plugins); } } const pluginDefaults = plugins.filter(plugin => plugin.defaultOptions && plugin.defaultOptions[option.name] !== undefined).reduce((reduced, plugin) => { reduced[plugin.name] = plugin.defaultOptions[option.name]; return reduced; }, {}); return Object.assign({}, option, { pluginDefaults }); }); return { languages, options }; function filterSince(object) { return showUnreleased || !("since" in object) || object.since && semver$1.gte(version, object.since); } function filterDeprecated(object) { return showDeprecated || !("deprecated" in object) || object.deprecated && semver$1.lt(version, object.deprecated); } function mapInternal(object) { if (showInternal) { return object; } const newObject = _objectWithoutPropertiesLoose(object, ["cliName", "cliCategory", "cliDescription"]); return newObject; } } function collectParsersFromLanguages(option, languages, plugins) { const existingValues = new Set(option.choices.map(choice => choice.value)); for (const language of languages) { if (language.parsers) { for (const value of language.parsers) { if (!existingValues.has(value)) { existingValues.add(value); const plugin = plugins.find(plugin => plugin.parsers && plugin.parsers[value]); let description = language.name; if (plugin && plugin.name) { description += " (plugin: ".concat(plugin.name, ")"); } option.choices.push({ value, description }); } } } } } var support = { getSupportInfo }; const notAsciiRegex = /[^\x20-\x7F]/; const getPenultimate = arr => arr[arr.length - 2]; /** * @typedef {{backwards?: boolean}} SkipOptions */ /** * @param {string | RegExp} chars * @returns {(text: string, index: number | false, opts?: SkipOptions) => number | false} */ function skip(chars) { return (text, index, opts) => { const backwards = opts && opts.backwards; // Allow `skip` functions to be threaded together without having // to check for failures (did someone say monads?). /* istanbul ignore next */ if (index === false) { return false; } const { length } = text; let cursor = index; while (cursor >= 0 && cursor < length) { const c = text.charAt(cursor); if (chars instanceof RegExp) { if (!chars.test(c)) { return cursor; } } else if (!chars.includes(c)) { return cursor; } backwards ? cursor-- : cursor++; } if (cursor === -1 || cursor === length) { // If we reached the beginning or end of the file, return the // out-of-bounds cursor. It's up to the caller to handle this // correctly. We don't want to indicate `false` though if it // actually skipped valid characters. return cursor; } return false; }; } /** * @type {(text: string, index: number | false, opts?: SkipOptions) => number | false} */ const skipWhitespace = skip(/\s/); /** * @type {(text: string, index: number | false, opts?: SkipOptions) => number | false} */ const skipSpaces = skip(" \t"); /** * @type {(text: string, index: number | false, opts?: SkipOptions) => number | false} */ const skipToLineEnd = skip(",; \t"); /** * @type {(text: string, index: number | false, opts?: SkipOptions) => number | false} */ const skipEverythingButNewLine = skip(/[^\n\r]/); /** * @param {string} text * @param {number | false} index * @returns {number | false} */ function skipInlineComment(text, index) { /* istanbul ignore next */ if (index === false) { return false; } if (text.charAt(index) === "/" && text.charAt(index + 1) === "*") { for (let i = index + 2; i < text.length; ++i) { if (text.charAt(i) === "*" && text.charAt(i + 1) === "/") { return i + 2; } } } return index; } /** * @param {string} text * @param {number | false} index * @returns {number | false} */ function skipTrailingComment(text, index) { /* istanbul ignore next */ if (index === false) { return false; } if (text.charAt(index) === "/" && text.charAt(index + 1) === "/") { return skipEverythingButNewLine(text, index); } return index; } // This one doesn't use the above helper function because it wants to // test \r\n in order and `skip` doesn't support ordering and we only // want to skip one newline. It's simple to implement. /** * @param {string} text * @param {number | false} index * @param {SkipOptions=} opts * @returns {number | false} */ function skipNewline(text, index, opts) { const backwards = opts && opts.backwards; if (index === false) { return false; } const atIndex = text.charAt(index); if (backwards) { // We already replace `\r\n` with `\n` before parsing /* istanbul ignore next */ if (text.charAt(index - 1) === "\r" && atIndex === "\n") { return index - 2; } if (atIndex === "\n" || atIndex === "\r" || atIndex === "\u2028" || atIndex === "\u2029") { return index - 1; } } else { // We already replace `\r\n` with `\n` before parsing /* istanbul ignore next */ if (atIndex === "\r" && text.charAt(index + 1) === "\n") { return index + 2; } if (atIndex === "\n" || atIndex === "\r" || atIndex === "\u2028" || atIndex === "\u2029") { return index + 1; } } return index; } /** * @param {string} text * @param {number} index * @param {SkipOptions=} opts * @returns {boolean} */ function hasNewline(text, index, opts) { opts = opts || {}; const idx = skipSpaces(text, opts.backwards ? index - 1 : index, opts); const idx2 = skipNewline(text, idx, opts); return idx !== idx2; } /** * @param {string} text * @param {number} start * @param {number} end * @returns {boolean} */ function hasNewlineInRange(text, start, end) { for (let i = start; i < end; ++i) { if (text.charAt(i) === "\n") { return true; } } return false; } // Note: this function doesn't ignore leading comments unlike isNextLineEmpty /** * @template N * @param {string} text * @param {N} node * @param {(node: N) => number} locStart */ function isPreviousLineEmpty(text, node, locStart) { /** @type {number | false} */ let idx = locStart(node) - 1; idx = skipSpaces(text, idx, { backwards: true }); idx = skipNewline(text, idx, { backwards: true }); idx = skipSpaces(text, idx, { backwards: true }); const idx2 = skipNewline(text, idx, { backwards: true }); return idx !== idx2; } /** * @param {string} text * @param {number} index * @returns {boolean} */ function isNextLineEmptyAfterIndex(text, index) { /** @type {number | false} */ let oldIdx = null; /** @type {number | false} */ let idx = index; while (idx !== oldIdx) { // We need to skip all the potential trailing inline comments oldIdx = idx; idx = skipToLineEnd(text, idx); idx = skipInlineComment(text, idx); idx = skipSpaces(text, idx); } idx = skipTrailingComment(text, idx); idx = skipNewline(text, idx); return idx !== false && hasNewline(text, idx); } /** * @template N * @param {string} text * @param {N} node * @param {(node: N) => number} locEnd * @returns {boolean} */ function isNextLineEmpty(text, node, locEnd) { return isNextLineEmptyAfterIndex(text, locEnd(node)); } /** * @param {string} text * @param {number} idx * @returns {number | false} */ function getNextNonSpaceNonCommentCharacterIndexWithStartIndex(text, idx) { /** @type {number | false} */ let oldIdx = null; /** @type {number | false} */ let nextIdx = idx; while (nextIdx !== oldIdx) { oldIdx = nextIdx; nextIdx = skipSpaces(text, nextIdx); nextIdx = skipInlineComment(text, nextIdx); nextIdx = skipTrailingComment(text, nextIdx); nextIdx = skipNewline(text, nextIdx); } return nextIdx; } /** * @template N * @param {string} text * @param {N} node * @param {(node: N) => number} locEnd * @returns {number | false} */ function getNextNonSpaceNonCommentCharacterIndex(text, node, locEnd) { return getNextNonSpaceNonCommentCharacterIndexWithStartIndex(text, locEnd(node)); } /** * @template N * @param {string} text * @param {N} node * @param {(node: N) => number} locEnd * @returns {string} */ function getNextNonSpaceNonCommentCharacter(text, node, locEnd) { return text.charAt( // @ts-ignore => TBD: can return false, should we define a fallback? getNextNonSpaceNonCommentCharacterIndex(text, node, locEnd)); } // Not using, but it's public utils /* istanbul ignore next */ /** * @param {string} text * @param {number} index * @param {SkipOptions=} opts * @returns {boolean} */ function hasSpaces(text, index, opts) { opts = opts || {}; const idx = skipSpaces(text, opts.backwards ? index - 1 : index, opts); return idx !== index; } /** * @param {string} value * @param {number} tabWidth * @param {number=} startIndex * @returns {number} */ function getAlignmentSize(value, tabWidth, startIndex) { startIndex = startIndex || 0; let size = 0; for (let i = startIndex; i < value.length; ++i) { if (value[i] === "\t") { // Tabs behave in a way that they are aligned to the nearest // multiple of tabWidth: // 0 -> 4, 1 -> 4, 2 -> 4, 3 -> 4 // 4 -> 8, 5 -> 8, 6 -> 8, 7 -> 8 ... size = size + tabWidth - size % tabWidth; } else { size++; } } return size; } /** * @param {string} value * @param {number} tabWidth * @returns {number} */ function getIndentSize(value, tabWidth) { const lastNewlineIndex = value.lastIndexOf("\n"); if (lastNewlineIndex === -1) { return 0; } return getAlignmentSize( // All the leading whitespaces value.slice(lastNewlineIndex + 1).match(/^[\t ]*/)[0], tabWidth); } /** * @typedef {'"' | "'"} Quote */ /** * * @param {string} raw * @param {Quote} preferredQuote * @returns {Quote} */ function getPreferredQuote(raw, preferredQuote) { // `rawContent` is the string exactly like it appeared in the input source // code, without its enclosing quotes. const rawContent = raw.slice(1, -1); /** @type {{ quote: '"', regex: RegExp }} */ const double = { quote: '"', regex: /"/g }; /** @type {{ quote: "'", regex: RegExp }} */ const single = { quote: "'", regex: /'/g }; const preferred = preferredQuote === "'" ? single : double; const alternate = preferred === single ? double : single; let result = preferred.quote; // If `rawContent` contains at least one of the quote preferred for enclosing // the string, we might want to enclose with the alternate quote instead, to // minimize the number of escaped quotes. if (rawContent.includes(preferred.quote) || rawContent.includes(alternate.quote)) { const numPreferredQuotes = (rawContent.match(preferred.regex) || []).length; const numAlternateQuotes = (rawContent.match(alternate.regex) || []).length; result = numPreferredQuotes > numAlternateQuotes ? alternate.quote : preferred.quote; } return result; } function printString(raw, options, isDirectiveLiteral) { // `rawContent` is the string exactly like it appeared in the input source // code, without its enclosing quotes. const rawContent = raw.slice(1, -1); // Check for the alternate quote, to determine if we're allowed to swap // the quotes on a DirectiveLiteral. const canChangeDirectiveQuotes = !rawContent.includes('"') && !rawContent.includes("'"); /** @type {Quote} */ const enclosingQuote = options.parser === "json" ? '"' : options.__isInHtmlAttribute ? "'" : getPreferredQuote(raw, options.singleQuote ? "'" : '"'); // Directives are exact code unit sequences, which means that you can't // change the escape sequences they use. // See https://github.com/prettier/prettier/issues/1555 // and https://tc39.github.io/ecma262/#directive-prologue if (isDirectiveLiteral) { if (canChangeDirectiveQuotes) { return enclosingQuote + rawContent + enclosingQuote; } return raw; } // It might sound unnecessary to use `makeString` even if the string already // is enclosed with `enclosingQuote`, but it isn't. The string could contain // unnecessary escapes (such as in `"\'"`). Always using `makeString` makes // sure that we consistently output the minimum amount of escaped quotes. return makeString(rawContent, enclosingQuote, !(options.parser === "css" || options.parser === "less" || options.parser === "scss" || options.embeddedInHtml)); } /** * @param {string} rawContent * @param {Quote} enclosingQuote * @param {boolean=} unescapeUnnecessaryEscapes * @returns {string} */ function makeString(rawContent, enclosingQuote, unescapeUnnecessaryEscapes) { const otherQuote = enclosingQuote === '"' ? "'" : '"'; // Matches _any_ escape and unescaped quotes (both single and double). const regex = /\\([\S\s])|(["'])/g; // Escape and unescape single and double quotes as needed to be able to // enclose `rawContent` with `enclosingQuote`. const newContent = rawContent.replace(regex, (match, escaped, quote) => { // If we matched an escape, and the escaped character is a quote of the // other type than we intend to enclose the string with, there's no need for // it to be escaped, so return it _without_ the backslash. if (escaped === otherQuote) { return escaped; } // If we matched an unescaped quote and it is of the _same_ type as we // intend to enclose the string with, it must be escaped, so return it with // a backslash. if (quote === enclosingQuote) { return "\\" + quote; } if (quote) { return quote; } // Unescape any unnecessarily escaped character. // Adapted from https://github.com/eslint/eslint/blob/de0b4ad7bd820ade41b1f606008bea68683dc11a/lib/rules/no-useless-escape.js#L27 return unescapeUnnecessaryEscapes && /^[^\n\r"'0-7\\bfnrt-vx\u2028\u2029]$/.test(escaped) ? escaped : "\\" + escaped; }); return enclosingQuote + newContent + enclosingQuote; } function printNumber(rawNumber) { return rawNumber.toLowerCase() // Remove unnecessary plus and zeroes from scientific notation. .replace(/^([+-]?[\d.]+e)(?:\+|(-))?0*(\d)/, "$1$2$3") // Remove unnecessary scientific notation (1e0). .replace(/^([+-]?[\d.]+)e[+-]?0+$/, "$1") // Make sure numbers always start with a digit. .replace(/^([+-])?\./, "$10.") // Remove extraneous trailing decimal zeroes. .replace(/(\.\d+?)0+(?=e|$)/, "$1") // Remove trailing dot. .replace(/\.(?=e|$)/, ""); } /** * @param {string} str * @param {string} target * @returns {number} */ function getMaxContinuousCount(str, target) { const results = str.match(new RegExp("(".concat(escapeStringRegexp(target), ")+"), "g")); if (results === null) { return 0; } return results.reduce((maxCount, result) => Math.max(maxCount, result.length / target.length), 0); } function getMinNotPresentContinuousCount(str, target) { const matches = str.match(new RegExp("(".concat(escapeStringRegexp(target), ")+"), "g")); if (matches === null) { return 0; } const countPresent = new Map(); let max = 0; for (const match of matches) { const count = match.length / target.length; countPresent.set(count, true); if (count > max) { max = count; } } for (let i = 1; i < max; i++) { if (!countPresent.get(i)) { return i; } } return max + 1; } /** * @param {string} text * @returns {number} */ function getStringWidth(text) { if (!text) { return 0; } // shortcut to avoid needless string `RegExp`s, replacements, and allocations within `string-width` if (!notAsciiRegex.test(text)) { return text.length; } return stringWidth_1(text); } function hasIgnoreComment(path) { const node = path.getValue(); return hasNodeIgnoreComment(node); } function hasNodeIgnoreComment(node) { return node && (node.comments && node.comments.length > 0 && node.comments.some(comment => isNodeIgnoreComment(comment) && !comment.unignore) || node.prettierIgnore); } function isNodeIgnoreComment(comment) { return comment.value.trim() === "prettier-ignore"; } function addCommentHelper(node, comment) { const comments = node.comments || (node.comments = []); comments.push(comment); comment.printed = false; // For some reason, TypeScript parses `// x` inside of JSXText as a comment // We already "print" it via the raw text, we don't need to re-print it as a // comment /* istanbul ignore next */ if (node.type === "JSXText") { comment.printed = true; } } function addLeadingComment(node, comment) { comment.leading = true; comment.trailing = false; addCommentHelper(node, comment); } function addDanglingComment(node, comment, marker) { comment.leading = false; comment.trailing = false; if (marker) { comment.marker = marker; } addCommentHelper(node, comment); } function addTrailingComment(node, comment) { comment.leading = false; comment.trailing = true; addCommentHelper(node, comment); } // Not using /* istanbul ignore next */ function isWithinParentArrayProperty(path, propertyName) { const node = path.getValue(); const parent = path.getParentNode(); if (parent == null) { return false; } if (!Array.isArray(parent[propertyName])) { return false; } const key = path.getName(); return parent[propertyName][key] === node; } function replaceEndOfLineWith(text, replacement) { const parts = []; for (const part of text.split("\n")) { if (parts.length !== 0) { parts.push(replacement); } parts.push(part); } return parts; } function getParserName(lang, options) { const supportInfo = support.getSupportInfo({ plugins: options.plugins }); const language = supportInfo.languages.find(language => language.name.toLowerCase() === lang || language.aliases && language.aliases.includes(lang) || language.extensions && language.extensions.some(ext => ext === ".".concat(lang))); if (language) { return language.parsers[0]; } return null; } function isFrontMatterNode(node) { return node && node.type === "front-matter"; } function getShebang(text) { if (!text.startsWith("#!")) { return ""; } const index = text.indexOf("\n"); if (index === -1) { return text; } return text.slice(0, index); } var util = { replaceEndOfLineWith, getStringWidth, getMaxContinuousCount, getMinNotPresentContinuousCount, getParserName, getPenultimate, getLast, getNextNonSpaceNonCommentCharacterIndexWithStartIndex, getNextNonSpaceNonCommentCharacterIndex, getNextNonSpaceNonCommentCharacter, skip, skipWhitespace, skipSpaces, skipToLineEnd, skipEverythingButNewLine, skipInlineComment, skipTrailingComment, skipNewline, isNextLineEmptyAfterIndex, isNextLineEmpty, isPreviousLineEmpty, hasNewline, hasNewlineInRange, hasSpaces, getAlignmentSize, getIndentSize, getPreferredQuote, printString, printNumber, hasIgnoreComment, hasNodeIgnoreComment, isNodeIgnoreComment, makeString, addLeadingComment, addDanglingComment, addTrailingComment, isWithinParentArrayProperty, isFrontMatterNode, getShebang }; function guessEndOfLine(text) { const index = text.indexOf("\r"); if (index >= 0) { return text.charAt(index + 1) === "\n" ? "crlf" : "cr"; } return "lf"; } function convertEndOfLineToChars(value) { switch (value) { case "cr": return "\r"; case "crlf": return "\r\n"; default: return "\n"; } } function countEndOfLineChars(text, eol) { let regex; /* istanbul ignore else */ if (eol === "\n") { regex = /\n/g; } else if (eol === "\r") { regex = /\r/g; } else if (eol === "\r\n") { regex = /\r\n/g; } else { throw new Error("Unexpected \"eol\" ".concat(JSON.stringify(eol), ".")); } const endOfLines = text.match(regex); return endOfLines ? endOfLines.length : 0; } function normalizeEndOfLine(text) { return text.replace(/\r\n?/g, "\n"); } var endOfLine = { guessEndOfLine, convertEndOfLineToChars, countEndOfLineChars, normalizeEndOfLine }; const { getStringWidth: getStringWidth$1 } = util; const { convertEndOfLineToChars: convertEndOfLineToChars$1 } = endOfLine; const { concat: concat$1, fill: fill$1, cursor: cursor$1 } = docBuilders; /** @type {Record} */ let groupModeMap; const MODE_BREAK = 1; const MODE_FLAT = 2; function rootIndent() { return { value: "", length: 0, queue: [] }; } function makeIndent(ind, options) { return generateInd(ind, { type: "indent" }, options); } function makeAlign(indent, n, options) { if (n === -Infinity) { return indent.root || rootIndent(); } if (n < 0) { return generateInd(indent, { type: "dedent" }, options); } if (!n) { return indent; } if (n.type === "root") { return Object.assign({}, indent, { root: indent }); } const alignType = typeof n === "string" ? "stringAlign" : "numberAlign"; return generateInd(indent, { type: alignType, n }, options); } function generateInd(ind, newPart, options) { const queue = newPart.type === "dedent" ? ind.queue.slice(0, -1) : ind.queue.concat(newPart); let value = ""; let length = 0; let lastTabs = 0; let lastSpaces = 0; for (const part of queue) { switch (part.type) { case "indent": flush(); if (options.useTabs) { addTabs(1); } else { addSpaces(options.tabWidth); } break; case "stringAlign": flush(); value += part.n; length += part.n.length; break; case "numberAlign": lastTabs += 1; lastSpaces += part.n; break; /* istanbul ignore next */ default: throw new Error("Unexpected type '".concat(part.type, "'")); } } flushSpaces(); return Object.assign({}, ind, { value, length, queue }); function addTabs(count) { value += "\t".repeat(count); length += options.tabWidth * count; } function addSpaces(count) { value += " ".repeat(count); length += count; } function flush() { if (options.useTabs) { flushTabs(); } else { flushSpaces(); } } function flushTabs() { if (lastTabs > 0) { addTabs(lastTabs); } resetLast(); } function flushSpaces() { if (lastSpaces > 0) { addSpaces(lastSpaces); } resetLast(); } function resetLast() { lastTabs = 0; lastSpaces = 0; } } function trim$1(out) { if (out.length === 0) { return 0; } let trimCount = 0; // Trim whitespace at the end of line while (out.length > 0 && typeof out[out.length - 1] === "string" && out[out.length - 1].match(/^[\t ]*$/)) { trimCount += out.pop().length; } if (out.length && typeof out[out.length - 1] === "string") { const trimmed = out[out.length - 1].replace(/[\t ]*$/, ""); trimCount += out[out.length - 1].length - trimmed.length; out[out.length - 1] = trimmed; } return trimCount; } function fits(next, restCommands, width, options, mustBeFlat) { let restIdx = restCommands.length; const cmds = [next]; // `out` is only used for width counting because `trim` requires to look // backwards for space characters. const out = []; while (width >= 0) { if (cmds.length === 0) { if (restIdx === 0) { return true; } cmds.push(restCommands[restIdx - 1]); restIdx--; continue; } const [ind, mode, doc] = cmds.pop(); if (typeof doc === "string") { out.push(doc); width -= getStringWidth$1(doc); } else { switch (doc.type) { case "concat": for (let i = doc.parts.length - 1; i >= 0; i--) { cmds.push([ind, mode, doc.parts[i]]); } break; case "indent": cmds.push([makeIndent(ind, options), mode, doc.contents]); break; case "align": cmds.push([makeAlign(ind, doc.n, options), mode, doc.contents]); break; case "trim": width += trim$1(out); break; case "group": if (mustBeFlat && doc.break) { return false; } cmds.push([ind, doc.break ? MODE_BREAK : mode, doc.contents]); if (doc.id) { groupModeMap[doc.id] = cmds[cmds.length - 1][1]; } break; case "fill": for (let i = doc.parts.length - 1; i >= 0; i--) { cmds.push([ind, mode, doc.parts[i]]); } break; case "if-break": { const groupMode = doc.groupId ? groupModeMap[doc.groupId] : mode; if (groupMode === MODE_BREAK) { if (doc.breakContents) { cmds.push([ind, mode, doc.breakContents]); } } if (groupMode === MODE_FLAT) { if (doc.flatContents) { cmds.push([ind, mode, doc.flatContents]); } } break; } case "line": switch (mode) { // fallthrough case MODE_FLAT: if (!doc.hard) { if (!doc.soft) { out.push(" "); width -= 1; } break; } return true; case MODE_BREAK: return true; } break; } } } return false; } function printDocToString(doc, options) { groupModeMap = {}; const width = options.printWidth; const newLine = convertEndOfLineToChars$1(options.endOfLine); let pos = 0; // cmds is basically a stack. We've turned a recursive call into a // while loop which is much faster. The while loop below adds new // cmds to the array instead of recursively calling `print`. const cmds = [[rootIndent(), MODE_BREAK, doc]]; const out = []; let shouldRemeasure = false; let lineSuffix = []; while (cmds.length !== 0) { const [ind, mode, doc] = cmds.pop(); if (typeof doc === "string") { const formatted = newLine !== "\n" && doc.includes("\n") ? doc.replace(/\n/g, newLine) : doc; out.push(formatted); pos += getStringWidth$1(formatted); } else { switch (doc.type) { case "cursor": out.push(cursor$1.placeholder); break; case "concat": for (let i = doc.parts.length - 1; i >= 0; i--) { cmds.push([ind, mode, doc.parts[i]]); } break; case "indent": cmds.push([makeIndent(ind, options), mode, doc.contents]); break; case "align": cmds.push([makeAlign(ind, doc.n, options), mode, doc.contents]); break; case "trim": pos -= trim$1(out); break; case "group": switch (mode) { case MODE_FLAT: if (!shouldRemeasure) { cmds.push([ind, doc.break ? MODE_BREAK : MODE_FLAT, doc.contents]); break; } // fallthrough case MODE_BREAK: { shouldRemeasure = false; const next = [ind, MODE_FLAT, doc.contents]; const rem = width - pos; if (!doc.break && fits(next, cmds, rem, options)) { cmds.push(next); } else { // Expanded states are a rare case where a document // can manually provide multiple representations of // itself. It provides an array of documents // going from the least expanded (most flattened) // representation first to the most expanded. If a // group has these, we need to manually go through // these states and find the first one that fits. if (doc.expandedStates) { const mostExpanded = doc.expandedStates[doc.expandedStates.length - 1]; if (doc.break) { cmds.push([ind, MODE_BREAK, mostExpanded]); break; } else { for (let i = 1; i < doc.expandedStates.length + 1; i++) { if (i >= doc.expandedStates.length) { cmds.push([ind, MODE_BREAK, mostExpanded]); break; } else { const state = doc.expandedStates[i]; const cmd = [ind, MODE_FLAT, state]; if (fits(cmd, cmds, rem, options)) { cmds.push(cmd); break; } } } } } else { cmds.push([ind, MODE_BREAK, doc.contents]); } } break; } } if (doc.id) { groupModeMap[doc.id] = cmds[cmds.length - 1][1]; } break; // Fills each line with as much code as possible before moving to a new // line with the same indentation. // // Expects doc.parts to be an array of alternating content and // whitespace. The whitespace contains the linebreaks. // // For example: // ["I", line, "love", line, "monkeys"] // or // [{ type: group, ... }, softline, { type: group, ... }] // // It uses this parts structure to handle three main layout cases: // * The first two content items fit on the same line without // breaking // -> output the first content item and the whitespace "flat". // * Only the first content item fits on the line without breaking // -> output the first content item "flat" and the whitespace with // "break". // * Neither content item fits on the line without breaking // -> output the first content item and the whitespace with "break". case "fill": { const rem = width - pos; const { parts } = doc; if (parts.length === 0) { break; } const [content, whitespace] = parts; const contentFlatCmd = [ind, MODE_FLAT, content]; const contentBreakCmd = [ind, MODE_BREAK, content]; const contentFits = fits(contentFlatCmd, [], rem, options, true); if (parts.length === 1) { if (contentFits) { cmds.push(contentFlatCmd); } else { cmds.push(contentBreakCmd); } break; } const whitespaceFlatCmd = [ind, MODE_FLAT, whitespace]; const whitespaceBreakCmd = [ind, MODE_BREAK, whitespace]; if (parts.length === 2) { if (contentFits) { cmds.push(whitespaceFlatCmd); cmds.push(contentFlatCmd); } else { cmds.push(whitespaceBreakCmd); cmds.push(contentBreakCmd); } break; } // At this point we've handled the first pair (context, separator) // and will create a new fill doc for the rest of the content. // Ideally we wouldn't mutate the array here but copying all the // elements to a new array would make this algorithm quadratic, // which is unusable for large arrays (e.g. large texts in JSX). parts.splice(0, 2); const remainingCmd = [ind, mode, fill$1(parts)]; const secondContent = parts[0]; const firstAndSecondContentFlatCmd = [ind, MODE_FLAT, concat$1([content, whitespace, secondContent])]; const firstAndSecondContentFits = fits(firstAndSecondContentFlatCmd, [], rem, options, true); if (firstAndSecondContentFits) { cmds.push(remainingCmd); cmds.push(whitespaceFlatCmd); cmds.push(contentFlatCmd); } else if (contentFits) { cmds.push(remainingCmd); cmds.push(whitespaceBreakCmd); cmds.push(contentFlatCmd); } else { cmds.push(remainingCmd); cmds.push(whitespaceBreakCmd); cmds.push(contentBreakCmd); } break; } case "if-break": { const groupMode = doc.groupId ? groupModeMap[doc.groupId] : mode; if (groupMode === MODE_BREAK) { if (doc.breakContents) { cmds.push([ind, mode, doc.breakContents]); } } if (groupMode === MODE_FLAT) { if (doc.flatContents) { cmds.push([ind, mode, doc.flatContents]); } } break; } case "line-suffix": lineSuffix.push([ind, mode, doc.contents]); break; case "line-suffix-boundary": if (lineSuffix.length > 0) { cmds.push([ind, mode, { type: "line", hard: true }]); } break; case "line": switch (mode) { case MODE_FLAT: if (!doc.hard) { if (!doc.soft) { out.push(" "); pos += 1; } break; } else { // This line was forced into the output even if we // were in flattened mode, so we need to tell the next // group that no matter what, it needs to remeasure // because the previous measurement didn't accurately // capture the entire expression (this is necessary // for nested groups) shouldRemeasure = true; } // fallthrough case MODE_BREAK: if (lineSuffix.length) { cmds.push([ind, mode, doc]); cmds.push(...lineSuffix.reverse()); lineSuffix = []; break; } if (doc.literal) { if (ind.root) { out.push(newLine, ind.root.value); pos = ind.root.length; } else { out.push(newLine); pos = 0; } } else { pos -= trim$1(out); out.push(newLine + ind.value); pos = ind.length; } break; } break; } } } const cursorPlaceholderIndex = out.indexOf(cursor$1.placeholder); if (cursorPlaceholderIndex !== -1) { const otherCursorPlaceholderIndex = out.indexOf(cursor$1.placeholder, cursorPlaceholderIndex + 1); const beforeCursor = out.slice(0, cursorPlaceholderIndex).join(""); const aroundCursor = out.slice(cursorPlaceholderIndex + 1, otherCursorPlaceholderIndex).join(""); const afterCursor = out.slice(otherCursorPlaceholderIndex + 1).join(""); return { formatted: beforeCursor + aroundCursor + afterCursor, cursorNodeStart: beforeCursor.length, cursorNodeText: aroundCursor }; } return { formatted: out.join("") }; } var docPrinter = { printDocToString }; const traverseDocOnExitStackMarker = {}; function traverseDoc(doc, onEnter, onExit, shouldTraverseConditionalGroups) { const docsStack = [doc]; while (docsStack.length !== 0) { const doc = docsStack.pop(); if (doc === traverseDocOnExitStackMarker) { onExit(docsStack.pop()); continue; } if (onExit) { docsStack.push(doc, traverseDocOnExitStackMarker); } if ( // Should Recurse !onEnter || onEnter(doc) !== false) { // When there are multiple parts to process, // the parts need to be pushed onto the stack in reverse order, // so that they are processed in the original order // when the stack is popped. if (doc.type === "concat" || doc.type === "fill") { for (let ic = doc.parts.length, i = ic - 1; i >= 0; --i) { docsStack.push(doc.parts[i]); } } else if (doc.type === "if-break") { if (doc.flatContents) { docsStack.push(doc.flatContents); } if (doc.breakContents) { docsStack.push(doc.breakContents); } } else if (doc.type === "group" && doc.expandedStates) { if (shouldTraverseConditionalGroups) { for (let ic = doc.expandedStates.length, i = ic - 1; i >= 0; --i) { docsStack.push(doc.expandedStates[i]); } } else { docsStack.push(doc.contents); } } else if (doc.contents) { docsStack.push(doc.contents); } } } } function mapDoc(doc, cb) { if (doc.type === "concat" || doc.type === "fill") { const parts = doc.parts.map(part => mapDoc(part, cb)); return cb(Object.assign({}, doc, { parts })); } else if (doc.type === "if-break") { const breakContents = doc.breakContents && mapDoc(doc.breakContents, cb); const flatContents = doc.flatContents && mapDoc(doc.flatContents, cb); return cb(Object.assign({}, doc, { breakContents, flatContents })); } else if (doc.contents) { const contents = mapDoc(doc.contents, cb); return cb(Object.assign({}, doc, { contents })); } return cb(doc); } function findInDoc(doc, fn, defaultValue) { let result = defaultValue; let hasStopped = false; function findInDocOnEnterFn(doc) { const maybeResult = fn(doc); if (maybeResult !== undefined) { hasStopped = true; result = maybeResult; } if (hasStopped) { return false; } } traverseDoc(doc, findInDocOnEnterFn); return result; } function isEmpty(n) { return typeof n === "string" && n.length === 0; } function isLineNextFn(doc) { if (typeof doc === "string") { return false; } if (doc.type === "line") { return true; } } function isLineNext(doc) { return findInDoc(doc, isLineNextFn, false); } function willBreakFn(doc) { if (doc.type === "group" && doc.break) { return true; } if (doc.type === "line" && doc.hard) { return true; } if (doc.type === "break-parent") { return true; } } function willBreak(doc) { return findInDoc(doc, willBreakFn, false); } function breakParentGroup(groupStack) { if (groupStack.length > 0) { const parentGroup = groupStack[groupStack.length - 1]; // Breaks are not propagated through conditional groups because // the user is expected to manually handle what breaks. if (!parentGroup.expandedStates) { parentGroup.break = true; } } return null; } function propagateBreaks(doc) { const alreadyVisitedSet = new Set(); const groupStack = []; function propagateBreaksOnEnterFn(doc) { if (doc.type === "break-parent") { breakParentGroup(groupStack); } if (doc.type === "group") { groupStack.push(doc); if (alreadyVisitedSet.has(doc)) { return false; } alreadyVisitedSet.add(doc); } } function propagateBreaksOnExitFn(doc) { if (doc.type === "group") { const group = groupStack.pop(); if (group.break) { breakParentGroup(groupStack); } } } traverseDoc(doc, propagateBreaksOnEnterFn, propagateBreaksOnExitFn, /* shouldTraverseConditionalGroups */ true); } function removeLinesFn(doc) { // Force this doc into flat mode by statically converting all // lines into spaces (or soft lines into nothing). Hard lines // should still output because there's too great of a chance // of breaking existing assumptions otherwise. if (doc.type === "line" && !doc.hard) { return doc.soft ? "" : " "; } else if (doc.type === "if-break") { return doc.flatContents || ""; } return doc; } function removeLines(doc) { return mapDoc(doc, removeLinesFn); } function getInnerParts(doc) { let { parts } = doc; let lastPart; // Avoid a falsy element like "" for (let i = doc.parts.length; i > 0 && !lastPart; i--) { lastPart = parts[i - 1]; } if (lastPart.type === "group") { parts = lastPart.contents.parts; } return parts; } function stripTrailingHardline(doc, withInnerParts = false) { // HACK remove ending hardline, original PR: #1984 if (doc.type === "concat" && doc.parts.length !== 0) { const parts = withInnerParts ? getInnerParts(doc) : doc.parts; const lastPart = parts[parts.length - 1]; if (lastPart.type === "concat") { if (lastPart.parts.length === 2 && lastPart.parts[0].hard && lastPart.parts[1].type === "break-parent") { return { type: "concat", parts: parts.slice(0, -1) }; } return { type: "concat", parts: doc.parts.slice(0, -1).concat(stripTrailingHardline(lastPart)) }; } } return doc; } function normalizeParts(parts) { const newParts = []; const restParts = parts.filter(Boolean); while (restParts.length !== 0) { const part = restParts.shift(); if (!part) { continue; } if (part.type === "concat") { restParts.unshift(...part.parts); continue; } if (newParts.length !== 0 && typeof newParts[newParts.length - 1] === "string" && typeof part === "string") { newParts[newParts.length - 1] += part; continue; } newParts.push(part); } return newParts; } function normalizeDoc(doc) { return mapDoc(doc, currentDoc => { if (!currentDoc.parts) { return currentDoc; } return Object.assign({}, currentDoc, { parts: normalizeParts(currentDoc.parts) }); }); } var docUtils = { isEmpty, willBreak, isLineNext, traverseDoc, findInDoc, mapDoc, propagateBreaks, removeLines, stripTrailingHardline, normalizeParts, normalizeDoc }; function flattenDoc(doc) { if (doc.type === "concat") { const res = []; for (let i = 0; i < doc.parts.length; ++i) { const doc2 = doc.parts[i]; if (typeof doc2 !== "string" && doc2.type === "concat") { res.push(...flattenDoc(doc2).parts); } else { const flattened = flattenDoc(doc2); if (flattened !== "") { res.push(flattened); } } } return Object.assign({}, doc, { parts: res }); } else if (doc.type === "if-break") { return Object.assign({}, doc, { breakContents: doc.breakContents != null ? flattenDoc(doc.breakContents) : null, flatContents: doc.flatContents != null ? flattenDoc(doc.flatContents) : null }); } else if (doc.type === "group") { return Object.assign({}, doc, { contents: flattenDoc(doc.contents), expandedStates: doc.expandedStates ? doc.expandedStates.map(flattenDoc) : doc.expandedStates }); } else if (doc.contents) { return Object.assign({}, doc, { contents: flattenDoc(doc.contents) }); } return doc; } function printDoc(doc) { if (typeof doc === "string") { return JSON.stringify(doc); } if (doc.type === "line") { if (doc.literal) { return "literalline"; } if (doc.hard) { return "hardline"; } if (doc.soft) { return "softline"; } return "line"; } if (doc.type === "break-parent") { return "breakParent"; } if (doc.type === "trim") { return "trim"; } if (doc.type === "concat") { return "[" + doc.parts.map(printDoc).join(", ") + "]"; } if (doc.type === "indent") { return "indent(" + printDoc(doc.contents) + ")"; } if (doc.type === "align") { return doc.n === -Infinity ? "dedentToRoot(" + printDoc(doc.contents) + ")" : doc.n < 0 ? "dedent(" + printDoc(doc.contents) + ")" : doc.n.type === "root" ? "markAsRoot(" + printDoc(doc.contents) + ")" : "align(" + JSON.stringify(doc.n) + ", " + printDoc(doc.contents) + ")"; } if (doc.type === "if-break") { return "ifBreak(" + printDoc(doc.breakContents) + (doc.flatContents ? ", " + printDoc(doc.flatContents) : "") + ")"; } if (doc.type === "group") { if (doc.expandedStates) { return "conditionalGroup(" + "[" + doc.expandedStates.map(printDoc).join(",") + "])"; } return (doc.break ? "wrappedGroup" : "group") + "(" + printDoc(doc.contents) + ")"; } if (doc.type === "fill") { return "fill" + "(" + doc.parts.map(printDoc).join(", ") + ")"; } if (doc.type === "line-suffix") { return "lineSuffix(" + printDoc(doc.contents) + ")"; } if (doc.type === "line-suffix-boundary") { return "lineSuffixBoundary"; } throw new Error("Unknown doc type " + doc.type); } var docDebug = { printDocToDebug(doc) { return printDoc(flattenDoc(doc)); } }; /** * @typedef {import("./doc-builders").Doc} Doc */ var document = { builders: docBuilders, printer: docPrinter, utils: docUtils, debug: docDebug }; return document; })));