/*! fflate - fast JavaScript compression/decompression Licensed under MIT. https://github.com/101arrowz/fflate/blob/master/LICENSE version 0.8.2 */ // DEFLATE is a complex format; to read this code, you should probably check the RFC first: // https://tools.ietf.org/html/rfc1951 // You may also wish to take a look at the guide I made about this program: // https://gist.github.com/101arrowz/253f31eb5abc3d9275ab943003ffecad // Some of the following code is similar to that of UZIP.js: // https://github.com/photopea/UZIP.js // However, the vast majority of the codebase has diverged from UZIP.js to increase performance and reduce bundle size. // Sometimes 0 will appear where -1 would be more appropriate. This is because using a uint // is better for memory in most engines (I *think*). var ch2 = {}; var wk = (function (c, id, msg, transfer, cb) { var w = new Worker(ch2[id] || (ch2[id] = URL.createObjectURL(new Blob([ c + ';addEventListener("error",function(e){e=e.error;postMessage({$e$:[e.message,e.code,e.stack]})})' ], { type: 'text/javascript' })))); w.onmessage = function (e) { var d = e.data, ed = d.$e$; if (ed) { var err = new Error(ed[0]); err['code'] = ed[1]; err.stack = ed[2]; cb(err, null); } else cb(null, d); }; w.postMessage(msg, transfer); return w; }); // aliases for shorter compressed code (most minifers don't do this) var u8 = Uint8Array, u16 = Uint16Array, i32 = Int32Array; // fixed length extra bits var fleb = new u8([0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 2, 2, 2, 2, 3, 3, 3, 3, 4, 4, 4, 4, 5, 5, 5, 5, 0, /* unused */ 0, 0, /* impossible */ 0]); // fixed distance extra bits var fdeb = new u8([0, 0, 0, 0, 1, 1, 2, 2, 3, 3, 4, 4, 5, 5, 6, 6, 7, 7, 8, 8, 9, 9, 10, 10, 11, 11, 12, 12, 13, 13, /* unused */ 0, 0]); // code length index map var clim = new u8([16, 17, 18, 0, 8, 7, 9, 6, 10, 5, 11, 4, 12, 3, 13, 2, 14, 1, 15]); // get base, reverse index map from extra bits var freb = function (eb, start) { var b = new u16(31); for (var i = 0; i < 31; ++i) { b[i] = start += 1 << eb[i - 1]; } // numbers here are at max 18 bits var r = new i32(b[30]); for (var i = 1; i < 30; ++i) { for (var j = b[i]; j < b[i + 1]; ++j) { r[j] = ((j - b[i]) << 5) | i; } } return { b: b, r: r }; }; var _a = freb(fleb, 2), fl = _a.b, revfl = _a.r; // we can ignore the fact that the other numbers are wrong; they never happen anyway fl[28] = 258, revfl[258] = 28; var _b = freb(fdeb, 0), fd = _b.b, revfd = _b.r; // map of value to reverse (assuming 16 bits) var rev = new u16(32768); for (var i = 0; i < 32768; ++i) { // reverse table algorithm from SO var x = ((i & 0xAAAA) >> 1) | ((i & 0x5555) << 1); x = ((x & 0xCCCC) >> 2) | ((x & 0x3333) << 2); x = ((x & 0xF0F0) >> 4) | ((x & 0x0F0F) << 4); rev[i] = (((x & 0xFF00) >> 8) | ((x & 0x00FF) << 8)) >> 1; } // create huffman tree from u8 "map": index -> code length for code index // mb (max bits) must be at most 15 // TODO: optimize/split up? var hMap = (function (cd, mb, r) { var s = cd.length; // index var i = 0; // u16 "map": index -> # of codes with bit length = index var l = new u16(mb); // length of cd must be 288 (total # of codes) for (; i < s; ++i) { if (cd[i]) ++l[cd[i] - 1]; } // u16 "map": index -> minimum code for bit length = index var le = new u16(mb); for (i = 1; i < mb; ++i) { le[i] = (le[i - 1] + l[i - 1]) << 1; } var co; if (r) { // u16 "map": index -> number of actual bits, symbol for code co = new u16(1 << mb); // bits to remove for reverser var rvb = 15 - mb; for (i = 0; i < s; ++i) { // ignore 0 lengths if (cd[i]) { // num encoding both symbol and bits read var sv = (i << 4) | cd[i]; // free bits var r_1 = mb - cd[i]; // start value var v = le[cd[i] - 1]++ << r_1; // m is end value for (var m = v | ((1 << r_1) - 1); v <= m; ++v) { // every 16 bit value starting with the code yields the same result co[rev[v] >> rvb] = sv; } } } } else { co = new u16(s); for (i = 0; i < s; ++i) { if (cd[i]) { co[i] = rev[le[cd[i] - 1]++] >> (15 - cd[i]); } } } return co; }); // fixed length tree var flt = new u8(288); for (var i = 0; i < 144; ++i) flt[i] = 8; for (var i = 144; i < 256; ++i) flt[i] = 9; for (var i = 256; i < 280; ++i) flt[i] = 7; for (var i = 280; i < 288; ++i) flt[i] = 8; // fixed distance tree var fdt = new u8(32); for (var i = 0; i < 32; ++i) fdt[i] = 5; // fixed length map var flm = /*#__PURE__*/ hMap(flt, 9, 0), flrm = /*#__PURE__*/ hMap(flt, 9, 1); // fixed distance map var fdm = /*#__PURE__*/ hMap(fdt, 5, 0), fdrm = /*#__PURE__*/ hMap(fdt, 5, 1); // find max of array var max = function (a) { var m = a[0]; for (var i = 1; i < a.length; ++i) { if (a[i] > m) m = a[i]; } return m; }; // read d, starting at bit p and mask with m var bits = function (d, p, m) { var o = (p / 8) | 0; return ((d[o] | (d[o + 1] << 8)) >> (p & 7)) & m; }; // read d, starting at bit p continuing for at least 16 bits var bits16 = function (d, p) { var o = (p / 8) | 0; return ((d[o] | (d[o + 1] << 8) | (d[o + 2] << 16)) >> (p & 7)); }; // get end of byte var shft = function (p) { return ((p + 7) / 8) | 0; }; // typed array slice - allows garbage collector to free original reference, // while being more compatible than .slice var slc = function (v, s, e) { if (s == null || s < 0) s = 0; if (e == null || e > v.length) e = v.length; // can't use .constructor in case user-supplied return new u8(v.subarray(s, e)); }; /** * Codes for errors generated within this library */ export var FlateErrorCode = { UnexpectedEOF: 0, InvalidBlockType: 1, InvalidLengthLiteral: 2, InvalidDistance: 3, StreamFinished: 4, NoStreamHandler: 5, InvalidHeader: 6, NoCallback: 7, InvalidUTF8: 8, ExtraFieldTooLong: 9, InvalidDate: 10, FilenameTooLong: 11, StreamFinishing: 12, InvalidZipData: 13, UnknownCompressionMethod: 14 }; // error codes var ec = [ 'unexpected EOF', 'invalid block type', 'invalid length/literal', 'invalid distance', 'stream finished', 'no stream handler', , 'no callback', 'invalid UTF-8 data', 'extra field too long', 'date not in range 1980-2099', 'filename too long', 'stream finishing', 'invalid zip data' // determined by unknown compression method ]; ; var err = function (ind, msg, nt) { var e = new Error(msg || ec[ind]); e.code = ind; if (Error.captureStackTrace) Error.captureStackTrace(e, err); if (!nt) throw e; return e; }; // expands raw DEFLATE data var inflt = function (dat, st, buf, dict) { // source length dict length var sl = dat.length, dl = dict ? dict.length : 0; if (!sl || st.f && !st.l) return buf || new u8(0); var noBuf = !buf; // have to estimate size var resize = noBuf || st.i != 2; // no state var noSt = st.i; // Assumes roughly 33% compression ratio average if (noBuf) buf = new u8(sl * 3); // ensure buffer can fit at least l elements var cbuf = function (l) { var bl = buf.length; // need to increase size to fit if (l > bl) { // Double or set to necessary, whichever is greater var nbuf = new u8(Math.max(bl * 2, l)); nbuf.set(buf); buf = nbuf; } }; // last chunk bitpos bytes var final = st.f || 0, pos = st.p || 0, bt = st.b || 0, lm = st.l, dm = st.d, lbt = st.m, dbt = st.n; // total bits var tbts = sl * 8; do { if (!lm) { // BFINAL - this is only 1 when last chunk is next final = bits(dat, pos, 1); // type: 0 = no compression, 1 = fixed huffman, 2 = dynamic huffman var type = bits(dat, pos + 1, 3); pos += 3; if (!type) { // go to end of byte boundary var s = shft(pos) + 4, l = dat[s - 4] | (dat[s - 3] << 8), t = s + l; if (t > sl) { if (noSt) err(0); break; } // ensure size if (resize) cbuf(bt + l); // Copy over uncompressed data buf.set(dat.subarray(s, t), bt); // Get new bitpos, update byte count st.b = bt += l, st.p = pos = t * 8, st.f = final; continue; } else if (type == 1) lm = flrm, dm = fdrm, lbt = 9, dbt = 5; else if (type == 2) { // literal lengths var hLit = bits(dat, pos, 31) + 257, hcLen = bits(dat, pos + 10, 15) + 4; var tl = hLit + bits(dat, pos + 5, 31) + 1; pos += 14; // length+distance tree var ldt = new u8(tl); // code length tree var clt = new u8(19); for (var i = 0; i < hcLen; ++i) { // use index map to get real code clt[clim[i]] = bits(dat, pos + i * 3, 7); } pos += hcLen * 3; // code lengths bits var clb = max(clt), clbmsk = (1 << clb) - 1; // code lengths map var clm = hMap(clt, clb, 1); for (var i = 0; i < tl;) { var r = clm[bits(dat, pos, clbmsk)]; // bits read pos += r & 15; // symbol var s = r >> 4; // code length to copy if (s < 16) { ldt[i++] = s; } else { // copy count var c = 0, n = 0; if (s == 16) n = 3 + bits(dat, pos, 3), pos += 2, c = ldt[i - 1]; else if (s == 17) n = 3 + bits(dat, pos, 7), pos += 3; else if (s == 18) n = 11 + bits(dat, pos, 127), pos += 7; while (n--) ldt[i++] = c; } } // length tree distance tree var lt = ldt.subarray(0, hLit), dt = ldt.subarray(hLit); // max length bits lbt = max(lt); // max dist bits dbt = max(dt); lm = hMap(lt, lbt, 1); dm = hMap(dt, dbt, 1); } else err(1); if (pos > tbts) { if (noSt) err(0); break; } } // Make sure the buffer can hold this + the largest possible addition // Maximum chunk size (practically, theoretically infinite) is 2^17 if (resize) cbuf(bt + 131072); var lms = (1 << lbt) - 1, dms = (1 << dbt) - 1; var lpos = pos; for (;; lpos = pos) { // bits read, code var c = lm[bits16(dat, pos) & lms], sym = c >> 4; pos += c & 15; if (pos > tbts) { if (noSt) err(0); break; } if (!c) err(2); if (sym < 256) buf[bt++] = sym; else if (sym == 256) { lpos = pos, lm = null; break; } else { var add = sym - 254; // no extra bits needed if less if (sym > 264) { // index var i = sym - 257, b = fleb[i]; add = bits(dat, pos, (1 << b) - 1) + fl[i]; pos += b; } // dist var d = dm[bits16(dat, pos) & dms], dsym = d >> 4; if (!d) err(3); pos += d & 15; var dt = fd[dsym]; if (dsym > 3) { var b = fdeb[dsym]; dt += bits16(dat, pos) & (1 << b) - 1, pos += b; } if (pos > tbts) { if (noSt) err(0); break; } if (resize) cbuf(bt + 131072); var end = bt + add; if (bt < dt) { var shift = dl - dt, dend = Math.min(dt, end); if (shift + bt < 0) err(3); for (; bt < dend; ++bt) buf[bt] = dict[shift + bt]; } for (; bt < end; ++bt) buf[bt] = buf[bt - dt]; } } st.l = lm, st.p = lpos, st.b = bt, st.f = final; if (lm) final = 1, st.m = lbt, st.d = dm, st.n = dbt; } while (!final); // don't reallocate for streams or user buffers return bt != buf.length && noBuf ? slc(buf, 0, bt) : buf.subarray(0, bt); }; // starting at p, write the minimum number of bits that can hold v to d var wbits = function (d, p, v) { v <<= p & 7; var o = (p / 8) | 0; d[o] |= v; d[o + 1] |= v >> 8; }; // starting at p, write the minimum number of bits (>8) that can hold v to d var wbits16 = function (d, p, v) { v <<= p & 7; var o = (p / 8) | 0; d[o] |= v; d[o + 1] |= v >> 8; d[o + 2] |= v >> 16; }; // creates code lengths from a frequency table var hTree = function (d, mb) { // Need extra info to make a tree var t = []; for (var i = 0; i < d.length; ++i) { if (d[i]) t.push({ s: i, f: d[i] }); } var s = t.length; var t2 = t.slice(); if (!s) return { t: et, l: 0 }; if (s == 1) { var v = new u8(t[0].s + 1); v[t[0].s] = 1; return { t: v, l: 1 }; } t.sort(function (a, b) { return a.f - b.f; }); // after i2 reaches last ind, will be stopped // freq must be greater than largest possible number of symbols t.push({ s: -1, f: 25001 }); var l = t[0], r = t[1], i0 = 0, i1 = 1, i2 = 2; t[0] = { s: -1, f: l.f + r.f, l: l, r: r }; // efficient algorithm from UZIP.js // i0 is lookbehind, i2 is lookahead - after processing two low-freq // symbols that combined have high freq, will start processing i2 (high-freq, // non-composite) symbols instead // see https://reddit.com/r/photopea/comments/ikekht/uzipjs_questions/ while (i1 != s - 1) { l = t[t[i0].f < t[i2].f ? i0++ : i2++]; r = t[i0 != i1 && t[i0].f < t[i2].f ? i0++ : i2++]; t[i1++] = { s: -1, f: l.f + r.f, l: l, r: r }; } var maxSym = t2[0].s; for (var i = 1; i < s; ++i) { if (t2[i].s > maxSym) maxSym = t2[i].s; } // code lengths var tr = new u16(maxSym + 1); // max bits in tree var mbt = ln(t[i1 - 1], tr, 0); if (mbt > mb) { // more algorithms from UZIP.js // TODO: find out how this code works (debt) // ind debt var i = 0, dt = 0; // left cost var lft = mbt - mb, cst = 1 << lft; t2.sort(function (a, b) { return tr[b.s] - tr[a.s] || a.f - b.f; }); for (; i < s; ++i) { var i2_1 = t2[i].s; if (tr[i2_1] > mb) { dt += cst - (1 << (mbt - tr[i2_1])); tr[i2_1] = mb; } else break; } dt >>= lft; while (dt > 0) { var i2_2 = t2[i].s; if (tr[i2_2] < mb) dt -= 1 << (mb - tr[i2_2]++ - 1); else ++i; } for (; i >= 0 && dt; --i) { var i2_3 = t2[i].s; if (tr[i2_3] == mb) { --tr[i2_3]; ++dt; } } mbt = mb; } return { t: new u8(tr), l: mbt }; }; // get the max length and assign length codes var ln = function (n, l, d) { return n.s == -1 ? Math.max(ln(n.l, l, d + 1), ln(n.r, l, d + 1)) : (l[n.s] = d); }; // length codes generation var lc = function (c) { var s = c.length; // Note that the semicolon was intentional while (s && !c[--s]) ; var cl = new u16(++s); // ind num streak var cli = 0, cln = c[0], cls = 1; var w = function (v) { cl[cli++] = v; }; for (var i = 1; i <= s; ++i) { if (c[i] == cln && i != s) ++cls; else { if (!cln && cls > 2) { for (; cls > 138; cls -= 138) w(32754); if (cls > 2) { w(cls > 10 ? ((cls - 11) << 5) | 28690 : ((cls - 3) << 5) | 12305); cls = 0; } } else if (cls > 3) { w(cln), --cls; for (; cls > 6; cls -= 6) w(8304); if (cls > 2) w(((cls - 3) << 5) | 8208), cls = 0; } while (cls--) w(cln); cls = 1; cln = c[i]; } } return { c: cl.subarray(0, cli), n: s }; }; // calculate the length of output from tree, code lengths var clen = function (cf, cl) { var l = 0; for (var i = 0; i < cl.length; ++i) l += cf[i] * cl[i]; return l; }; // writes a fixed block // returns the new bit pos var wfblk = function (out, pos, dat) { // no need to write 00 as type: TypedArray defaults to 0 var s = dat.length; var o = shft(pos + 2); out[o] = s & 255; out[o + 1] = s >> 8; out[o + 2] = out[o] ^ 255; out[o + 3] = out[o + 1] ^ 255; for (var i = 0; i < s; ++i) out[o + i + 4] = dat[i]; return (o + 4 + s) * 8; }; // writes a block var wblk = function (dat, out, final, syms, lf, df, eb, li, bs, bl, p) { wbits(out, p++, final); ++lf[256]; var _a = hTree(lf, 15), dlt = _a.t, mlb = _a.l; var _b = hTree(df, 15), ddt = _b.t, mdb = _b.l; var _c = lc(dlt), lclt = _c.c, nlc = _c.n; var _d = lc(ddt), lcdt = _d.c, ndc = _d.n; var lcfreq = new u16(19); for (var i = 0; i < lclt.length; ++i) ++lcfreq[lclt[i] & 31]; for (var i = 0; i < lcdt.length; ++i) ++lcfreq[lcdt[i] & 31]; var _e = hTree(lcfreq, 7), lct = _e.t, mlcb = _e.l; var nlcc = 19; for (; nlcc > 4 && !lct[clim[nlcc - 1]]; --nlcc) ; var flen = (bl + 5) << 3; var ftlen = clen(lf, flt) + clen(df, fdt) + eb; var dtlen = clen(lf, dlt) + clen(df, ddt) + eb + 14 + 3 * nlcc + clen(lcfreq, lct) + 2 * lcfreq[16] + 3 * lcfreq[17] + 7 * lcfreq[18]; if (bs >= 0 && flen <= ftlen && flen <= dtlen) return wfblk(out, p, dat.subarray(bs, bs + bl)); var lm, ll, dm, dl; wbits(out, p, 1 + (dtlen < ftlen)), p += 2; if (dtlen < ftlen) { lm = hMap(dlt, mlb, 0), ll = dlt, dm = hMap(ddt, mdb, 0), dl = ddt; var llm = hMap(lct, mlcb, 0); wbits(out, p, nlc - 257); wbits(out, p + 5, ndc - 1); wbits(out, p + 10, nlcc - 4); p += 14; for (var i = 0; i < nlcc; ++i) wbits(out, p + 3 * i, lct[clim[i]]); p += 3 * nlcc; var lcts = [lclt, lcdt]; for (var it = 0; it < 2; ++it) { var clct = lcts[it]; for (var i = 0; i < clct.length; ++i) { var len = clct[i] & 31; wbits(out, p, llm[len]), p += lct[len]; if (len > 15) wbits(out, p, (clct[i] >> 5) & 127), p += clct[i] >> 12; } } } else { lm = flm, ll = flt, dm = fdm, dl = fdt; } for (var i = 0; i < li; ++i) { var sym = syms[i]; if (sym > 255) { var len = (sym >> 18) & 31; wbits16(out, p, lm[len + 257]), p += ll[len + 257]; if (len > 7) wbits(out, p, (sym >> 23) & 31), p += fleb[len]; var dst = sym & 31; wbits16(out, p, dm[dst]), p += dl[dst]; if (dst > 3) wbits16(out, p, (sym >> 5) & 8191), p += fdeb[dst]; } else { wbits16(out, p, lm[sym]), p += ll[sym]; } } wbits16(out, p, lm[256]); return p + ll[256]; }; // deflate options (nice << 13) | chain var deo = /*#__PURE__*/ new i32([65540, 131080, 131088, 131104, 262176, 1048704, 1048832, 2114560, 2117632]); // empty var et = /*#__PURE__*/ new u8(0); // compresses data into a raw DEFLATE buffer var dflt = function (dat, lvl, plvl, pre, post, st) { var s = st.z || dat.length; var o = new u8(pre + s + 5 * (1 + Math.ceil(s / 7000)) + post); // writing to this writes to the output buffer var w = o.subarray(pre, o.length - post); var lst = st.l; var pos = (st.r || 0) & 7; if (lvl) { if (pos) w[0] = st.r >> 3; var opt = deo[lvl - 1]; var n = opt >> 13, c = opt & 8191; var msk_1 = (1 << plvl) - 1; // prev 2-byte val map curr 2-byte val map var prev = st.p || new u16(32768), head = st.h || new u16(msk_1 + 1); var bs1_1 = Math.ceil(plvl / 3), bs2_1 = 2 * bs1_1; var hsh = function (i) { return (dat[i] ^ (dat[i + 1] << bs1_1) ^ (dat[i + 2] << bs2_1)) & msk_1; }; // 24576 is an arbitrary number of maximum symbols per block // 424 buffer for last block var syms = new i32(25000); // length/literal freq distance freq var lf = new u16(288), df = new u16(32); // l/lcnt exbits index l/lind waitdx blkpos var lc_1 = 0, eb = 0, i = st.i || 0, li = 0, wi = st.w || 0, bs = 0; for (; i + 2 < s; ++i) { // hash value var hv = hsh(i); // index mod 32768 previous index mod var imod = i & 32767, pimod = head[hv]; prev[imod] = pimod; head[hv] = imod; // We always should modify head and prev, but only add symbols if // this data is not yet processed ("wait" for wait index) if (wi <= i) { // bytes remaining var rem = s - i; if ((lc_1 > 7000 || li > 24576) && (rem > 423 || !lst)) { pos = wblk(dat, w, 0, syms, lf, df, eb, li, bs, i - bs, pos); li = lc_1 = eb = 0, bs = i; for (var j = 0; j < 286; ++j) lf[j] = 0; for (var j = 0; j < 30; ++j) df[j] = 0; } // len dist chain var l = 2, d = 0, ch_1 = c, dif = imod - pimod & 32767; if (rem > 2 && hv == hsh(i - dif)) { var maxn = Math.min(n, rem) - 1; var maxd = Math.min(32767, i); // max possible length // not capped at dif because decompressors implement "rolling" index population var ml = Math.min(258, rem); while (dif <= maxd && --ch_1 && imod != pimod) { if (dat[i + l] == dat[i + l - dif]) { var nl = 0; for (; nl < ml && dat[i + nl] == dat[i + nl - dif]; ++nl) ; if (nl > l) { l = nl, d = dif; // break out early when we reach "nice" (we are satisfied enough) if (nl > maxn) break; // now, find the rarest 2-byte sequence within this // length of literals and search for that instead. // Much faster than just using the start var mmd = Math.min(dif, nl - 2); var md = 0; for (var j = 0; j < mmd; ++j) { var ti = i - dif + j & 32767; var pti = prev[ti]; var cd = ti - pti & 32767; if (cd > md) md = cd, pimod = ti; } } } // check the previous match imod = pimod, pimod = prev[imod]; dif += imod - pimod & 32767; } } // d will be nonzero only when a match was found if (d) { // store both dist and len data in one int32 // Make sure this is recognized as a len/dist with 28th bit (2^28) syms[li++] = 268435456 | (revfl[l] << 18) | revfd[d]; var lin = revfl[l] & 31, din = revfd[d] & 31; eb += fleb[lin] + fdeb[din]; ++lf[257 + lin]; ++df[din]; wi = i + l; ++lc_1; } else { syms[li++] = dat[i]; ++lf[dat[i]]; } } } for (i = Math.max(i, wi); i < s; ++i) { syms[li++] = dat[i]; ++lf[dat[i]]; } pos = wblk(dat, w, lst, syms, lf, df, eb, li, bs, i - bs, pos); if (!lst) { st.r = (pos & 7) | w[(pos / 8) | 0] << 3; // shft(pos) now 1 less if pos & 7 != 0 pos -= 7; st.h = head, st.p = prev, st.i = i, st.w = wi; } } else { for (var i = st.w || 0; i < s + lst; i += 65535) { // end var e = i + 65535; if (e >= s) { // write final block w[(pos / 8) | 0] = lst; e = s; } pos = wfblk(w, pos + 1, dat.subarray(i, e)); } st.i = s; } return slc(o, 0, pre + shft(pos) + post); }; // CRC32 table var crct = /*#__PURE__*/ (function () { var t = new Int32Array(256); for (var i = 0; i < 256; ++i) { var c = i, k = 9; while (--k) c = ((c & 1) && -306674912) ^ (c >>> 1); t[i] = c; } return t; })(); // CRC32 var crc = function () { var c = -1; return { p: function (d) { // closures have awful performance var cr = c; for (var i = 0; i < d.length; ++i) cr = crct[(cr & 255) ^ d[i]] ^ (cr >>> 8); c = cr; }, d: function () { return ~c; } }; }; // Adler32 var adler = function () { var a = 1, b = 0; return { p: function (d) { // closures have awful performance var n = a, m = b; var l = d.length | 0; for (var i = 0; i != l;) { var e = Math.min(i + 2655, l); for (; i < e; ++i) m += n += d[i]; n = (n & 65535) + 15 * (n >> 16), m = (m & 65535) + 15 * (m >> 16); } a = n, b = m; }, d: function () { a %= 65521, b %= 65521; return (a & 255) << 24 | (a & 0xFF00) << 8 | (b & 255) << 8 | (b >> 8); } }; }; ; // deflate with opts var dopt = function (dat, opt, pre, post, st) { if (!st) { st = { l: 1 }; if (opt.dictionary) { var dict = opt.dictionary.subarray(-32768); var newDat = new u8(dict.length + dat.length); newDat.set(dict); newDat.set(dat, dict.length); dat = newDat; st.w = dict.length; } } return dflt(dat, opt.level == null ? 6 : opt.level, opt.mem == null ? (st.l ? Math.ceil(Math.max(8, Math.min(13, Math.log(dat.length))) * 1.5) : 20) : (12 + opt.mem), pre, post, st); }; // Walmart object spread var mrg = function (a, b) { var o = {}; for (var k in a) o[k] = a[k]; for (var k in b) o[k] = b[k]; return o; }; // worker clone // This is possibly the craziest part of the entire codebase, despite how simple it may seem. // The only parameter to this function is a closure that returns an array of variables outside of the function scope. // We're going to try to figure out the variable names used in the closure as strings because that is crucial for workerization. // We will return an object mapping of true variable name to value (basically, the current scope as a JS object). // The reason we can't just use the original variable names is minifiers mangling the toplevel scope. // This took me three weeks to figure out how to do. var wcln = function (fn, fnStr, td) { var dt = fn(); var st = fn.toString(); var ks = st.slice(st.indexOf('[') + 1, st.lastIndexOf(']')).replace(/\s+/g, '').split(','); for (var i = 0; i < dt.length; ++i) { var v = dt[i], k = ks[i]; if (typeof v == 'function') { fnStr += ';' + k + '='; var st_1 = v.toString(); if (v.prototype) { // for global objects if (st_1.indexOf('[native code]') != -1) { var spInd = st_1.indexOf(' ', 8) + 1; fnStr += st_1.slice(spInd, st_1.indexOf('(', spInd)); } else { fnStr += st_1; for (var t in v.prototype) fnStr += ';' + k + '.prototype.' + t + '=' + v.prototype[t].toString(); } } else fnStr += st_1; } else td[k] = v; } return fnStr; }; var ch = []; // clone bufs var cbfs = function (v) { var tl = []; for (var k in v) { if (v[k].buffer) { tl.push((v[k] = new v[k].constructor(v[k])).buffer); } } return tl; }; // use a worker to execute code var wrkr = function (fns, init, id, cb) { if (!ch[id]) { var fnStr = '', td_1 = {}, m = fns.length - 1; for (var i = 0; i < m; ++i) fnStr = wcln(fns[i], fnStr, td_1); ch[id] = { c: wcln(fns[m], fnStr, td_1), e: td_1 }; } var td = mrg({}, ch[id].e); return wk(ch[id].c + ';onmessage=function(e){for(var k in e.data)self[k]=e.data[k];onmessage=' + init.toString() + '}', id, td, cbfs(td), cb); }; // base async inflate fn var bInflt = function () { return [u8, u16, i32, fleb, fdeb, clim, fl, fd, flrm, fdrm, rev, ec, hMap, max, bits, bits16, shft, slc, err, inflt, inflateSync, pbf, gopt]; }; var bDflt = function () { return [u8, u16, i32, fleb, fdeb, clim, revfl, revfd, flm, flt, fdm, fdt, rev, deo, et, hMap, wbits, wbits16, hTree, ln, lc, clen, wfblk, wblk, shft, slc, dflt, dopt, deflateSync, pbf]; }; // gzip extra var gze = function () { return [gzh, gzhl, wbytes, crc, crct]; }; // gunzip extra var guze = function () { return [gzs, gzl]; }; // zlib extra var zle = function () { return [zlh, wbytes, adler]; }; // unzlib extra var zule = function () { return [zls]; }; // post buf var pbf = function (msg) { return postMessage(msg, [msg.buffer]); }; // get opts var gopt = function (o) { return o && { out: o.size && new u8(o.size), dictionary: o.dictionary }; }; // async helper var cbify = function (dat, opts, fns, init, id, cb) { var w = wrkr(fns, init, id, function (err, dat) { w.terminate(); cb(err, dat); }); w.postMessage([dat, opts], opts.consume ? [dat.buffer] : []); return function () { w.terminate(); }; }; // auto stream var astrm = function (strm) { strm.ondata = function (dat, final) { return postMessage([dat, final], [dat.buffer]); }; return function (ev) { if (ev.data.length) { strm.push(ev.data[0], ev.data[1]); postMessage([ev.data[0].length]); } else strm.flush(); }; }; // async stream attach var astrmify = function (fns, strm, opts, init, id, flush, ext) { var t; var w = wrkr(fns, init, id, function (err, dat) { if (err) w.terminate(), strm.ondata.call(strm, err); else if (!Array.isArray(dat)) ext(dat); else if (dat.length == 1) { strm.queuedSize -= dat[0]; if (strm.ondrain) strm.ondrain(dat[0]); } else { if (dat[1]) w.terminate(); strm.ondata.call(strm, err, dat[0], dat[1]); } }); w.postMessage(opts); strm.queuedSize = 0; strm.push = function (d, f) { if (!strm.ondata) err(5); if (t) strm.ondata(err(4, 0, 1), null, !!f); strm.queuedSize += d.length; w.postMessage([d, t = f], [d.buffer]); }; strm.terminate = function () { w.terminate(); }; if (flush) { strm.flush = function () { w.postMessage([]); }; } }; // read 2 bytes var b2 = function (d, b) { return d[b] | (d[b + 1] << 8); }; // read 4 bytes var b4 = function (d, b) { return (d[b] | (d[b + 1] << 8) | (d[b + 2] << 16) | (d[b + 3] << 24)) >>> 0; }; var b8 = function (d, b) { return b4(d, b) + (b4(d, b + 4) * 4294967296); }; // write bytes var wbytes = function (d, b, v) { for (; v; ++b) d[b] = v, v >>>= 8; }; // gzip header var gzh = function (c, o) { var fn = o.filename; c[0] = 31, c[1] = 139, c[2] = 8, c[8] = o.level < 2 ? 4 : o.level == 9 ? 2 : 0, c[9] = 3; // assume Unix if (o.mtime != 0) wbytes(c, 4, Math.floor(new Date(o.mtime || Date.now()) / 1000)); if (fn) { c[3] = 8; for (var i = 0; i <= fn.length; ++i) c[i + 10] = fn.charCodeAt(i); } }; // gzip footer: -8 to -4 = CRC, -4 to -0 is length // gzip start var gzs = function (d) { if (d[0] != 31 || d[1] != 139 || d[2] != 8) err(6, 'invalid gzip data'); var flg = d[3]; var st = 10; if (flg & 4) st += (d[10] | d[11] << 8) + 2; for (var zs = (flg >> 3 & 1) + (flg >> 4 & 1); zs > 0; zs -= !d[st++]) ; return st + (flg & 2); }; // gzip length var gzl = function (d) { var l = d.length; return (d[l - 4] | d[l - 3] << 8 | d[l - 2] << 16 | d[l - 1] << 24) >>> 0; }; // gzip header length var gzhl = function (o) { return 10 + (o.filename ? o.filename.length + 1 : 0); }; // zlib header var zlh = function (c, o) { var lv = o.level, fl = lv == 0 ? 0 : lv < 6 ? 1 : lv == 9 ? 3 : 2; c[0] = 120, c[1] = (fl << 6) | (o.dictionary && 32); c[1] |= 31 - ((c[0] << 8) | c[1]) % 31; if (o.dictionary) { var h = adler(); h.p(o.dictionary); wbytes(c, 2, h.d()); } }; // zlib start var zls = function (d, dict) { if ((d[0] & 15) != 8 || (d[0] >> 4) > 7 || ((d[0] << 8 | d[1]) % 31)) err(6, 'invalid zlib data'); if ((d[1] >> 5 & 1) == +!dict) err(6, 'invalid zlib data: ' + (d[1] & 32 ? 'need' : 'unexpected') + ' dictionary'); return (d[1] >> 3 & 4) + 2; }; function StrmOpt(opts, cb) { if (typeof opts == 'function') cb = opts, opts = {}; this.ondata = cb; return opts; } /** * Streaming DEFLATE compression */ var Deflate = /*#__PURE__*/ (function () { function Deflate(opts, cb) { if (typeof opts == 'function') cb = opts, opts = {}; this.ondata = cb; this.o = opts || {}; this.s = { l: 0, i: 32768, w: 32768, z: 32768 }; // Buffer length must always be 0 mod 32768 for index calculations to be correct when modifying head and prev // 98304 = 32768 (lookback) + 65536 (common chunk size) this.b = new u8(98304); if (this.o.dictionary) { var dict = this.o.dictionary.subarray(-32768); this.b.set(dict, 32768 - dict.length); this.s.i = 32768 - dict.length; } } Deflate.prototype.p = function (c, f) { this.ondata(dopt(c, this.o, 0, 0, this.s), f); }; /** * Pushes a chunk to be deflated * @param chunk The chunk to push * @param final Whether this is the last chunk */ Deflate.prototype.push = function (chunk, final) { if (!this.ondata) err(5); if (this.s.l) err(4); var endLen = chunk.length + this.s.z; if (endLen > this.b.length) { if (endLen > 2 * this.b.length - 32768) { var newBuf = new u8(endLen & -32768); newBuf.set(this.b.subarray(0, this.s.z)); this.b = newBuf; } var split = this.b.length - this.s.z; this.b.set(chunk.subarray(0, split), this.s.z); this.s.z = this.b.length; this.p(this.b, false); this.b.set(this.b.subarray(-32768)); this.b.set(chunk.subarray(split), 32768); this.s.z = chunk.length - split + 32768; this.s.i = 32766, this.s.w = 32768; } else { this.b.set(chunk, this.s.z); this.s.z += chunk.length; } this.s.l = final & 1; if (this.s.z > this.s.w + 8191 || final) { this.p(this.b, final || false); this.s.w = this.s.i, this.s.i -= 2; } }; /** * Flushes buffered uncompressed data. Useful to immediately retrieve the * deflated output for small inputs. */ Deflate.prototype.flush = function () { if (!this.ondata) err(5); if (this.s.l) err(4); this.p(this.b, false); this.s.w = this.s.i, this.s.i -= 2; }; return Deflate; }()); export { Deflate }; /** * Asynchronous streaming DEFLATE compression */ var AsyncDeflate = /*#__PURE__*/ (function () { function AsyncDeflate(opts, cb) { astrmify([ bDflt, function () { return [astrm, Deflate]; } ], this, StrmOpt.call(this, opts, cb), function (ev) { var strm = new Deflate(ev.data); onmessage = astrm(strm); }, 6, 1); } return AsyncDeflate; }()); export { AsyncDeflate }; export function deflate(data, opts, cb) { if (!cb) cb = opts, opts = {}; if (typeof cb != 'function') err(7); return cbify(data, opts, [ bDflt, ], function (ev) { return pbf(deflateSync(ev.data[0], ev.data[1])); }, 0, cb); } /** * Compresses data with DEFLATE without any wrapper * @param data The data to compress * @param opts The compression options * @returns The deflated version of the data */ export function deflateSync(data, opts) { return dopt(data, opts || {}, 0, 0); } /** * Streaming DEFLATE decompression */ var Inflate = /*#__PURE__*/ (function () { function Inflate(opts, cb) { // no StrmOpt here to avoid adding to workerizer if (typeof opts == 'function') cb = opts, opts = {}; this.ondata = cb; var dict = opts && opts.dictionary && opts.dictionary.subarray(-32768); this.s = { i: 0, b: dict ? dict.length : 0 }; this.o = new u8(32768); this.p = new u8(0); if (dict) this.o.set(dict); } Inflate.prototype.e = function (c) { if (!this.ondata) err(5); if (this.d) err(4); if (!this.p.length) this.p = c; else if (c.length) { var n = new u8(this.p.length + c.length); n.set(this.p), n.set(c, this.p.length), this.p = n; } }; Inflate.prototype.c = function (final) { this.s.i = +(this.d = final || false); var bts = this.s.b; var dt = inflt(this.p, this.s, this.o); this.ondata(slc(dt, bts, this.s.b), this.d); this.o = slc(dt, this.s.b - 32768), this.s.b = this.o.length; this.p = slc(this.p, (this.s.p / 8) | 0), this.s.p &= 7; }; /** * Pushes a chunk to be inflated * @param chunk The chunk to push * @param final Whether this is the final chunk */ Inflate.prototype.push = function (chunk, final) { this.e(chunk), this.c(final); }; return Inflate; }()); export { Inflate }; /** * Asynchronous streaming DEFLATE decompression */ var AsyncInflate = /*#__PURE__*/ (function () { function AsyncInflate(opts, cb) { astrmify([ bInflt, function () { return [astrm, Inflate]; } ], this, StrmOpt.call(this, opts, cb), function (ev) { var strm = new Inflate(ev.data); onmessage = astrm(strm); }, 7, 0); } return AsyncInflate; }()); export { AsyncInflate }; export function inflate(data, opts, cb) { if (!cb) cb = opts, opts = {}; if (typeof cb != 'function') err(7); return cbify(data, opts, [ bInflt ], function (ev) { return pbf(inflateSync(ev.data[0], gopt(ev.data[1]))); }, 1, cb); } /** * Expands DEFLATE data with no wrapper * @param data The data to decompress * @param opts The decompression options * @returns The decompressed version of the data */ export function inflateSync(data, opts) { return inflt(data, { i: 2 }, opts && opts.out, opts && opts.dictionary); } // before you yell at me for not just using extends, my reason is that TS inheritance is hard to workerize. /** * Streaming GZIP compression */ var Gzip = /*#__PURE__*/ (function () { function Gzip(opts, cb) { this.c = crc(); this.l = 0; this.v = 1; Deflate.call(this, opts, cb); } /** * Pushes a chunk to be GZIPped * @param chunk The chunk to push * @param final Whether this is the last chunk */ Gzip.prototype.push = function (chunk, final) { this.c.p(chunk); this.l += chunk.length; Deflate.prototype.push.call(this, chunk, final); }; Gzip.prototype.p = function (c, f) { var raw = dopt(c, this.o, this.v && gzhl(this.o), f && 8, this.s); if (this.v) gzh(raw, this.o), this.v = 0; if (f) wbytes(raw, raw.length - 8, this.c.d()), wbytes(raw, raw.length - 4, this.l); this.ondata(raw, f); }; /** * Flushes buffered uncompressed data. Useful to immediately retrieve the * GZIPped output for small inputs. */ Gzip.prototype.flush = function () { Deflate.prototype.flush.call(this); }; return Gzip; }()); export { Gzip }; /** * Asynchronous streaming GZIP compression */ var AsyncGzip = /*#__PURE__*/ (function () { function AsyncGzip(opts, cb) { astrmify([ bDflt, gze, function () { return [astrm, Deflate, Gzip]; } ], this, StrmOpt.call(this, opts, cb), function (ev) { var strm = new Gzip(ev.data); onmessage = astrm(strm); }, 8, 1); } return AsyncGzip; }()); export { AsyncGzip }; export function gzip(data, opts, cb) { if (!cb) cb = opts, opts = {}; if (typeof cb != 'function') err(7); return cbify(data, opts, [ bDflt, gze, function () { return [gzipSync]; } ], function (ev) { return pbf(gzipSync(ev.data[0], ev.data[1])); }, 2, cb); } /** * Compresses data with GZIP * @param data The data to compress * @param opts The compression options * @returns The gzipped version of the data */ export function gzipSync(data, opts) { if (!opts) opts = {}; var c = crc(), l = data.length; c.p(data); var d = dopt(data, opts, gzhl(opts), 8), s = d.length; return gzh(d, opts), wbytes(d, s - 8, c.d()), wbytes(d, s - 4, l), d; } /** * Streaming single or multi-member GZIP decompression */ var Gunzip = /*#__PURE__*/ (function () { function Gunzip(opts, cb) { this.v = 1; this.r = 0; Inflate.call(this, opts, cb); } /** * Pushes a chunk to be GUNZIPped * @param chunk The chunk to push * @param final Whether this is the last chunk */ Gunzip.prototype.push = function (chunk, final) { Inflate.prototype.e.call(this, chunk); this.r += chunk.length; if (this.v) { var p = this.p.subarray(this.v - 1); var s = p.length > 3 ? gzs(p) : 4; if (s > p.length) { if (!final) return; } else if (this.v > 1 && this.onmember) { this.onmember(this.r - p.length); } this.p = p.subarray(s), this.v = 0; } // necessary to prevent TS from using the closure value // This allows for workerization to function correctly Inflate.prototype.c.call(this, final); // process concatenated GZIP if (this.s.f && !this.s.l && !final) { this.v = shft(this.s.p) + 9; this.s = { i: 0 }; this.o = new u8(0); this.push(new u8(0), final); } }; return Gunzip; }()); export { Gunzip }; /** * Asynchronous streaming single or multi-member GZIP decompression */ var AsyncGunzip = /*#__PURE__*/ (function () { function AsyncGunzip(opts, cb) { var _this = this; astrmify([ bInflt, guze, function () { return [astrm, Inflate, Gunzip]; } ], this, StrmOpt.call(this, opts, cb), function (ev) { var strm = new Gunzip(ev.data); strm.onmember = function (offset) { return postMessage(offset); }; onmessage = astrm(strm); }, 9, 0, function (offset) { return _this.onmember && _this.onmember(offset); }); } return AsyncGunzip; }()); export { AsyncGunzip }; export function gunzip(data, opts, cb) { if (!cb) cb = opts, opts = {}; if (typeof cb != 'function') err(7); return cbify(data, opts, [ bInflt, guze, function () { return [gunzipSync]; } ], function (ev) { return pbf(gunzipSync(ev.data[0], ev.data[1])); }, 3, cb); } /** * Expands GZIP data * @param data The data to decompress * @param opts The decompression options * @returns The decompressed version of the data */ export function gunzipSync(data, opts) { var st = gzs(data); if (st + 8 > data.length) err(6, 'invalid gzip data'); return inflt(data.subarray(st, -8), { i: 2 }, opts && opts.out || new u8(gzl(data)), opts && opts.dictionary); } /** * Streaming Zlib compression */ var Zlib = /*#__PURE__*/ (function () { function Zlib(opts, cb) { this.c = adler(); this.v = 1; Deflate.call(this, opts, cb); } /** * Pushes a chunk to be zlibbed * @param chunk The chunk to push * @param final Whether this is the last chunk */ Zlib.prototype.push = function (chunk, final) { this.c.p(chunk); Deflate.prototype.push.call(this, chunk, final); }; Zlib.prototype.p = function (c, f) { var raw = dopt(c, this.o, this.v && (this.o.dictionary ? 6 : 2), f && 4, this.s); if (this.v) zlh(raw, this.o), this.v = 0; if (f) wbytes(raw, raw.length - 4, this.c.d()); this.ondata(raw, f); }; /** * Flushes buffered uncompressed data. Useful to immediately retrieve the * zlibbed output for small inputs. */ Zlib.prototype.flush = function () { Deflate.prototype.flush.call(this); }; return Zlib; }()); export { Zlib }; /** * Asynchronous streaming Zlib compression */ var AsyncZlib = /*#__PURE__*/ (function () { function AsyncZlib(opts, cb) { astrmify([ bDflt, zle, function () { return [astrm, Deflate, Zlib]; } ], this, StrmOpt.call(this, opts, cb), function (ev) { var strm = new Zlib(ev.data); onmessage = astrm(strm); }, 10, 1); } return AsyncZlib; }()); export { AsyncZlib }; export function zlib(data, opts, cb) { if (!cb) cb = opts, opts = {}; if (typeof cb != 'function') err(7); return cbify(data, opts, [ bDflt, zle, function () { return [zlibSync]; } ], function (ev) { return pbf(zlibSync(ev.data[0], ev.data[1])); }, 4, cb); } /** * Compress data with Zlib * @param data The data to compress * @param opts The compression options * @returns The zlib-compressed version of the data */ export function zlibSync(data, opts) { if (!opts) opts = {}; var a = adler(); a.p(data); var d = dopt(data, opts, opts.dictionary ? 6 : 2, 4); return zlh(d, opts), wbytes(d, d.length - 4, a.d()), d; } /** * Streaming Zlib decompression */ var Unzlib = /*#__PURE__*/ (function () { function Unzlib(opts, cb) { Inflate.call(this, opts, cb); this.v = opts && opts.dictionary ? 2 : 1; } /** * Pushes a chunk to be unzlibbed * @param chunk The chunk to push * @param final Whether this is the last chunk */ Unzlib.prototype.push = function (chunk, final) { Inflate.prototype.e.call(this, chunk); if (this.v) { if (this.p.length < 6 && !final) return; this.p = this.p.subarray(zls(this.p, this.v - 1)), this.v = 0; } if (final) { if (this.p.length < 4) err(6, 'invalid zlib data'); this.p = this.p.subarray(0, -4); } // necessary to prevent TS from using the closure value // This allows for workerization to function correctly Inflate.prototype.c.call(this, final); }; return Unzlib; }()); export { Unzlib }; /** * Asynchronous streaming Zlib decompression */ var AsyncUnzlib = /*#__PURE__*/ (function () { function AsyncUnzlib(opts, cb) { astrmify([ bInflt, zule, function () { return [astrm, Inflate, Unzlib]; } ], this, StrmOpt.call(this, opts, cb), function (ev) { var strm = new Unzlib(ev.data); onmessage = astrm(strm); }, 11, 0); } return AsyncUnzlib; }()); export { AsyncUnzlib }; export function unzlib(data, opts, cb) { if (!cb) cb = opts, opts = {}; if (typeof cb != 'function') err(7); return cbify(data, opts, [ bInflt, zule, function () { return [unzlibSync]; } ], function (ev) { return pbf(unzlibSync(ev.data[0], gopt(ev.data[1]))); }, 5, cb); } /** * Expands Zlib data * @param data The data to decompress * @param opts The decompression options * @returns The decompressed version of the data */ export function unzlibSync(data, opts) { return inflt(data.subarray(zls(data, opts && opts.dictionary), -4), { i: 2 }, opts && opts.out, opts && opts.dictionary); } // Default algorithm for compression (used because having a known output size allows faster decompression) export { gzip as compress, AsyncGzip as AsyncCompress }; export { gzipSync as compressSync, Gzip as Compress }; /** * Streaming GZIP, Zlib, or raw DEFLATE decompression */ var Decompress = /*#__PURE__*/ (function () { function Decompress(opts, cb) { this.o = StrmOpt.call(this, opts, cb) || {}; this.G = Gunzip; this.I = Inflate; this.Z = Unzlib; } // init substream // overriden by AsyncDecompress Decompress.prototype.i = function () { var _this = this; this.s.ondata = function (dat, final) { _this.ondata(dat, final); }; }; /** * Pushes a chunk to be decompressed * @param chunk The chunk to push * @param final Whether this is the last chunk */ Decompress.prototype.push = function (chunk, final) { if (!this.ondata) err(5); if (!this.s) { if (this.p && this.p.length) { var n = new u8(this.p.length + chunk.length); n.set(this.p), n.set(chunk, this.p.length); } else this.p = chunk; if (this.p.length > 2) { this.s = (this.p[0] == 31 && this.p[1] == 139 && this.p[2] == 8) ? new this.G(this.o) : ((this.p[0] & 15) != 8 || (this.p[0] >> 4) > 7 || ((this.p[0] << 8 | this.p[1]) % 31)) ? new this.I(this.o) : new this.Z(this.o); this.i(); this.s.push(this.p, final); this.p = null; } } else this.s.push(chunk, final); }; return Decompress; }()); export { Decompress }; /** * Asynchronous streaming GZIP, Zlib, or raw DEFLATE decompression */ var AsyncDecompress = /*#__PURE__*/ (function () { function AsyncDecompress(opts, cb) { Decompress.call(this, opts, cb); this.queuedSize = 0; this.G = AsyncGunzip; this.I = AsyncInflate; this.Z = AsyncUnzlib; } AsyncDecompress.prototype.i = function () { var _this = this; this.s.ondata = function (err, dat, final) { _this.ondata(err, dat, final); }; this.s.ondrain = function (size) { _this.queuedSize -= size; if (_this.ondrain) _this.ondrain(size); }; }; /** * Pushes a chunk to be decompressed * @param chunk The chunk to push * @param final Whether this is the last chunk */ AsyncDecompress.prototype.push = function (chunk, final) { this.queuedSize += chunk.length; Decompress.prototype.push.call(this, chunk, final); }; return AsyncDecompress; }()); export { AsyncDecompress }; export function decompress(data, opts, cb) { if (!cb) cb = opts, opts = {}; if (typeof cb != 'function') err(7); return (data[0] == 31 && data[1] == 139 && data[2] == 8) ? gunzip(data, opts, cb) : ((data[0] & 15) != 8 || (data[0] >> 4) > 7 || ((data[0] << 8 | data[1]) % 31)) ? inflate(data, opts, cb) : unzlib(data, opts, cb); } /** * Expands compressed GZIP, Zlib, or raw DEFLATE data, automatically detecting the format * @param data The data to decompress * @param opts The decompression options * @returns The decompressed version of the data */ export function decompressSync(data, opts) { return (data[0] == 31 && data[1] == 139 && data[2] == 8) ? gunzipSync(data, opts) : ((data[0] & 15) != 8 || (data[0] >> 4) > 7 || ((data[0] << 8 | data[1]) % 31)) ? inflateSync(data, opts) : unzlibSync(data, opts); } // flatten a directory structure var fltn = function (d, p, t, o) { for (var k in d) { var val = d[k], n = p + k, op = o; if (Array.isArray(val)) op = mrg(o, val[1]), val = val[0]; if (val instanceof u8) t[n] = [val, op]; else { t[n += '/'] = [new u8(0), op]; fltn(val, n, t, o); } } }; // text encoder var te = typeof TextEncoder != 'undefined' && /*#__PURE__*/ new TextEncoder(); // text decoder var td = typeof TextDecoder != 'undefined' && /*#__PURE__*/ new TextDecoder(); // text decoder stream var tds = 0; try { td.decode(et, { stream: true }); tds = 1; } catch (e) { } // decode UTF8 var dutf8 = function (d) { for (var r = '', i = 0;;) { var c = d[i++]; var eb = (c > 127) + (c > 223) + (c > 239); if (i + eb > d.length) return { s: r, r: slc(d, i - 1) }; if (!eb) r += String.fromCharCode(c); else if (eb == 3) { c = ((c & 15) << 18 | (d[i++] & 63) << 12 | (d[i++] & 63) << 6 | (d[i++] & 63)) - 65536, r += String.fromCharCode(55296 | (c >> 10), 56320 | (c & 1023)); } else if (eb & 1) r += String.fromCharCode((c & 31) << 6 | (d[i++] & 63)); else r += String.fromCharCode((c & 15) << 12 | (d[i++] & 63) << 6 | (d[i++] & 63)); } }; /** * Streaming UTF-8 decoding */ var DecodeUTF8 = /*#__PURE__*/ (function () { /** * Creates a UTF-8 decoding stream * @param cb The callback to call whenever data is decoded */ function DecodeUTF8(cb) { this.ondata = cb; if (tds) this.t = new TextDecoder(); else this.p = et; } /** * Pushes a chunk to be decoded from UTF-8 binary * @param chunk The chunk to push * @param final Whether this is the last chunk */ DecodeUTF8.prototype.push = function (chunk, final) { if (!this.ondata) err(5); final = !!final; if (this.t) { this.ondata(this.t.decode(chunk, { stream: true }), final); if (final) { if (this.t.decode().length) err(8); this.t = null; } return; } if (!this.p) err(4); var dat = new u8(this.p.length + chunk.length); dat.set(this.p); dat.set(chunk, this.p.length); var _a = dutf8(dat), s = _a.s, r = _a.r; if (final) { if (r.length) err(8); this.p = null; } else this.p = r; this.ondata(s, final); }; return DecodeUTF8; }()); export { DecodeUTF8 }; /** * Streaming UTF-8 encoding */ var EncodeUTF8 = /*#__PURE__*/ (function () { /** * Creates a UTF-8 decoding stream * @param cb The callback to call whenever data is encoded */ function EncodeUTF8(cb) { this.ondata = cb; } /** * Pushes a chunk to be encoded to UTF-8 * @param chunk The string data to push * @param final Whether this is the last chunk */ EncodeUTF8.prototype.push = function (chunk, final) { if (!this.ondata) err(5); if (this.d) err(4); this.ondata(strToU8(chunk), this.d = final || false); }; return EncodeUTF8; }()); export { EncodeUTF8 }; /** * Converts a string into a Uint8Array for use with compression/decompression methods * @param str The string to encode * @param latin1 Whether or not to interpret the data as Latin-1. This should * not need to be true unless decoding a binary string. * @returns The string encoded in UTF-8/Latin-1 binary */ export function strToU8(str, latin1) { if (latin1) { var ar_1 = new u8(str.length); for (var i = 0; i < str.length; ++i) ar_1[i] = str.charCodeAt(i); return ar_1; } if (te) return te.encode(str); var l = str.length; var ar = new u8(str.length + (str.length >> 1)); var ai = 0; var w = function (v) { ar[ai++] = v; }; for (var i = 0; i < l; ++i) { if (ai + 5 > ar.length) { var n = new u8(ai + 8 + ((l - i) << 1)); n.set(ar); ar = n; } var c = str.charCodeAt(i); if (c < 128 || latin1) w(c); else if (c < 2048) w(192 | (c >> 6)), w(128 | (c & 63)); else if (c > 55295 && c < 57344) c = 65536 + (c & 1023 << 10) | (str.charCodeAt(++i) & 1023), w(240 | (c >> 18)), w(128 | ((c >> 12) & 63)), w(128 | ((c >> 6) & 63)), w(128 | (c & 63)); else w(224 | (c >> 12)), w(128 | ((c >> 6) & 63)), w(128 | (c & 63)); } return slc(ar, 0, ai); } /** * Converts a Uint8Array to a string * @param dat The data to decode to string * @param latin1 Whether or not to interpret the data as Latin-1. This should * not need to be true unless encoding to binary string. * @returns The original UTF-8/Latin-1 string */ export function strFromU8(dat, latin1) { if (latin1) { var r = ''; for (var i = 0; i < dat.length; i += 16384) r += String.fromCharCode.apply(null, dat.subarray(i, i + 16384)); return r; } else if (td) { return td.decode(dat); } else { var _a = dutf8(dat), s = _a.s, r = _a.r; if (r.length) err(8); return s; } } ; // deflate bit flag var dbf = function (l) { return l == 1 ? 3 : l < 6 ? 2 : l == 9 ? 1 : 0; }; // skip local zip header var slzh = function (d, b) { return b + 30 + b2(d, b + 26) + b2(d, b + 28); }; // read zip header var zh = function (d, b, z) { var fnl = b2(d, b + 28), fn = strFromU8(d.subarray(b + 46, b + 46 + fnl), !(b2(d, b + 8) & 2048)), es = b + 46 + fnl, bs = b4(d, b + 20); var _a = z && bs == 4294967295 ? z64e(d, es) : [bs, b4(d, b + 24), b4(d, b + 42)], sc = _a[0], su = _a[1], off = _a[2]; return [b2(d, b + 10), sc, su, fn, es + b2(d, b + 30) + b2(d, b + 32), off]; }; // read zip64 extra field var z64e = function (d, b) { for (; b2(d, b) != 1; b += 4 + b2(d, b + 2)) ; return [b8(d, b + 12), b8(d, b + 4), b8(d, b + 20)]; }; // extra field length var exfl = function (ex) { var le = 0; if (ex) { for (var k in ex) { var l = ex[k].length; if (l > 65535) err(9); le += l + 4; } } return le; }; // write zip header var wzh = function (d, b, f, fn, u, c, ce, co) { var fl = fn.length, ex = f.extra, col = co && co.length; var exl = exfl(ex); wbytes(d, b, ce != null ? 0x2014B50 : 0x4034B50), b += 4; if (ce != null) d[b++] = 20, d[b++] = f.os; d[b] = 20, b += 2; // spec compliance? what's that? d[b++] = (f.flag << 1) | (c < 0 && 8), d[b++] = u && 8; d[b++] = f.compression & 255, d[b++] = f.compression >> 8; var dt = new Date(f.mtime == null ? Date.now() : f.mtime), y = dt.getFullYear() - 1980; if (y < 0 || y > 119) err(10); wbytes(d, b, (y << 25) | ((dt.getMonth() + 1) << 21) | (dt.getDate() << 16) | (dt.getHours() << 11) | (dt.getMinutes() << 5) | (dt.getSeconds() >> 1)), b += 4; if (c != -1) { wbytes(d, b, f.crc); wbytes(d, b + 4, c < 0 ? -c - 2 : c); wbytes(d, b + 8, f.size); } wbytes(d, b + 12, fl); wbytes(d, b + 14, exl), b += 16; if (ce != null) { wbytes(d, b, col); wbytes(d, b + 6, f.attrs); wbytes(d, b + 10, ce), b += 14; } d.set(fn, b); b += fl; if (exl) { for (var k in ex) { var exf = ex[k], l = exf.length; wbytes(d, b, +k); wbytes(d, b + 2, l); d.set(exf, b + 4), b += 4 + l; } } if (col) d.set(co, b), b += col; return b; }; // write zip footer (end of central directory) var wzf = function (o, b, c, d, e) { wbytes(o, b, 0x6054B50); // skip disk wbytes(o, b + 8, c); wbytes(o, b + 10, c); wbytes(o, b + 12, d); wbytes(o, b + 16, e); }; /** * A pass-through stream to keep data uncompressed in a ZIP archive. */ var ZipPassThrough = /*#__PURE__*/ (function () { /** * Creates a pass-through stream that can be added to ZIP archives * @param filename The filename to associate with this data stream */ function ZipPassThrough(filename) { this.filename = filename; this.c = crc(); this.size = 0; this.compression = 0; } /** * Processes a chunk and pushes to the output stream. You can override this * method in a subclass for custom behavior, but by default this passes * the data through. You must call this.ondata(err, chunk, final) at some * point in this method. * @param chunk The chunk to process * @param final Whether this is the last chunk */ ZipPassThrough.prototype.process = function (chunk, final) { this.ondata(null, chunk, final); }; /** * Pushes a chunk to be added. If you are subclassing this with a custom * compression algorithm, note that you must push data from the source * file only, pre-compression. * @param chunk The chunk to push * @param final Whether this is the last chunk */ ZipPassThrough.prototype.push = function (chunk, final) { if (!this.ondata) err(5); this.c.p(chunk); this.size += chunk.length; if (final) this.crc = this.c.d(); this.process(chunk, final || false); }; return ZipPassThrough; }()); export { ZipPassThrough }; // I don't extend because TypeScript extension adds 1kB of runtime bloat /** * Streaming DEFLATE compression for ZIP archives. Prefer using AsyncZipDeflate * for better performance */ var ZipDeflate = /*#__PURE__*/ (function () { /** * Creates a DEFLATE stream that can be added to ZIP archives * @param filename The filename to associate with this data stream * @param opts The compression options */ function ZipDeflate(filename, opts) { var _this = this; if (!opts) opts = {}; ZipPassThrough.call(this, filename); this.d = new Deflate(opts, function (dat, final) { _this.ondata(null, dat, final); }); this.compression = 8; this.flag = dbf(opts.level); } ZipDeflate.prototype.process = function (chunk, final) { try { this.d.push(chunk, final); } catch (e) { this.ondata(e, null, final); } }; /** * Pushes a chunk to be deflated * @param chunk The chunk to push * @param final Whether this is the last chunk */ ZipDeflate.prototype.push = function (chunk, final) { ZipPassThrough.prototype.push.call(this, chunk, final); }; return ZipDeflate; }()); export { ZipDeflate }; /** * Asynchronous streaming DEFLATE compression for ZIP archives */ var AsyncZipDeflate = /*#__PURE__*/ (function () { /** * Creates an asynchronous DEFLATE stream that can be added to ZIP archives * @param filename The filename to associate with this data stream * @param opts The compression options */ function AsyncZipDeflate(filename, opts) { var _this = this; if (!opts) opts = {}; ZipPassThrough.call(this, filename); this.d = new AsyncDeflate(opts, function (err, dat, final) { _this.ondata(err, dat, final); }); this.compression = 8; this.flag = dbf(opts.level); this.terminate = this.d.terminate; } AsyncZipDeflate.prototype.process = function (chunk, final) { this.d.push(chunk, final); }; /** * Pushes a chunk to be deflated * @param chunk The chunk to push * @param final Whether this is the last chunk */ AsyncZipDeflate.prototype.push = function (chunk, final) { ZipPassThrough.prototype.push.call(this, chunk, final); }; return AsyncZipDeflate; }()); export { AsyncZipDeflate }; // TODO: Better tree shaking /** * A zippable archive to which files can incrementally be added */ var Zip = /*#__PURE__*/ (function () { /** * Creates an empty ZIP archive to which files can be added * @param cb The callback to call whenever data for the generated ZIP archive * is available */ function Zip(cb) { this.ondata = cb; this.u = []; this.d = 1; } /** * Adds a file to the ZIP archive * @param file The file stream to add */ Zip.prototype.add = function (file) { var _this = this; if (!this.ondata) err(5); // finishing or finished if (this.d & 2) this.ondata(err(4 + (this.d & 1) * 8, 0, 1), null, false); else { var f = strToU8(file.filename), fl_1 = f.length; var com = file.comment, o = com && strToU8(com); var u = fl_1 != file.filename.length || (o && (com.length != o.length)); var hl_1 = fl_1 + exfl(file.extra) + 30; if (fl_1 > 65535) this.ondata(err(11, 0, 1), null, false); var header = new u8(hl_1); wzh(header, 0, file, f, u, -1); var chks_1 = [header]; var pAll_1 = function () { for (var _i = 0, chks_2 = chks_1; _i < chks_2.length; _i++) { var chk = chks_2[_i]; _this.ondata(null, chk, false); } chks_1 = []; }; var tr_1 = this.d; this.d = 0; var ind_1 = this.u.length; var uf_1 = mrg(file, { f: f, u: u, o: o, t: function () { if (file.terminate) file.terminate(); }, r: function () { pAll_1(); if (tr_1) { var nxt = _this.u[ind_1 + 1]; if (nxt) nxt.r(); else _this.d = 1; } tr_1 = 1; } }); var cl_1 = 0; file.ondata = function (err, dat, final) { if (err) { _this.ondata(err, dat, final); _this.terminate(); } else { cl_1 += dat.length; chks_1.push(dat); if (final) { var dd = new u8(16); wbytes(dd, 0, 0x8074B50); wbytes(dd, 4, file.crc); wbytes(dd, 8, cl_1); wbytes(dd, 12, file.size); chks_1.push(dd); uf_1.c = cl_1, uf_1.b = hl_1 + cl_1 + 16, uf_1.crc = file.crc, uf_1.size = file.size; if (tr_1) uf_1.r(); tr_1 = 1; } else if (tr_1) pAll_1(); } }; this.u.push(uf_1); } }; /** * Ends the process of adding files and prepares to emit the final chunks. * This *must* be called after adding all desired files for the resulting * ZIP file to work properly. */ Zip.prototype.end = function () { var _this = this; if (this.d & 2) { this.ondata(err(4 + (this.d & 1) * 8, 0, 1), null, true); return; } if (this.d) this.e(); else this.u.push({ r: function () { if (!(_this.d & 1)) return; _this.u.splice(-1, 1); _this.e(); }, t: function () { } }); this.d = 3; }; Zip.prototype.e = function () { var bt = 0, l = 0, tl = 0; for (var _i = 0, _a = this.u; _i < _a.length; _i++) { var f = _a[_i]; tl += 46 + f.f.length + exfl(f.extra) + (f.o ? f.o.length : 0); } var out = new u8(tl + 22); for (var _b = 0, _c = this.u; _b < _c.length; _b++) { var f = _c[_b]; wzh(out, bt, f, f.f, f.u, -f.c - 2, l, f.o); bt += 46 + f.f.length + exfl(f.extra) + (f.o ? f.o.length : 0), l += f.b; } wzf(out, bt, this.u.length, tl, l); this.ondata(null, out, true); this.d = 2; }; /** * A method to terminate any internal workers used by the stream. Subsequent * calls to add() will fail. */ Zip.prototype.terminate = function () { for (var _i = 0, _a = this.u; _i < _a.length; _i++) { var f = _a[_i]; f.t(); } this.d = 2; }; return Zip; }()); export { Zip }; export function zip(data, opts, cb) { if (!cb) cb = opts, opts = {}; if (typeof cb != 'function') err(7); var r = {}; fltn(data, '', r, opts); var k = Object.keys(r); var lft = k.length, o = 0, tot = 0; var slft = lft, files = new Array(lft); var term = []; var tAll = function () { for (var i = 0; i < term.length; ++i) term[i](); }; var cbd = function (a, b) { mt(function () { cb(a, b); }); }; mt(function () { cbd = cb; }); var cbf = function () { var out = new u8(tot + 22), oe = o, cdl = tot - o; tot = 0; for (var i = 0; i < slft; ++i) { var f = files[i]; try { var l = f.c.length; wzh(out, tot, f, f.f, f.u, l); var badd = 30 + f.f.length + exfl(f.extra); var loc = tot + badd; out.set(f.c, loc); wzh(out, o, f, f.f, f.u, l, tot, f.m), o += 16 + badd + (f.m ? f.m.length : 0), tot = loc + l; } catch (e) { return cbd(e, null); } } wzf(out, o, files.length, cdl, oe); cbd(null, out); }; if (!lft) cbf(); var _loop_1 = function (i) { var fn = k[i]; var _a = r[fn], file = _a[0], p = _a[1]; var c = crc(), size = file.length; c.p(file); var f = strToU8(fn), s = f.length; var com = p.comment, m = com && strToU8(com), ms = m && m.length; var exl = exfl(p.extra); var compression = p.level == 0 ? 0 : 8; var cbl = function (e, d) { if (e) { tAll(); cbd(e, null); } else { var l = d.length; files[i] = mrg(p, { size: size, crc: c.d(), c: d, f: f, m: m, u: s != fn.length || (m && (com.length != ms)), compression: compression }); o += 30 + s + exl + l; tot += 76 + 2 * (s + exl) + (ms || 0) + l; if (!--lft) cbf(); } }; if (s > 65535) cbl(err(11, 0, 1), null); if (!compression) cbl(null, file); else if (size < 160000) { try { cbl(null, deflateSync(file, p)); } catch (e) { cbl(e, null); } } else term.push(deflate(file, p, cbl)); }; // Cannot use lft because it can decrease for (var i = 0; i < slft; ++i) { _loop_1(i); } return tAll; } /** * Synchronously creates a ZIP file. Prefer using `zip` for better performance * with more than one file. * @param data The directory structure for the ZIP archive * @param opts The main options, merged with per-file options * @returns The generated ZIP archive */ export function zipSync(data, opts) { if (!opts) opts = {}; var r = {}; var files = []; fltn(data, '', r, opts); var o = 0; var tot = 0; for (var fn in r) { var _a = r[fn], file = _a[0], p = _a[1]; var compression = p.level == 0 ? 0 : 8; var f = strToU8(fn), s = f.length; var com = p.comment, m = com && strToU8(com), ms = m && m.length; var exl = exfl(p.extra); if (s > 65535) err(11); var d = compression ? deflateSync(file, p) : file, l = d.length; var c = crc(); c.p(file); files.push(mrg(p, { size: file.length, crc: c.d(), c: d, f: f, m: m, u: s != fn.length || (m && (com.length != ms)), o: o, compression: compression })); o += 30 + s + exl + l; tot += 76 + 2 * (s + exl) + (ms || 0) + l; } var out = new u8(tot + 22), oe = o, cdl = tot - o; for (var i = 0; i < files.length; ++i) { var f = files[i]; wzh(out, f.o, f, f.f, f.u, f.c.length); var badd = 30 + f.f.length + exfl(f.extra); out.set(f.c, f.o + badd); wzh(out, o, f, f.f, f.u, f.c.length, f.o, f.m), o += 16 + badd + (f.m ? f.m.length : 0); } wzf(out, o, files.length, cdl, oe); return out; } /** * Streaming pass-through decompression for ZIP archives */ var UnzipPassThrough = /*#__PURE__*/ (function () { function UnzipPassThrough() { } UnzipPassThrough.prototype.push = function (data, final) { this.ondata(null, data, final); }; UnzipPassThrough.compression = 0; return UnzipPassThrough; }()); export { UnzipPassThrough }; /** * Streaming DEFLATE decompression for ZIP archives. Prefer AsyncZipInflate for * better performance. */ var UnzipInflate = /*#__PURE__*/ (function () { /** * Creates a DEFLATE decompression that can be used in ZIP archives */ function UnzipInflate() { var _this = this; this.i = new Inflate(function (dat, final) { _this.ondata(null, dat, final); }); } UnzipInflate.prototype.push = function (data, final) { try { this.i.push(data, final); } catch (e) { this.ondata(e, null, final); } }; UnzipInflate.compression = 8; return UnzipInflate; }()); export { UnzipInflate }; /** * Asynchronous streaming DEFLATE decompression for ZIP archives */ var AsyncUnzipInflate = /*#__PURE__*/ (function () { /** * Creates a DEFLATE decompression that can be used in ZIP archives */ function AsyncUnzipInflate(_, sz) { var _this = this; if (sz < 320000) { this.i = new Inflate(function (dat, final) { _this.ondata(null, dat, final); }); } else { this.i = new AsyncInflate(function (err, dat, final) { _this.ondata(err, dat, final); }); this.terminate = this.i.terminate; } } AsyncUnzipInflate.prototype.push = function (data, final) { if (this.i.terminate) data = slc(data, 0); this.i.push(data, final); }; AsyncUnzipInflate.compression = 8; return AsyncUnzipInflate; }()); export { AsyncUnzipInflate }; /** * A ZIP archive decompression stream that emits files as they are discovered */ var Unzip = /*#__PURE__*/ (function () { /** * Creates a ZIP decompression stream * @param cb The callback to call whenever a file in the ZIP archive is found */ function Unzip(cb) { this.onfile = cb; this.k = []; this.o = { 0: UnzipPassThrough }; this.p = et; } /** * Pushes a chunk to be unzipped * @param chunk The chunk to push * @param final Whether this is the last chunk */ Unzip.prototype.push = function (chunk, final) { var _this = this; if (!this.onfile) err(5); if (!this.p) err(4); if (this.c > 0) { var len = Math.min(this.c, chunk.length); var toAdd = chunk.subarray(0, len); this.c -= len; if (this.d) this.d.push(toAdd, !this.c); else this.k[0].push(toAdd); chunk = chunk.subarray(len); if (chunk.length) return this.push(chunk, final); } else { var f = 0, i = 0, is = void 0, buf = void 0; if (!this.p.length) buf = chunk; else if (!chunk.length) buf = this.p; else { buf = new u8(this.p.length + chunk.length); buf.set(this.p), buf.set(chunk, this.p.length); } var l = buf.length, oc = this.c, add = oc && this.d; var _loop_2 = function () { var _a; var sig = b4(buf, i); if (sig == 0x4034B50) { f = 1, is = i; this_1.d = null; this_1.c = 0; var bf = b2(buf, i + 6), cmp_1 = b2(buf, i + 8), u = bf & 2048, dd = bf & 8, fnl = b2(buf, i + 26), es = b2(buf, i + 28); if (l > i + 30 + fnl + es) { var chks_3 = []; this_1.k.unshift(chks_3); f = 2; var sc_1 = b4(buf, i + 18), su_1 = b4(buf, i + 22); var fn_1 = strFromU8(buf.subarray(i + 30, i += 30 + fnl), !u); if (sc_1 == 4294967295) { _a = dd ? [-2] : z64e(buf, i), sc_1 = _a[0], su_1 = _a[1]; } else if (dd) sc_1 = -1; i += es; this_1.c = sc_1; var d_1; var file_1 = { name: fn_1, compression: cmp_1, start: function () { if (!file_1.ondata) err(5); if (!sc_1) file_1.ondata(null, et, true); else { var ctr = _this.o[cmp_1]; if (!ctr) file_1.ondata(err(14, 'unknown compression type ' + cmp_1, 1), null, false); d_1 = sc_1 < 0 ? new ctr(fn_1) : new ctr(fn_1, sc_1, su_1); d_1.ondata = function (err, dat, final) { file_1.ondata(err, dat, final); }; for (var _i = 0, chks_4 = chks_3; _i < chks_4.length; _i++) { var dat = chks_4[_i]; d_1.push(dat, false); } if (_this.k[0] == chks_3 && _this.c) _this.d = d_1; else d_1.push(et, true); } }, terminate: function () { if (d_1 && d_1.terminate) d_1.terminate(); } }; if (sc_1 >= 0) file_1.size = sc_1, file_1.originalSize = su_1; this_1.onfile(file_1); } return "break"; } else if (oc) { if (sig == 0x8074B50) { is = i += 12 + (oc == -2 && 8), f = 3, this_1.c = 0; return "break"; } else if (sig == 0x2014B50) { is = i -= 4, f = 3, this_1.c = 0; return "break"; } } }; var this_1 = this; for (; i < l - 4; ++i) { var state_1 = _loop_2(); if (state_1 === "break") break; } this.p = et; if (oc < 0) { var dat = f ? buf.subarray(0, is - 12 - (oc == -2 && 8) - (b4(buf, is - 16) == 0x8074B50 && 4)) : buf.subarray(0, i); if (add) add.push(dat, !!f); else this.k[+(f == 2)].push(dat); } if (f & 2) return this.push(buf.subarray(i), final); this.p = buf.subarray(i); } if (final) { if (this.c) err(13); this.p = null; } }; /** * Registers a decoder with the stream, allowing for files compressed with * the compression type provided to be expanded correctly * @param decoder The decoder constructor */ Unzip.prototype.register = function (decoder) { this.o[decoder.compression] = decoder; }; return Unzip; }()); export { Unzip }; var mt = typeof queueMicrotask == 'function' ? queueMicrotask : typeof setTimeout == 'function' ? setTimeout : function (fn) { fn(); }; export function unzip(data, opts, cb) { if (!cb) cb = opts, opts = {}; if (typeof cb != 'function') err(7); var term = []; var tAll = function () { for (var i = 0; i < term.length; ++i) term[i](); }; var files = {}; var cbd = function (a, b) { mt(function () { cb(a, b); }); }; mt(function () { cbd = cb; }); var e = data.length - 22; for (; b4(data, e) != 0x6054B50; --e) { if (!e || data.length - e > 65558) { cbd(err(13, 0, 1), null); return tAll; } } ; var lft = b2(data, e + 8); if (lft) { var c = lft; var o = b4(data, e + 16); var z = o == 4294967295 || c == 65535; if (z) { var ze = b4(data, e - 12); z = b4(data, ze) == 0x6064B50; if (z) { c = lft = b4(data, ze + 32); o = b4(data, ze + 48); } } var fltr = opts && opts.filter; var _loop_3 = function (i) { var _a = zh(data, o, z), c_1 = _a[0], sc = _a[1], su = _a[2], fn = _a[3], no = _a[4], off = _a[5], b = slzh(data, off); o = no; var cbl = function (e, d) { if (e) { tAll(); cbd(e, null); } else { if (d) files[fn] = d; if (!--lft) cbd(null, files); } }; if (!fltr || fltr({ name: fn, size: sc, originalSize: su, compression: c_1 })) { if (!c_1) cbl(null, slc(data, b, b + sc)); else if (c_1 == 8) { var infl = data.subarray(b, b + sc); // Synchronously decompress under 512KB, or barely-compressed data if (su < 524288 || sc > 0.8 * su) { try { cbl(null, inflateSync(infl, { out: new u8(su) })); } catch (e) { cbl(e, null); } } else term.push(inflate(infl, { size: su }, cbl)); } else cbl(err(14, 'unknown compression type ' + c_1, 1), null); } else cbl(null, null); }; for (var i = 0; i < c; ++i) { _loop_3(i); } } else cbd(null, {}); return tAll; } /** * Synchronously decompresses a ZIP archive. Prefer using `unzip` for better * performance with more than one file. * @param data The raw compressed ZIP file * @param opts The ZIP extraction options * @returns The decompressed files */ export function unzipSync(data, opts) { var files = {}; var e = data.length - 22; for (; b4(data, e) != 0x6054B50; --e) { if (!e || data.length - e > 65558) err(13); } ; var c = b2(data, e + 8); if (!c) return {}; var o = b4(data, e + 16); var z = o == 4294967295 || c == 65535; if (z) { var ze = b4(data, e - 12); z = b4(data, ze) == 0x6064B50; if (z) { c = b4(data, ze + 32); o = b4(data, ze + 48); } } var fltr = opts && opts.filter; for (var i = 0; i < c; ++i) { var _a = zh(data, o, z), c_2 = _a[0], sc = _a[1], su = _a[2], fn = _a[3], no = _a[4], off = _a[5], b = slzh(data, off); o = no; if (!fltr || fltr({ name: fn, size: sc, originalSize: su, compression: c_2 })) { if (!c_2) files[fn] = slc(data, b, b + sc); else if (c_2 == 8) files[fn] = inflateSync(data.subarray(b, b + sc), { out: new u8(su) }); else err(14, 'unknown compression type ' + c_2); } } return files; }