This repository has been archived on 2024-09-11. You can view files and clone it, but cannot push or open issues or pull requests.
svrjs-blog-newsletter/cronjob/node_modules/nodemailer/lib/dkim/message-parser.js
2024-05-26 22:54:55 +02:00

155 lines
4.7 KiB
JavaScript

'use strict';
const Transform = require('stream').Transform;
/**
* MessageParser instance is a transform stream that separates message headers
* from the rest of the body. Headers are emitted with the 'headers' event. Message
* body is passed on as the resulting stream.
*/
class MessageParser extends Transform {
constructor(options) {
super(options);
this.lastBytes = Buffer.alloc(4);
this.headersParsed = false;
this.headerBytes = 0;
this.headerChunks = [];
this.rawHeaders = false;
this.bodySize = 0;
}
/**
* Keeps count of the last 4 bytes in order to detect line breaks on chunk boundaries
*
* @param {Buffer} data Next data chunk from the stream
*/
updateLastBytes(data) {
let lblen = this.lastBytes.length;
let nblen = Math.min(data.length, lblen);
// shift existing bytes
for (let i = 0, len = lblen - nblen; i < len; i++) {
this.lastBytes[i] = this.lastBytes[i + nblen];
}
// add new bytes
for (let i = 1; i <= nblen; i++) {
this.lastBytes[lblen - i] = data[data.length - i];
}
}
/**
* Finds and removes message headers from the remaining body. We want to keep
* headers separated until final delivery to be able to modify these
*
* @param {Buffer} data Next chunk of data
* @return {Boolean} Returns true if headers are already found or false otherwise
*/
checkHeaders(data) {
if (this.headersParsed) {
return true;
}
let lblen = this.lastBytes.length;
let headerPos = 0;
this.curLinePos = 0;
for (let i = 0, len = this.lastBytes.length + data.length; i < len; i++) {
let chr;
if (i < lblen) {
chr = this.lastBytes[i];
} else {
chr = data[i - lblen];
}
if (chr === 0x0a && i) {
let pr1 = i - 1 < lblen ? this.lastBytes[i - 1] : data[i - 1 - lblen];
let pr2 = i > 1 ? (i - 2 < lblen ? this.lastBytes[i - 2] : data[i - 2 - lblen]) : false;
if (pr1 === 0x0a) {
this.headersParsed = true;
headerPos = i - lblen + 1;
this.headerBytes += headerPos;
break;
} else if (pr1 === 0x0d && pr2 === 0x0a) {
this.headersParsed = true;
headerPos = i - lblen + 1;
this.headerBytes += headerPos;
break;
}
}
}
if (this.headersParsed) {
this.headerChunks.push(data.slice(0, headerPos));
this.rawHeaders = Buffer.concat(this.headerChunks, this.headerBytes);
this.headerChunks = null;
this.emit('headers', this.parseHeaders());
if (data.length - 1 > headerPos) {
let chunk = data.slice(headerPos);
this.bodySize += chunk.length;
// this would be the first chunk of data sent downstream
setImmediate(() => this.push(chunk));
}
return false;
} else {
this.headerBytes += data.length;
this.headerChunks.push(data);
}
// store last 4 bytes to catch header break
this.updateLastBytes(data);
return false;
}
_transform(chunk, encoding, callback) {
if (!chunk || !chunk.length) {
return callback();
}
if (typeof chunk === 'string') {
chunk = Buffer.from(chunk, encoding);
}
let headersFound;
try {
headersFound = this.checkHeaders(chunk);
} catch (E) {
return callback(E);
}
if (headersFound) {
this.bodySize += chunk.length;
this.push(chunk);
}
setImmediate(callback);
}
_flush(callback) {
if (this.headerChunks) {
let chunk = Buffer.concat(this.headerChunks, this.headerBytes);
this.bodySize += chunk.length;
this.push(chunk);
this.headerChunks = null;
}
callback();
}
parseHeaders() {
let lines = (this.rawHeaders || '').toString().split(/\r?\n/);
for (let i = lines.length - 1; i > 0; i--) {
if (/^\s/.test(lines[i])) {
lines[i - 1] += '\n' + lines[i];
lines.splice(i, 1);
}
}
return lines
.filter(line => line.trim())
.map(line => ({
key: line.substr(0, line.indexOf(':')).trim().toLowerCase(),
line
}));
}
}
module.exports = MessageParser;