imapkit 0.0.0-stage → 4.0.0

package/lib/mimeparser.js

@@ -0,0 +1,778 @@
+'use strict';
+
+// Parses a message into a MIME tree. Ported from WildDuck (imap-core/lib/indexer/parse-mime-tree.js and
+// tree-walker.js), adapted to ImapKit, where message sources are binary strings (one char per octet).
+//
+// The tree describes the message exactly, so every section a client can FETCH is rendered from it:
+//
+//     entity    = header-lines [ CRLF body ]            ; "CRLF body" present unless hasBody is false
+//     body      = bytes | preamble *( delimiter entity CRLF ) close-delimiter epilogue
+//     delimiter = "--" boundary pad CRLF
+//
+// The CRLF after an entity belongs to the delimiter that follows it (RFC 2046 5.1.1), a part body never
+// includes it.
+
+const addressparser = require('./addressparser');
+
+// header field and parameter names that are kept in the parsed header, anything else is only available
+// through the raw header lines
+const FIELD_NAME = /^[a-zA-Z0-9\-*]{1,99}$/;
+
+// content transfer encodings that leave the data as it is (RFC 2045 section 6.2)
+const IDENTITY_ENCODINGS = ['7bit', '8bit', 'binary'];
+
+// message/rfc822 parts nested deeper than this are not parsed, they are described as plain parts
+const MAX_MESSAGE_DEPTH = 32;
+
+/**
+ * Converts every line break of a message to CRLF. IMAP serves messages with CRLF line breaks, so sizes and
+ * contents are computed from this form
+ *
+ * @param {String} raw Message source as a binary string
+ * @return {String} Message source with CRLF line breaks
+ */
+function normalizeLineBreaks(raw) {
+    const str = (raw || '').toString('binary');
+    // most messages already use CRLF, so only replace if there is a bare LF
+    return /(?:^|[^\r])\n/.test(str) ? str.replace(/\r?\n/g, '\r\n') : str;
+}
+
+/**
+ * Splits a structured header value (RFC 2045 5.1 Content-Type and friends) at a separator character,
+ * leaving quoted strings intact and dropping RFC 822 comments outside of them. With no separator the
+ * whole value comes back as one part with its comments removed.
+ *
+ * @param {String} value Header value
+ * @param {String} [separator] Character to split at, outside quotes and comments
+ * @returns {Array} Parts
+ */
+function splitStructuredValue(value, separator) {
+    const parts = [];
+    let current = '';
+    let quoted = false;
+    let depth = 0;
+
+    for (let i = 0; i < value.length; i++) {
+        const chr = value.charAt(i);
+
+        if (quoted) {
+            current += chr;
+            if (chr === '\\' && i + 1 < value.length) {
+                // quoted-pair, keep the escaped character for the value parser
+                current += value.charAt(++i);
+            } else if (chr === '"') {
+                quoted = false;
+            }
+            continue;
+        }
+
+        if (depth) {
+            // inside a comment, which may nest and may hold quoted-pairs
+            if (chr === '\\') {
+                i++;
+            } else if (chr === '(') {
+                depth++;
+            } else if (chr === ')') {
+                depth--;
+            }
+            continue;
+        }
+
+        if (chr === '"') {
+            quoted = true;
+            current += chr;
+        } else if (chr === '(') {
+            depth = 1;
+        } else if (separator && chr === separator) {
+            parts.push(current);
+            current = '';
+        } else {
+            current += chr;
+        }
+    }
+
+    parts.push(current);
+    return parts;
+}
+
+/**
+ * Decodes an RFC 2231 extended parameter value (percent encoded octets in a charset). UTF-8 and US-ASCII
+ * values become the UTF-8 octets, ISO-8859-1 values are converted to UTF-8, other charsets are kept as
+ * an RFC 2047 encoded word
+ *
+ * @param {String} charset Charset name from the first segment
+ * @param {String} value Percent encoded octets
+ * @return {String} Decoded value as a binary string
+ */
+function decodeExtendedValue(charset, value) {
+    const octets = value.replace(/%([0-9a-fA-F]{2})/g, (match, hex) => String.fromCharCode(parseInt(hex, 16)));
+    switch ((charset || '').trim().toLowerCase()) {
+        case '':
+        case 'utf-8':
+        case 'utf8':
+        case 'us-ascii':
+        case 'ascii':
+            return octets;
+        case 'iso-8859-1':
+        case 'latin1':
+            // every char of the binary string is an ISO-8859-1 character
+            return Buffer.from(octets, 'utf-8').toString('binary');
+        default:
+            return '=?' + charset.toUpperCase() + '?Q?' + value.replace(/%/g, '=') + '?=';
+    }
+}
+
+class MIMEParser {
+    /**
+     * @param {String} rfc822 Message source as a binary string with CRLF line breaks
+     * @param {Number} [depth] How many message/rfc822 levels this message is nested in
+     */
+    constructor(rfc822, depth) {
+        this.rfc822 = rfc822 || '';
+        this.depth = depth || 0;
+
+        // the line break that ended the last line read, false once the input is used up
+        this._br = '';
+        this._pos = 0;
+
+        this.tree = {
+            childNodes: []
+        };
+        this._node = this.createNode(this.tree);
+    }
+
+    /**
+     * Parses the message, line by line
+     */
+    parse() {
+        let line;
+        let prevBr = '';
+
+        // keep parsing until the last linebreak is not a string (no linebreaks anymore)
+        while (typeof this._br === 'string') {
+            line = this.readLine();
+
+            const delimiter = this.matchDelimiter(line);
+            if (delimiter) {
+                this.endPart(delimiter, prevBr);
+            } else {
+                switch (this._node.state) {
+                    case 'header':
+                        if (!line) {
+                            // the blank line that separates the header from the body
+                            this.endHeader();
+                        } else {
+                            this._node.header.push(line);
+                        }
+                        break;
+
+                    case 'body':
+                        // push the line with previous linebreak value, joined together the lines
+                        // give the original body
+                        this._node.body.push((this._node.body.length ? prevBr : '') + line);
+                        break;
+
+                    case 'epilogue':
+                        // RFC 2046 5.1.1: everything after the close delimiter up to the next delimiter
+                        // of the enclosing multipart (or the end of the message) is the epilogue. Every
+                        // epilogue line keeps the line break that precedes it, the first one being the
+                        // line break that ends the close delimiter line
+                        if (!this._node.epilogue) {
+                            this._node.epilogue = [];
+                        }
+                        this._node.epilogue.push(prevBr + line);
+                        break;
+
+                    default:
+                        // never should be reached
+                        throw new Error('Unexpected state');
+                }
+            }
+
+            // store the linebreak for later usage
+            prevBr = this._br;
+        }
+    }
+
+    /**
+     * Reads a line from the message
+     *
+     * @return {String} The line, without its line break
+     */
+    readLine() {
+        const end = this.rfc822.indexOf('\n', this._pos);
+        if (end < 0) {
+            // the remainder, which is empty when the input ended with a line break
+            const line = this.rfc822.slice(this._pos);
+            this._pos = this.rfc822.length;
+            this._br = false;
+            return line;
+        }
+
+        let lineEnd = end;
+        if (lineEnd > this._pos && this.rfc822.charCodeAt(lineEnd - 1) === 0x0d) {
+            lineEnd--;
+        }
+
+        const line = this.rfc822.slice(this._pos, lineEnd);
+        this._br = this.rfc822.slice(lineEnd, end + 1);
+        this._pos = end + 1;
+        return line;
+    }
+
+    /**
+     * Ends the header of the current node: its blank line was read, or a delimiter took its place
+     */
+    endHeader() {
+        this.processNodeHeader();
+        this.processContentType();
+        this._node.state = 'body';
+    }
+
+    /**
+     * Checks whether a line is a delimiter of an open multipart: the current node while it collects its
+     * preamble or epilogue, the multipart the current node belongs to, or any multipart above that (a
+     * lost close delimiter leaves the inner multipart open, the enclosing delimiter still ends it).
+     * RFC 2046 5.1.1: `--boundary` or `--boundary--`, optionally followed by transport padding (spaces
+     * and tabs) that receivers must accept. After the close delimiter, lines that look like the own
+     * delimiter are epilogue text.
+     *
+     * @param {String} line Line of the message
+     * @return {Object|Boolean} `{ multipart, close, pad }`, or false when the line is not a delimiter
+     */
+    matchDelimiter(line) {
+        if (!line.startsWith('--')) {
+            return false;
+        }
+
+        const node = this._node;
+        // a boundary is only known once the header has ended
+        const innermost = node.boundary ? node : node.parentNode;
+        for (let multipart = innermost; multipart.boundary; multipart = multipart.parentNode) {
+            const delimiter = '--' + multipart.boundary;
+            if (!line.startsWith(delimiter)) {
+                continue;
+            }
+
+            let rest = line.slice(delimiter.length);
+            const close = rest.startsWith('--');
+            if (close) {
+                rest = rest.slice(2);
+            }
+
+            if (/[^ \t]/.test(rest)) {
+                // something other than padding after the boundary, so a different boundary or text
+                continue;
+            }
+
+            if (multipart.state === 'epilogue') {
+                return false;
+            }
+
+            return { multipart, close, pad: rest };
+        }
+
+        return false;
+    }
+
+    /**
+     * Ends whatever the current node was collecting at a delimiter line and moves on to the next
+     * part of that multipart, or to its epilogue
+     *
+     * @param {Object} delimiter Result of matchDelimiter()
+     * @param {String} prevBr The line break that ended the line before the delimiter
+     */
+    endPart(delimiter, prevBr) {
+        const node = this._node;
+        const multipart = delimiter.multipart;
+
+        if (node !== multipart) {
+            // a part ends here
+            if (node.state === 'header') {
+                // a delimiter right after the header lines, with no blank line and no line break of
+                // its own (RFC 2046 5.1.1 wants one before every delimiter): the line break that ended
+                // the last header line, or the previous delimiter line, is the only one there is
+                this.endHeader();
+                // the multiparts a lost close delimiter leaves open between this part and the one the
+                // delimiter belongs to end with this part, so they end bare too
+                for (let open = node; open !== multipart; open = open.parentNode) {
+                    open.bare = true;
+                }
+            }
+        } else if (node.body.length) {
+            // the preamble ends here. The line break between the preamble and the delimiter belongs to
+            // the delimiter, but the preamble lines end with their own
+            node.body[node.body.length - 1] += prevBr;
+        }
+
+        if (!delimiter.close) {
+            const next = this.createNode(multipart);
+            if (delimiter.pad) {
+                next.pad = delimiter.pad;
+            }
+            this._node = next;
+        } else {
+            if (delimiter.pad) {
+                multipart.closePad = delimiter.pad;
+            }
+            delete multipart.unterminated;
+            multipart.state = 'epilogue';
+            this._node = multipart;
+        }
+    }
+
+    /**
+     * Parses the body of a message/rfc822 node into `node.message`. RFC 2046 5.2.1 only allows the
+     * identity encodings for such a body, RFC 2045 5.1 and 6.1 make the type and encoding names case
+     * insensitive. A message/global body (RFC 6532 section 3.5) goes to `node.globalMessage`, as only
+     * IMAP4rev2 treats it like message/rfc822 (RFC 9051 section 7.5.2), see embeddedMessage()
+     */
+    parseEmbeddedMessage(node) {
+        const contentType = node.parsedHeader['content-type'];
+        const type = ((contentType && contentType.value) || '').toLowerCase();
+        if ((type !== 'message/rfc822' && type !== 'message/global') || this.depth >= MAX_MESSAGE_DEPTH) {
+            return;
+        }
+        const encoding = (node.parsedHeader['content-transfer-encoding'] || '').toString().trim().toLowerCase();
+        if (encoding && !IDENTITY_ENCODINGS.includes(encoding)) {
+            return;
+        }
+        node[type === 'message/rfc822' ? 'message' : 'globalMessage'] = parseTree(node.body, this.depth + 1);
+    }
+
+    /**
+     * Joins body arrays into strings. Removes unnecessary fields from the tree
+     */
+    finalizeTree() {
+        if (this._node.state === 'header') {
+            this.endHeader();
+            if (this._node.header.length && !this.rfc822.endsWith('\n')) {
+                // the input ended within the last header line, there is no line break to render
+                this._node.headerUnterminated = true;
+            }
+        }
+
+        // iterative, so that deeply nested multiparts do not exhaust the stack
+        const stack = [].concat(this.tree.childNodes);
+        while (stack.length) {
+            const node = stack.pop();
+
+            // a body section holds lines, parts, or at least a close delimiter. A blank line followed by
+            // the end of the input leaves the trailing empty line in the body, one followed directly by
+            // a delimiter leaves nothing
+            const closed = node.boundary && !node.unterminated;
+            if (!(node.body.length || node.childNodes.length || closed)) {
+                node.hasBody = false;
+            }
+
+            // RFC 3501 7.4.2 body-fld-lines, counted like Dovecot does: the line breaks in the body
+            node.lineCount = node.body.length ? node.body.length - 1 : 0;
+            node.body = node.body.join('');
+            node.size = node.body.length;
+
+            // a message/rfc822 entity carries its parsed message
+            this.parseEmbeddedMessage(node);
+
+            if (node.epilogue) {
+                node.epilogue = node.epilogue.join('');
+            }
+
+            node.childNodes.forEach(child => stack.push(child));
+
+            // remove unneeded properties
+            delete node.parentNode;
+            delete node.state;
+            if (!node.childNodes.length) {
+                delete node.childNodes;
+            }
+        }
+    }
+
+    /**
+     * Creates a new node with default values for the parse tree
+     */
+    createNode(parentNode) {
+        const node = {
+            state: 'header',
+            childNodes: [],
+            header: [],
+            parsedHeader: Object.create(null),
+            body: [],
+            multipart: false,
+            boundary: false,
+            parentNode
+        };
+        parentNode.childNodes.push(node);
+        return node;
+    }
+
+    /**
+     * Processes header lines. Splits lines to key-value pairs
+     * and processes special values
+     */
+    processNodeHeader() {
+        const node = this._node;
+
+        // RFC 5322 2.2.3: a line that starts with whitespace continues the previous header field
+        const header = [];
+        for (const line of node.header) {
+            if (header.length && /^\s/.test(line)) {
+                header[header.length - 1] += '\r\n' + line;
+            } else {
+                header.push(line);
+            }
+        }
+        node.header = header;
+
+        for (const line of header) {
+            let value = line.split(':');
+            const key = (value.shift() || '').trim().toLowerCase();
+            value = value.join(':').trim();
+
+            // Do not touch headers that have strange looking keys, keep these
+            // only in the unparsed array
+            if (!FIELD_NAME.test(key)) {
+                continue;
+            }
+
+            if (key in node.parsedHeader) {
+                if (Array.isArray(node.parsedHeader[key])) {
+                    node.parsedHeader[key].push(value);
+                } else {
+                    node.parsedHeader[key] = [node.parsedHeader[key], value];
+                }
+            } else {
+                node.parsedHeader[key] = value.replace(/\s*\r?\n\s*/g, ' ');
+            }
+        }
+
+        // always ensure the presence of Content-Type. RFC 2046 5.1.5: inside a digest the default
+        // is message/rfc822 instead of text/plain
+        if (!node.parsedHeader['content-type']) {
+            const parentSubtype = ((node.parentNode && node.parentNode.multipart) || '').toString().toLowerCase();
+            node.parsedHeader['content-type'] = parentSubtype === 'digest' ? 'message/rfc822' : 'text/plain';
+        }
+
+        // parse additional params for Content-Type and Content-Disposition, the last header wins
+        ['content-type', 'content-disposition'].forEach(key => {
+            if (node.parsedHeader[key]) {
+                node.parsedHeader[key] = this.parseValueParams([].concat(node.parsedHeader[key]).pop());
+            }
+        });
+
+        // ensure single value for selected fields, the last header wins
+        [
+            'subject',
+            'date',
+            'in-reply-to',
+            'message-id',
+            'content-transfer-encoding',
+            'content-id',
+            'content-description',
+            'content-language',
+            'content-md5',
+            'content-location'
+        ].forEach(key => {
+            if (Array.isArray(node.parsedHeader[key])) {
+                node.parsedHeader[key] = node.parsedHeader[key].pop().replace(/\s*\r?\n\s*/g, ' ');
+            }
+        });
+
+        if (node.parsedHeader['content-transfer-encoding']) {
+            // RFC 2045 6.1: the mechanism token may be followed by a comment, which is not part of it
+            node.parsedHeader['content-transfer-encoding'] = splitStructuredValue(node.parsedHeader['content-transfer-encoding'])[0].trim();
+        }
+
+        // Parse address fields (join several fields with same key)
+        ['from', 'sender', 'reply-to', 'to', 'cc', 'bcc'].forEach(key => {
+            if (node.parsedHeader[key]) {
+                node.parsedHeader[key] = [].concat(node.parsedHeader[key]).flatMap(value => (value && addressparser(value.replace(/\s*\r?\n\s*/g, ' '))) || []);
+            }
+        });
+    }
+
+    /**
+     * Splits a value to an object.
+     * eg. 'text/plain; charset=utf-8' -> {value: 'text/plain', params:{charset: 'utf-8'}}
+     *
+     * @param {String} headerValue A string value for a header key
+     * @return {Object} Parsed value
+     */
+    parseValueParams(headerValue) {
+        const data = {
+            value: '',
+            type: '',
+            subtype: '',
+            params: Object.create(null)
+        };
+
+        // RFC 2231 continuations and extended values, by parameter name
+        const continuations = Object.create(null);
+
+        splitStructuredValue(headerValue || '', ';').forEach((part, i) => {
+            if (!i) {
+                data.value = part.trim();
+                const subtype = data.value.split('/');
+                data.type = (subtype.shift() || '').toLowerCase();
+                data.subtype = subtype.join('/');
+                return;
+            }
+
+            let value = part.split('=');
+            const key = (value.shift() || '').trim().toLowerCase();
+            value = value.join('=').trim();
+            if (value.length >= 2 && value.startsWith('"') && value.endsWith('"')) {
+                // RFC 2045 5.1 quoted-string: the quotes are not part of the value and a backslash
+                // quotes the character after it (RFC 822 3.3)
+                value = value.slice(1, -1).replace(/\\([\s\S])/g, '$1');
+            } else {
+                value = value.replace(/^['"\s]*|['"\s]*$/g, '');
+            }
+
+            // Do not touch parameters that have strange looking keys
+            if (!FIELD_NAME.test(key)) {
+                return;
+            }
+
+            // RFC 2231 3 and 4: `name*N` is segment N of a continued value, a trailing asterisk marks
+            // a segment as extended (percent encoded, the first one prefixed with charset'language')
+            const match = key.match(/^([^*]+)(?:\*(\d{1,3}))?(\*)?$/);
+            if (match && (match[2] !== undefined || match[3])) {
+                const name = match[1];
+                if (!continuations[name]) {
+                    continuations[name] = [];
+                }
+                continuations[name][Number(match[2]) || 0] = { value, extended: !!match[3] };
+            } else {
+                data.params[key] = value;
+            }
+            data.hasParams = true;
+        });
+
+        Object.keys(continuations).forEach(key => {
+            let charset = '';
+            let encoded = '';
+            let extended = false;
+
+            continuations[key].forEach((segment, i) => {
+                if (!segment) {
+                    return;
+                }
+                let value = segment.value;
+                if (segment.extended && !i) {
+                    const parts = value.split("'");
+                    if (parts.length >= 3) {
+                        charset = parts.shift();
+                        parts.shift(); // language, ignored
+                        value = parts.join("'");
+                    }
+                }
+                if (segment.extended) {
+                    extended = true;
+                    encoded += value;
+                } else {
+                    // a segment that is not extended is plain text, encode it like the rest
+                    encoded += value.replace(/[^\x21-\x24\x26-\x7e]/g, chr => '%' + ('0' + chr.charCodeAt(0).toString(16)).substr(-2));
+                }
+            });
+
+            data.params[key] = extended ? decodeExtendedValue(charset, encoded) : decodeExtendedValue('', encoded);
+        });
+
+        return data;
+    }
+
+    /**
+     * Checks Content-Type value for the current tree node.
+     */
+    processContentType() {
+        const node = this._node;
+        // processNodeHeader() always sets a Content-Type
+        const contentType = node.parsedHeader['content-type'];
+
+        if (contentType.type === 'multipart' && contentType.params.boundary) {
+            node.multipart = contentType.subtype;
+            node.boundary = contentType.params.boundary;
+            // until the close delimiter is seen
+            node.unterminated = true;
+        }
+    }
+}
+
+/**
+ * Parses a message into a MIME tree
+ *
+ * @param {String} rfc822 Message source as a binary string with CRLF line breaks
+ * @param {Number} [depth] Nesting level of message/rfc822 parts
+ * @return {Object} Root node of the tree
+ */
+function parseTree(rfc822, depth) {
+    const parser = new MIMEParser(rfc822, depth);
+    parser.parse();
+    parser.finalizeTree();
+    return parser.tree.childNodes[0];
+}
+
+// RFC 3501 9: body-type-mpart = 1*body SP media-subtype. A multipart whose boundary never appeared has
+// no parts, so this empty text part stands in for them
+const PLACEHOLDER_PART = parseTree('Content-Type: text/plain\r\n\r\n');
+
+/**
+ * The parts of a node as IMAP numbers them: the parts of a multipart (the placeholder when it has none),
+ * undefined for anything else
+ *
+ * @param {Object} node A tree node
+ * @returns {Array|undefined} Part nodes
+ */
+function partsOf(node) {
+    if (node.childNodes) {
+        return node.childNodes;
+    }
+    return node.boundary ? [PLACEHOLDER_PART] : undefined;
+}
+
+/**
+ * The message encapsulated in a node: the message of a message/rfc822 part, and with `global` also of a
+ * message/global part (RFC 9051 sections 6.4.5.1 and 7.5.2, IMAP4rev1 treats message/global as a basic part)
+ *
+ * @param {Object} node A tree node
+ * @param {Boolean} [global] Treat message/global like message/rfc822
+ * @returns {Object|undefined} Root node of the encapsulated message
+ */
+function embeddedMessage(node, global) {
+    return node.message || (global ? node.globalMessage : undefined);
+}
+
+/**
+ * Resolves a numeric part path to a node. RFC 3501 6.4.5: the parts of a multipart are numbered from 1,
+ * a non-multipart message has one part which is the message itself, and the parts of a message/rfc822
+ * part are numbered under it
+ *
+ * @param {Object} tree Root node
+ * @param {String} path Dot-separated numeric path
+ * @param {Boolean} [global] Number the parts of message/global parts too, like IMAP4rev2 does
+ * @return {Object|Boolean} Node, or false when there is no such part
+ */
+function resolveNode(tree, path, global) {
+    // the message whose parts the next number counts
+    let scope = tree;
+    let node = tree;
+
+    for (const number of (path || '').toString().split('.')) {
+        const index = Number(number) - 1;
+        if (!scope || !(index >= 0)) {
+            return false;
+        }
+
+        const parts = partsOf(scope);
+        node = parts ? parts[index] : index === 0 ? scope : undefined;
+        if (!node) {
+            return false;
+        }
+
+        scope = embeddedMessage(node, global) || (partsOf(node) ? node : false);
+    }
+
+    return node;
+}
+
+/**
+ * The header of a node as the HEADER and MIME sections return it, with the blank line that ends it
+ *
+ * @param {Object} node Tree node
+ * @return {String} Header section
+ */
+function headerSection(node) {
+    const header = node.header || [];
+    if (node.hasBody === false) {
+        // RFC 3501 7.4.2: the blank line is part of the header, except for a message with no body and no blank line
+        return header.length ? header.join('\r\n') + (node.headerUnterminated ? '' : '\r\n') : '';
+    }
+    return header.length ? header.join('\r\n') + '\r\n\r\n' : '\r\n';
+}
+
+/**
+ * Renders a node of the tree back to the octets it was parsed from
+ *
+ * @param {Object} node Tree node
+ * @param {Boolean} [textOnly] Leave out the header of the node (BODY[TEXT], BODY[n])
+ * @return {String} Binary string
+ */
+function render(node, textOnly) {
+    const output = [];
+
+    const walk = (node, withHeader) => {
+        if (withHeader) {
+            const header = node.header || [];
+            if (header.length) {
+                output.push(header.join('\r\n') + (node.headerUnterminated ? '' : '\r\n'));
+            }
+        }
+
+        if (node.hasBody === false) {
+            return;
+        }
+
+        if (withHeader) {
+            // the blank line between header and body
+            output.push('\r\n');
+        }
+
+        output.push(node.body || '');
+
+        if (!node.boundary) {
+            return;
+        }
+
+        const children = node.childNodes || [];
+        for (let i = 0; i < children.length; i++) {
+            const child = children[i];
+            output.push('--' + node.boundary + (child.pad || '') + '\r\n');
+            walk(child, true);
+            if ((!node.unterminated || i < children.length - 1) && !child.bare) {
+                // the line break that belongs to the next delimiter
+                output.push('\r\n');
+            }
+        }
+
+        if (!node.unterminated) {
+            output.push('--' + node.boundary + '--' + (node.closePad || ''));
+            output.push(node.epilogue || '');
+        }
+    };
+
+    walk(node, !textOnly);
+    return output.join('');
+}
+
+// Parsed data of messages, so a message is parsed only once and not again until its source changes
+const cache = new WeakMap();
+
+/**
+ * Returns the source with CRLF line breaks and the MIME tree of a message
+ *
+ * @param {Object} message Message object with a `raw` property
+ * @return {Object} `{ raw, tree }`
+ */
+function getMessageData(message) {
+    let cached = cache.get(message);
+    if (!cached || cached.source !== message.raw) {
+        const raw = normalizeLineBreaks(message.raw);
+        cached = { source: message.raw, raw, tree: parseTree(raw) };
+        cache.set(message, cached);
+    }
+    return cached;
+}
+
+module.exports = function (rfc822) {
+    return parseTree(normalizeLineBreaks(rfc822));
+};
+module.exports.parseTree = parseTree;
+module.exports.IDENTITY_ENCODINGS = IDENTITY_ENCODINGS;
+module.exports.getMessageData = getMessageData;
+module.exports.normalizeLineBreaks = normalizeLineBreaks;
+module.exports.partsOf = partsOf;
+module.exports.resolveNode = resolveNode;
+module.exports.embeddedMessage = embeddedMessage;
+module.exports.headerSection = headerSection;
+module.exports.render = render;