transaction.js

'use strict'
// An SMTP Transaction

const util = require('node:util')

const Notes = require('haraka-notes')
const utils = require('haraka-utils')
const message = require('haraka-email-message')

class Transaction {
    constructor(uuid, cfg = {}) {
        this.uuid = uuid || utils.uuid()
        this.cfg = cfg
        this.mail_from = null
        this.rcpt_to = []
        this.header_lines = []
        this.data_lines = []
        this.attachment_start_hooks = []
        this.banner = null
        this.body_filters = []
        this.data_bytes = 0
        this.header_pos = 0
        this.found_hb_sep = false
        this.body = null
        this.parse_body = false
        this.notes = new Notes()
        this.notes.skip_plugins = []
        this.header = new message.Header()
        this.message_stream = new message.stream(this.cfg, this.uuid, this.header.header_list)
        this.discard_data = false
        this.resetting = false
        this.rcpt_count = {
            accept: 0,
            tempfail: 0,
            reject: 0,
        }
        this.msg_status = undefined
        this.data_post_start = null
        this.data_post_delay = 0
        this.encoding = 'utf8'
        this.mime_part_count = 0
    }

    ensure_body() {
        if (this.body) return

        this.body = new message.Body(this.header)
        this.body.on('mime_boundary', (m) => this.incr_mime_count())

        for (const hook of this.attachment_start_hooks) {
            this.body.on('attachment_start', hook)
        }

        if (this.banner) this.body.set_banner(this.banner)

        for (const o of this.body_filters) {
            this.body.add_filter((ct, enc, buf) => {
                const re_match = util.types.isRegExp(o.ct_match) && o.ct_match.test(ct.toLowerCase())
                const ct_begins = ct.toLowerCase().indexOf(String(o.ct_match).toLowerCase()) === 0
                if (re_match || ct_begins) return o.filter(ct, enc, buf)
            })
        }
    }

    // Removes the CR of a CRLF newline at the end of the buffer.
    remove_final_cr(data) {
        if (data.length < 2) return data
        if (!Buffer.isBuffer(data)) data = Buffer.from(data)

        if (data[data.length - 2] === 0x0d && data[data.length - 1] === 0x0a) {
            data[data.length - 2] = 0x0a
            return data.slice(0, data.length - 1)
        }
        return data
    }

    // Duplicates any '.' chars at the beginning of a line (dot-stuffing) and
    // ensures all newlines are CRLF.
    add_dot_stuffing_and_ensure_crlf_newlines(data) {
        if (!data.length) return data
        if (!Buffer.isBuffer(data)) data = Buffer.from(data)

        // Make a new buffer big enough to hold two bytes for every one input
        // byte.  At most, we add one extra character per input byte, so this
        // is always big enough.  We allocate it "unsafe" (i.e. no memset) for
        // speed because we're about to fill it with data, and the remainder of
        // the space we don't fill will be sliced away before we return this.
        const output = Buffer.allocUnsafe(data.length * 2)
        let output_pos = 0

        let input_pos = 0
        let next_dot = data.indexOf(0x2e)
        let next_lf = data.indexOf(0x0a)
        while (next_dot !== -1 || next_lf !== -1) {
            const run_end = next_dot !== -1 && (next_lf === -1 || next_dot < next_lf) ? next_dot : next_lf

            // Copy up to whichever comes first, '.' or '\n' (but don't
            // copy the '.' or '\n' itself).
            data.copy(output, output_pos, input_pos, run_end)
            output_pos += run_end - input_pos

            if (data[run_end] === 0x2e && (run_end === 0 || data[run_end - 1] === 0x0a)) {
                // Replace /^\./ with '..'
                output[output_pos++] = 0x2e
            } else if (data[run_end] === 0x0a && (run_end === 0 || data[run_end - 1] !== 0x0d)) {
                // Replace /\r?\n/ with '\r\n'
                output[output_pos++] = 0x0d
            }
            output[output_pos++] = data[run_end]

            input_pos = run_end + 1

            if (run_end === next_dot) {
                next_dot = data.indexOf(0x2e, input_pos)
            } else {
                next_lf = data.indexOf(0x0a, input_pos)
            }
        }

        if (input_pos < data.length) {
            data.copy(output, output_pos, input_pos)
            output_pos += data.length - input_pos
        }

        return output.slice(0, output_pos)
    }

    add_data(line) {
        if (typeof line === 'string') {
            // This shouldn't ever happen.
            line = Buffer.from(line, this.encoding)
        }
        // is this the end of headers line?
        if (this.header_pos === 0 && (line[0] === 0x0a || (line[0] === 0x0d && line[1] === 0x0a))) {
            this.header.parse(this.header_lines)
            this.header_pos = this.header_lines.length
            this.found_hb_sep = true
            if (this.parse_body) this.ensure_body()
        } else if (this.header_pos === 0) {
            // Build up headers
            if (this.header_lines.length < (this.cfg?.headers?.max_lines || 1000)) {
                if (line[0] === 0x2e) line = line.slice(1) // Strip leading '.'
                this.header_lines.push(line.toString(this.encoding).replace(/\r\n$/, '\n'))
            }
        } else if (this.parse_body) {
            let new_line = line
            if (new_line[0] === 0x2e) new_line = new_line.slice(1) // Strip leading "."

            line = this.add_dot_stuffing_and_ensure_crlf_newlines(this.body.parse_more(this.remove_final_cr(new_line)))

            if (!line.length) return // buffering for banners
        }

        if (!this.discard_data) this.message_stream.add_line(line)
    }

    end_data(cb) {
        if (!this.found_hb_sep && this.header_lines.length) {
            // Headers not parsed yet - must be a busted email
            // Strategy: Find the first line that doesn't look like a header.
            // Treat anything before that as headers, anything after as body.
            let header_pos = 0
            for (let i = 0; i < this.header_lines.length; i++) {
                // Anything that doesn't match a header or continuation
                if (!/^(?:([^\s:]*):\s*([\s\S]*)$|[ \t])/.test(this.header_lines[i])) {
                    break
                }
                header_pos = i
            }
            const body_lines = this.header_lines.splice(header_pos + 1)
            this.header.parse(this.header_lines)
            this.header_pos = header_pos
            if (this.parse_body) {
                this.ensure_body()
                for (const bodyLine of body_lines) {
                    this.body.parse_more(bodyLine)
                }
            }
        }
        if (this.header_pos && this.parse_body) {
            const line = this.add_dot_stuffing_and_ensure_crlf_newlines(this.body.parse_end())
            if (line.length) {
                this.body.force_end()

                if (!this.discard_data) this.message_stream.add_line(line)
            }
        }

        if (this.discard_data) {
            cb()
        } else {
            this.message_stream.add_line_end(cb)
        }
    }

    add_header(key, value) {
        this.header.add_end(key, value)
        if (this.header_pos > 0) this.reset_headers()
    }

    add_leading_header(key, value) {
        this.header.add(key, value)
        if (this.header_pos > 0) this.reset_headers()
    }

    reset_headers() {
        const header_lines = this.header.lines()
        this.header_pos = header_lines.length
    }

    remove_header(key) {
        this.header.remove(key)
        if (this.header_pos > 0) this.reset_headers()
    }

    attachment_hooks(start, data, end) {
        this.parse_body = true
        this.attachment_start_hooks.push(start)
    }

    set_banner(text, html) {
        // throw "transaction.set_banner is currently non-functional";
        this.parse_body = true
        if (!html) {
            html = text.replace(/\n/g, '<br/>\n')
        }
        this.banner = [text, html]
    }

    add_body_filter(ct_match, filter) {
        this.parse_body = true
        this.body_filters.push({ ct_match, filter })
    }

    incr_mime_count(line) {
        this.mime_part_count++
    }
}

exports.Transaction = Transaction

exports.createTransaction = (uuid, cfg) => {
    return new Transaction(uuid, cfg)
}