main.go

package main

import (
	"bufio"
	"crypto/sha256"
	"encoding/binary"
	"encoding/hex"
	"encoding/json"
	"errors"
	"flag"
	"fmt"
	"log"
	"math"
	"net/url"
	"os"
	"os/user"
	"path"
	"path/filepath"
	"strconv"
	"strings"
	"unicode"

	"github.com/coreos/go-semver/semver"
	"github.com/curusarn/resh/pkg/records"
	giturls "github.com/whilp/git-urls"
)

// version from git set during build
var version string

// commit from git set during build
var commit string

func main() {
	usr, _ := user.Current()
	dir := usr.HomeDir
	historyPath := filepath.Join(dir, ".resh_history.json")
	// outputPath := filepath.Join(dir, "resh_history_sanitized.json")
	sanitizerDataPath := filepath.Join(dir, ".resh", "sanitizer_data")

	showVersion := flag.Bool("version", false, "Show version and exit")
	showRevision := flag.Bool("revision", false, "Show git revision and exit")
	trimHashes := flag.Int("trim-hashes", 12, "Trim hashes to N characters, '0' turns off trimming")
	inputPath := flag.String("input", historyPath, "Input file")
	outputPath := flag.String("output", "", "Output file (default: use stdout)")

	flag.Parse()

	if *showVersion == true {
		fmt.Println(version)
		os.Exit(0)
	}
	if *showRevision == true {
		fmt.Println(commit)
		os.Exit(0)
	}
	sanitizer := sanitizer{hashLength: *trimHashes}
	err := sanitizer.init(sanitizerDataPath)
	if err != nil {
		log.Fatal("Sanitizer init() error:", err)
	}

	inputFile, err := os.Open(*inputPath)
	if err != nil {
		log.Fatal("Open() resh history file error:", err)
	}
	defer inputFile.Close()

	var writer *bufio.Writer
	if *outputPath == "" {
		writer = bufio.NewWriter(os.Stdout)
	} else {
		outputFile, err := os.Create(*outputPath)
		if err != nil {
			log.Fatal("Create() output file error:", err)
		}
		defer outputFile.Close()
		writer = bufio.NewWriter(outputFile)
	}
	defer writer.Flush()

	scanner := bufio.NewScanner(inputFile)
	for scanner.Scan() {
		record := records.Record{}
		fallbackRecord := records.FallbackRecord{}
		line := scanner.Text()
		err = json.Unmarshal([]byte(line), &record)
		if err != nil {
			err = json.Unmarshal([]byte(line), &fallbackRecord)
			if err != nil {
				log.Println("Line:", line)
				log.Fatal("Decoding error:", err)
			}
			record = records.Convert(&fallbackRecord)
		}
		err = sanitizer.sanitizeRecord(&record)
		if err != nil {
			log.Println("Line:", line)
			log.Fatal("Sanitization error:", err)
		}
		outLine, err := json.Marshal(&record)
		if err != nil {
			log.Println("Line:", line)
			log.Fatal("Encoding error:", err)
		}
		// fmt.Println(string(outLine))
		n, err := writer.WriteString(string(outLine) + "\n")
		if err != nil {
			log.Fatal(err)
		}
		if n == 0 {
			log.Fatal("Nothing was written", n)
		}
	}
}

type sanitizer struct {
	hashLength int
	whitelist  map[string]bool
}

func (s *sanitizer) init(dataPath string) error {
	globalData := path.Join(dataPath, "whitelist.txt")
	s.whitelist = loadData(globalData)
	return nil
}

func loadData(fname string) map[string]bool {
	file, err := os.Open(fname)
	if err != nil {
		log.Fatal("Open() file error:", err)
	}
	defer file.Close()

	scanner := bufio.NewScanner(file)
	data := make(map[string]bool)
	for scanner.Scan() {
		line := scanner.Text()
		data[line] = true
	}
	return data
}

func (s *sanitizer) sanitizeRecord(record *records.Record) error {
	// hash directories of the paths
	record.Pwd = s.sanitizePath(record.Pwd)
	record.RealPwd = s.sanitizePath(record.RealPwd)
	record.PwdAfter = s.sanitizePath(record.PwdAfter)
	record.RealPwdAfter = s.sanitizePath(record.RealPwdAfter)
	record.GitDir = s.sanitizePath(record.GitDir)
	record.GitDirAfter = s.sanitizePath(record.GitDirAfter)
	record.GitRealDir = s.sanitizePath(record.GitRealDir)
	record.GitRealDirAfter = s.sanitizePath(record.GitRealDirAfter)
	record.Home = s.sanitizePath(record.Home)
	record.ShellEnv = s.sanitizePath(record.ShellEnv)

	// hash the most sensitive info, do not tokenize
	record.Host = s.hashToken(record.Host)
	record.Login = s.hashToken(record.Login)
	record.MachineID = s.hashToken(record.MachineID)

	var err error
	// this changes git url a bit but I'm still happy with the result
	// e.g. "git@github.com:curusarn/resh" becomes "ssh://git@github.com/3385162f14d7/5a7b2909005c"
	// 		notice the "ssh://" prefix
	record.GitOriginRemote, err = s.sanitizeGitURL(record.GitOriginRemote)
	if err != nil {
		log.Println("Error while snitizing GitOriginRemote url", record.GitOriginRemote, ":", err)
		return err
	}
	record.GitOriginRemoteAfter, err = s.sanitizeGitURL(record.GitOriginRemoteAfter)
	if err != nil {
		log.Println("Error while snitizing GitOriginRemoteAfter url", record.GitOriginRemoteAfter, ":", err)
		return err
	}

	// sanitization destroys original CmdLine length -> save it
	record.CmdLength = len(record.CmdLine)

	record.CmdLine, err = s.sanitizeCmdLine(record.CmdLine)
	if err != nil {
		log.Fatal("Cmd:", record.CmdLine, "; sanitization error:", err)
	}
	record.RecallLastCmdLine, err = s.sanitizeCmdLine(record.RecallLastCmdLine)
	if err != nil {
		log.Fatal("RecallLastCmdLine:", record.RecallLastCmdLine, "; sanitization error:", err)
	}

	if len(record.RecallActionsRaw) > 0 {
		record.RecallActionsRaw, err = s.sanitizeRecallActions(record.RecallActionsRaw, record.ReshVersion)
		if err != nil {
			log.Println("RecallActionsRaw:", record.RecallActionsRaw, "; sanitization error:", err)
		}
	}
	// add a flag to signify that the record has been sanitized
	record.Sanitized = true
	return nil
}

func fixSeparator(str string) string {
	if len(str) > 0 && str[0] == ';' {
		return "|||" + str[1:]
	}
	return str
}

func minIndex(str string, substrs []string) (idx, substrIdx int) {
	minMatch := math.MaxInt32
	for i, sep := range substrs {
		match := strings.Index(str, sep)
		if match != -1 && match < minMatch {
			minMatch = match
			substrIdx = i
		}
	}
	idx = minMatch
	return
}

// sanitizes the recall actions by replacing the recall prefix with it's length
func (s *sanitizer) sanitizeRecallActions(str string, reshVersion string) (string, error) {
	if len(str) == 0 {
		return "", nil
	}
	var separators []string
	seps := []string{"|||"}
	refVersion, err := semver.NewVersion("2.5.14")
	if err != nil {
		return str, fmt.Errorf("sanitizeRecallActions: semver error: %s", err.Error())
	}
	if len(reshVersion) == 0 {
		return str, errors.New("sanitizeRecallActions: record.ReshVersion is an empty string")
	}
	if reshVersion == "dev" {
		reshVersion = "0.0.0"
	}
	if reshVersion[0] == 'v' {
		reshVersion = reshVersion[1:]
	}
	recordVersion, err := semver.NewVersion(reshVersion)
	if err != nil {
		return str, fmt.Errorf("sanitizeRecallActions: semver error: %s; version string: %s", err.Error(), reshVersion)
	}
	if recordVersion.LessThan(*refVersion) {
		seps = append(seps, ";")
	}

	actions := []string{"arrow_up", "arrow_down", "control_R"}
	for _, sep := range seps {
		for _, action := range actions {
			separators = append(separators, sep+action+":")
		}
	}
	/*
		- find any of {|||,;}{arrow_up,arrow_down,control_R}: in the recallActions (on the lowest index)
		- use found substring to parse out the next prefix
		- sanitize prefix
		- add fixed substring and sanitized prefix to output
	*/
	doBreak := false
	sanStr := ""
	idx := 0
	var currSeparator string
	tokenLen, sepIdx := minIndex(str, separators)
	if tokenLen != 0 {
		return str, errors.New("sanitizeReacallActions: unexpected string before first action/separator")
	}
	currSeparator = separators[sepIdx]
	idx += len(currSeparator)
	for !doBreak {
		tokenLen, sepIdx := minIndex(str[idx:], separators)
		if tokenLen > len(str[idx:]) {
			tokenLen = len(str[idx:])
			doBreak = true
		}
		// token := str[idx : idx+tokenLen]
		sanStr += fixSeparator(currSeparator) + strconv.Itoa(tokenLen)
		currSeparator = separators[sepIdx]
		idx += tokenLen + len(currSeparator)
	}
	return sanStr, nil
}

func (s *sanitizer) sanitizeCmdLine(cmdLine string) (string, error) {
	const optionEndingChars = "\"$'\\#[]!><|;{}()*,?~&=`:@^/+%." // all bash control characters, '=', ...
	const optionAllowedChars = "-_"                              // characters commonly found inside of options
	sanCmdLine := ""
	buff := ""

	// simple options shouldn't be sanitized
	// 1) whitespace 2) "-" or "--" 3) letters, digits, "-", "_" 4) ending whitespace or any of "=;)"
	var optionDetected bool

	prevR3 := ' '
	prevR2 := ' '
	prevR := ' '
	for _, r := range cmdLine {
		switch optionDetected {
		case true:
			if unicode.IsSpace(r) || strings.ContainsRune(optionEndingChars, r) {
				// whitespace or option ends the option
				// => add option unsanitized
				optionDetected = false
				if len(buff) > 0 {
					sanCmdLine += buff
					buff = ""
				}
				sanCmdLine += string(r)
			} else if unicode.IsLetter(r) == false && unicode.IsDigit(r) == false &&
				strings.ContainsRune(optionAllowedChars, r) == false {
				// r is not any of allowed chars for an option: letter, digit, "-" or "_"
				// => sanitize
				if len(buff) > 0 {
					sanToken, err := s.sanitizeCmdToken(buff)
					if err != nil {
						log.Println("WARN: got error while sanitizing cmdLine:", cmdLine)
						// return cmdLine, err
					}
					sanCmdLine += sanToken
					buff = ""
				}
				sanCmdLine += string(r)
			} else {
				buff += string(r)
			}
		case false:
			// split command on all non-letter and non-digit characters
			if unicode.IsLetter(r) == false && unicode.IsDigit(r) == false {
				// split token
				if len(buff) > 0 {
					sanToken, err := s.sanitizeCmdToken(buff)
					if err != nil {
						log.Println("WARN: got error while sanitizing cmdLine:", cmdLine)
						// return cmdLine, err
					}
					sanCmdLine += sanToken
					buff = ""
				}
				sanCmdLine += string(r)
			} else {
				if (unicode.IsSpace(prevR2) && prevR == '-') ||
					(unicode.IsSpace(prevR3) && prevR2 == '-' && prevR == '-') {
					optionDetected = true
				}
				buff += string(r)
			}
		}
		prevR3 = prevR2
		prevR2 = prevR
		prevR = r
	}
	if len(buff) <= 0 {
		// nothing in the buffer => work is done
		return sanCmdLine, nil
	}
	if optionDetected {
		// option detected => dont sanitize
		sanCmdLine += buff
		return sanCmdLine, nil
	}
	// sanitize
	sanToken, err := s.sanitizeCmdToken(buff)
	if err != nil {
		log.Println("WARN: got error while sanitizing cmdLine:", cmdLine)
		// return cmdLine, err
	}
	sanCmdLine += sanToken
	return sanCmdLine, nil
}

func (s *sanitizer) sanitizeGitURL(rawURL string) (string, error) {
	if len(rawURL) <= 0 {
		return rawURL, nil
	}
	parsedURL, err := giturls.Parse(rawURL)
	if err != nil {
		return rawURL, err
	}
	return s.sanitizeParsedURL(parsedURL)
}

func (s *sanitizer) sanitizeURL(rawURL string) (string, error) {
	if len(rawURL) <= 0 {
		return rawURL, nil
	}
	parsedURL, err := url.Parse(rawURL)
	if err != nil {
		return rawURL, err
	}
	return s.sanitizeParsedURL(parsedURL)
}

func (s *sanitizer) sanitizeParsedURL(parsedURL *url.URL) (string, error) {
	parsedURL.Opaque = s.sanitizeToken(parsedURL.Opaque)

	userinfo := parsedURL.User.Username() // only get username => password won't even make it to the sanitized data
	if len(userinfo) > 0 {
		parsedURL.User = url.User(s.sanitizeToken(userinfo))
	} else {
		// we need to do this because `gitUrls.Parse()` sets `User` to `url.User("")` instead of `nil`
		parsedURL.User = nil
	}
	var err error
	parsedURL.Host, err = s.sanitizeTwoPartToken(parsedURL.Host, ":")
	if err != nil {
		return parsedURL.String(), err
	}
	parsedURL.Path = s.sanitizePath(parsedURL.Path)
	// ForceQuery bool
	parsedURL.RawQuery = s.sanitizeToken(parsedURL.RawQuery)
	parsedURL.Fragment = s.sanitizeToken(parsedURL.Fragment)

	return parsedURL.String(), nil
}

func (s *sanitizer) sanitizePath(path string) string {
	var sanPath string
	for _, token := range strings.Split(path, "/") {
		if s.whitelist[token] != true {
			token = s.hashToken(token)
		}
		sanPath += token + "/"
	}
	if len(sanPath) > 0 {
		sanPath = sanPath[:len(sanPath)-1]
	}
	return sanPath
}

func (s *sanitizer) sanitizeTwoPartToken(token string, delimeter string) (string, error) {
	tokenParts := strings.Split(token, delimeter)
	if len(tokenParts) <= 1 {
		return s.sanitizeToken(token), nil
	}
	if len(tokenParts) == 2 {
		return s.sanitizeToken(tokenParts[0]) + delimeter + s.sanitizeToken(tokenParts[1]), nil
	}
	return token, errors.New("Token has more than two parts")
}

func (s *sanitizer) sanitizeCmdToken(token string) (string, error) {
	// there shouldn't be tokens with letters or digits mixed together with symbols
	if len(token) <= 1 {
		// NOTE: do not sanitize single letter tokens
		return token, nil
	}
	if s.isInWhitelist(token) == true {
		return token, nil
	}

	isLettersOrDigits := true
	// isDigits := true
	isOtherCharacters := true
	for _, r := range token {
		if unicode.IsDigit(r) == false && unicode.IsLetter(r) == false {
			isLettersOrDigits = false
			// isDigits = false
		}
		// if unicode.IsDigit(r) == false {
		// 	isDigits = false
		// }
		if unicode.IsDigit(r) || unicode.IsLetter(r) {
			isOtherCharacters = false
		}
	}
	// NOTE: I decided that I don't want a special sanitization for numbers
	// if isDigits {
	// 	return s.hashNumericToken(token), nil
	// }
	if isLettersOrDigits {
		return s.hashToken(token), nil
	}
	if isOtherCharacters {
		return token, nil
	}
	log.Println("WARN: cmd token is made of mix of letters or digits and other characters; token:", token)
	// return token, errors.New("cmd token is made of mix of letters or digits and other characters")
	return s.hashToken(token), errors.New("cmd token is made of mix of letters or digits and other characters")
}

func (s *sanitizer) sanitizeToken(token string) string {
	if len(token) <= 1 {
		// NOTE: do not sanitize single letter tokens
		return token
	}
	if s.isInWhitelist(token) {
		return token
	}
	return s.hashToken(token)
}

func (s *sanitizer) hashToken(token string) string {
	if len(token) <= 0 {
		return token
	}
	// hash with sha256
	sum := sha256.Sum256([]byte(token))
	return s.trimHash(hex.EncodeToString(sum[:]))
}

func (s *sanitizer) hashNumericToken(token string) string {
	if len(token) <= 0 {
		return token
	}
	sum := sha256.Sum256([]byte(token))
	sumInt := int(binary.LittleEndian.Uint64(sum[:]))
	if sumInt < 0 {
		return strconv.Itoa(sumInt * -1)
	}
	return s.trimHash(strconv.Itoa(sumInt))
}

func (s *sanitizer) trimHash(hash string) string {
	length := s.hashLength
	if length <= 0 || len(hash) < length {
		length = len(hash)
	}
	return hash[:length]
}

func (s *sanitizer) isInWhitelist(token string) bool {
	return s.whitelist[strings.ToLower(token)] == true
}