gpt-plus-gpt/pkg/stripe/sha256.go

package stripe

import (
	"crypto/sha256"
	"encoding/base64"
	"fmt"
	"net/url"
	"regexp"
	"strings"
	"time"
)

// SHA256URLSafe computes SHA-256 and returns URL-safe base64 (no padding).
// Matches JS: btoa(sha256_raw(input)).replace(/\+/g,"-").replace(/\//g,"_").replace(/=+$/,"")
func SHA256URLSafe(input string) string {
	h := sha256.Sum256([]byte(input))
	return base64.RawURLEncoding.EncodeToString(h[:])
}

const urlSalt = "7766e861-8279-424d-87a1-07a6022fd8cd"

// SHA256WithSalt hashes input with the Stripe URL salt.
// Matches JS: sha256WithSalt which does sha256(unescape(encodeURIComponent(e)) + URL_SALT)
func SHA256WithSalt(input string) string {
	if input == "" {
		return ""
	}
	// JS unescape(encodeURIComponent(s)) converts to UTF-8 bytes, which Go strings already are.
	return SHA256URLSafe(input + urlSalt)
}

// EncodeURIComponent mimics JavaScript's encodeURIComponent.
func EncodeURIComponent(s string) string {
	result := url.QueryEscape(s)
	result = strings.ReplaceAll(result, "+", "%20")
	// JS encodeURIComponent does NOT encode: - _ . ! ~ * ' ( )
	for _, c := range []string{"!", "'", "(", ")", "*", "~"} {
		result = strings.ReplaceAll(result, url.QueryEscape(c), c)
	}
	return result
}

// ---------- URL Hashing ----------

const (
	defaultFullHashLimit = 10
	totalPartsLimit      = 40
	partialHashLen       = 6
	pathPartsLimit       = 30
)

func isStripeAuthority(s string) bool {
	if s == "//stripe.com" || s == "//stripe.com." {
		return true
	}
	return strings.HasSuffix(s, ".stripe.com") || strings.HasSuffix(s, ".stripe.com.")
}

func isStripeCheckoutAuthority(s string) bool {
	candidates := []string{
		"//checkout.stripe.com",
		"//qa-checkout.stripe.com",
		"//edge-checkout.stripe.com",
	}
	for _, c := range candidates {
		if s == c {
			return true
		}
	}
	return false
}

// removeUserInfo strips user-info from an authority component.
func removeUserInfo(e string) string {
	if e == "" {
		return e
	}
	t := strings.LastIndex(e, "@")
	if t == -1 {
		return e
	}
	return e[:2] + e[t+1:] // keep "//" prefix
}

// PartitionedUrl parses a URL into RFC-3986 components.
type PartitionedUrl struct {
	Scheme    string
	Authority string
	Path      string
	Query     string
	Fragment  string
}

var urlParseRe = regexp.MustCompile(`^(([^:/?#]+):)?(//([^/?#]*))?([^?#]*)(\?([^#]*))?(#(.*))?`)

// NewPartitionedUrl parses a URL string.
func NewPartitionedUrl(rawURL string) *PartitionedUrl {
	p := &PartitionedUrl{}
	if rawURL == "" {
		return p
	}
	m := urlParseRe.FindStringSubmatch(rawURL)
	if m == nil {
		return p
	}
	p.Scheme = m[1]    // e.g. "https:"
	if m[3] != "" {
		p.Authority = removeUserInfo(m[3]) // e.g. "//chatgpt.com"
	}
	p.Path = m[5]      // e.g. "/"
	p.Query = m[6]     // e.g. "?foo=bar"
	p.Fragment = m[8]  // e.g. "#section"
	return p
}

func (p *PartitionedUrl) String() string {
	var parts []string
	for _, s := range []string{p.Scheme, p.Authority, p.Path, p.Query, p.Fragment} {
		if s != "" {
			parts = append(parts, s)
		}
	}
	return strings.Join(parts, "")
}

// SequentialHashWithLimit manages individual segment hashing with full/partial limits.
type SequentialHashWithLimit struct {
	s              string
	cur            int
	hashedCount    int
	fullHashLimit  int
	totalHashLimit int
}

func newSequentialHashWithLimit(s string, fullHashLimit, totalHashLimit int) *SequentialHashWithLimit {
	return &SequentialHashWithLimit{
		s:              s,
		cur:            0,
		hashedCount:    0,
		fullHashLimit:  fullHashLimit,
		totalHashLimit: totalHashLimit,
	}
}

func (h *SequentialHashWithLimit) shouldHash() bool {
	return h.hashedCount < h.totalHashLimit
}

func (h *SequentialHashWithLimit) isLastHash() bool {
	return h.hashedCount == h.totalHashLimit-1
}

func (h *SequentialHashWithLimit) shouldPartialHash() bool {
	return !h.isLastHash() && h.hashedCount >= h.fullHashLimit
}

func (h *SequentialHashWithLimit) replace(e string) {
	t := e
	n := strings.Index(h.s[h.cur:], e)
	if n == -1 {
		return
	}
	n += h.cur

	if h.isLastHash() {
		t = h.s[n:]
	}

	r := SHA256WithSalt(t)
	if h.shouldPartialHash() {
		if len(r) > partialHashLen {
			r = r[:partialHashLen]
		}
	}

	h.s = h.s[:n] + r + h.s[n+len(t):]
	h.cur = n + len(r)
	h.hashedCount++
}

// SequentialSplitterAndHasher tracks remaining hash budget across URL parts.
type SequentialSplitterAndHasher struct {
	remainingHashes int
	fullHashLimit   int
}

func newSequentialSplitterAndHasher(fullHashLimit int) *SequentialSplitterAndHasher {
	return &SequentialSplitterAndHasher{
		remainingHashes: totalPartsLimit,
		fullHashLimit:   fullHashLimit,
	}
}

func (s *SequentialSplitterAndHasher) getFullHashLimit(part string) int {
	if part == "authority" {
		return totalPartsLimit
	}
	return s.fullHashLimit
}

func (s *SequentialSplitterAndHasher) totalHashLimitForPart(part string) int {
	switch part {
	case "authority":
		return totalPartsLimit
	case "path":
		v := s.remainingHashes
		if v > pathPartsLimit {
			v = pathPartsLimit
		}
		if v < 1 {
			v = 1
		}
		return v
	case "query", "fragment":
		if s.remainingHashes < 1 {
			return 1
		}
		return s.remainingHashes
	default:
		return 0
	}
}

func (s *SequentialSplitterAndHasher) splitAndHash(input, partType string, delim *regexp.Regexp) string {
	if partType == "authority" && input != "" && isStripeCheckoutAuthority(input) {
		return input
	}
	if input == "" {
		return input
	}

	h := newSequentialHashWithLimit(input, s.getFullHashLimit(partType), s.totalHashLimitForPart(partType))

	// Split by delimiter, filter empty strings
	segments := delim.Split(input, -1)
	for _, seg := range segments {
		if seg == "" {
			continue
		}
		if h.shouldHash() {
			h.replace(seg)
		}
	}
	s.remainingHashes -= h.hashedCount
	return h.s
}

// hashURL hashes URL components using SequentialSplitterAndHasher.
func hashURL(p *PartitionedUrl, fullHashLimit int) *PartitionedUrl {
	r := newSequentialSplitterAndHasher(fullHashLimit)

	authorityRe := regexp.MustCompile(`[/.:]`)
	otherRe := regexp.MustCompile(`[/#?!&+,=]`)

	p.Authority = r.splitAndHash(p.Authority, "authority", authorityRe)
	p.Path = r.splitAndHash(p.Path, "path", otherRe)
	p.Query = r.splitAndHash(p.Query, "query", otherRe)
	p.Fragment = r.splitAndHash(p.Fragment, "fragment", otherRe)
	return p
}

// hashURLWithAuthorityCheck applies Stripe-specific authority logic before hashing.
func hashURLWithAuthorityCheck(p *PartitionedUrl, fullHashLimit int) *PartitionedUrl {
	n := p.Authority
	if n != "" && isStripeCheckoutAuthority(n) {
		return hashURL(p, defaultFullHashLimit)
	}
	if n != "" && isStripeAuthority(n) {
		return p // don't hash stripe.com URLs
	}
	return hashURL(p, fullHashLimit)
}

// HashURL is the public entry point for URL hashing.
func HashURL(urlStr string) string {
	p := NewPartitionedUrl(urlStr)
	return hashURLWithAuthorityCheck(p, defaultFullHashLimit).String()
}

// GetHashTimestampWithSalt returns "timestamp:hash" string.
func GetHashTimestampWithSalt(randomValue string) string {
	now := time.Now().UnixMilli()
	hash := SHA256URLSafe(randomValue + fmt.Sprintf("%d", now+1))
	return fmt.Sprintf("%d:%s", now, hash)
}