feedback formatter

pkg/richtext/richtext.go

@@ -0,0 +1,121 @@
+// Package richtext renders the light formatting the admin panel's message
+// fields allow, for the two places that text leaves the app:
+//
+//   - ToHTML, for email bodies.
+//   - Plain, for push notifications and toasts, which can't render markup.
+//
+// The format is a handful of markers typed into a plain textarea:
+//
+//	*bold*   _italic_   __underline__   ~strike~   [label](https://…)
+//
+// plus bare http(s) URLs, which become links on their own. *bold* and ~strike~
+// match what chat already uses, so the two read the same.
+//
+// Safety comes from the order of operations: the input is HTML-escaped FIRST,
+// and only then are the markers turned into tags. Nothing an author types can
+// therefore produce markup of its own — the only tags in the output are the
+// ones this package writes. Links are limited to http and https, so a
+// `javascript:` URL is never made clickable.
+package richtext
+
+import (
+	"fmt"
+	"html"
+	"html/template"
+	"regexp"
+	"strings"
+)
+
+var (
+	// A labelled link, or a bare URL. One alternation so a URL is never
+	// linked twice — once as the label's target and once on its own.
+	linkRe = regexp.MustCompile(`\[([^\]\n]+)\]\((https?://[^\s)]+)\)|(https?://[^\s<]+)`)
+
+	underlineRe = regexp.MustCompile(`__([^_\n]+?)__`)
+	boldRe      = regexp.MustCompile(`\*([^*\n]+?)\*`)
+	strikeRe    = regexp.MustCompile(`~([^~\n]+?)~`)
+	// Italic needs a non-word character (or an edge) on each side, so the
+	// underscores inside snake_case or a handle aren't read as markers. RE2
+	// has no lookbehind, so the neighbours are captured and put back.
+	italicRe = regexp.MustCompile(`(^|[^\w])_([^_\n]+?)_($|[^\w])`)
+
+	// Punctuation that ends a sentence rather than belonging to the URL.
+	trailingPunct = ".,;:!?)"
+)
+
+// placeholder marks where a finished link sits while markers are applied, so
+// the underscores and tildes common in URLs aren't read as formatting.
+func placeholder(i int) string { return fmt.Sprintf("\x00%d\x00", i) }
+
+// splitTrailing separates sentence punctuation from the end of a bare URL.
+func splitTrailing(u string) (url, tail string) {
+	end := len(u)
+	for end > 0 && strings.ContainsRune(trailingPunct, rune(u[end-1])) {
+		end--
+	}
+	return u[:end], u[end:]
+}
+
+// ToHTML renders the markers as HTML for an email body.
+func ToHTML(s string) template.HTML {
+	escaped := html.EscapeString(strings.ReplaceAll(s, "\r\n", "\n"))
+
+	var links []string
+	withHolders := linkRe.ReplaceAllStringFunc(escaped, func(m string) string {
+		parts := linkRe.FindStringSubmatch(m)
+		var label, href, tail string
+		if parts[1] != "" {
+			label, href = parts[1], parts[2]
+		} else {
+			href, tail = splitTrailing(parts[3])
+			label = href
+		}
+		// href is already escaped (it came from the escaped input), so it is
+		// safe inside the attribute as-is.
+		links = append(links, fmt.Sprintf(
+			`<a href="%s" style="color:#9a6b2f;font-weight:600">%s</a>`, href, label))
+		return placeholder(len(links)-1) + tail
+	})
+
+	out := underlineRe.ReplaceAllString(withHolders, `<u>$1</u>`)
+	out = boldRe.ReplaceAllString(out, `<strong>$1</strong>`)
+	out = strikeRe.ReplaceAllString(out, `<s>$1</s>`)
+	out = italicRe.ReplaceAllString(out, `$1<em>$2</em>$3`)
+
+	for i, l := range links {
+		out = strings.Replace(out, placeholder(i), l, 1)
+	}
+	out = strings.ReplaceAll(out, "\n", "<br>")
+
+	// Every tag here was written by this function, and all author text was
+	// escaped before any of them were added.
+	return template.HTML(out) //nolint:gosec // see above
+}
+
+// Plain strips the markers for surfaces that can't render them — a push
+// notification or a toast. A labelled link keeps only its label; a bare URL
+// stays as written, since it is still useful to see.
+func Plain(s string) string {
+	s = strings.ReplaceAll(s, "\r\n", "\n")
+
+	var links []string
+	withHolders := linkRe.ReplaceAllStringFunc(s, func(m string) string {
+		parts := linkRe.FindStringSubmatch(m)
+		if parts[1] != "" {
+			links = append(links, parts[1])
+		} else {
+			links = append(links, parts[3])
+		}
+		return placeholder(len(links) - 1)
+	})
+
+	out := underlineRe.ReplaceAllString(withHolders, `$1`)
+	out = boldRe.ReplaceAllString(out, `$1`)
+	out = strikeRe.ReplaceAllString(out, `$1`)
+	out = italicRe.ReplaceAllString(out, `$1$2$3`)
+
+	for i, l := range links {
+		out = strings.Replace(out, placeholder(i), l, 1)
+	}
+	return out
+}