internal/web: pre-expand tabs in markdown
Goldmark has a bug that causes crashes in inputs
containing mixed spaces and tabs. Avoid it by removing tabs.
Change-Id: I4c678fce6c68bf0d448ed0b75a2ac12a42891ec6
Reviewed-on: https://go-review.googlesource.com/c/website/+/331349
Trust: Russ Cox <rsc@golang.org>
Run-TryBot: Russ Cox <rsc@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Steve Traut <straut@google.com>
diff --git a/internal/web/markdown.go b/internal/web/markdown.go
index 1973461..2cee319 100644
--- a/internal/web/markdown.go
+++ b/internal/web/markdown.go
@@ -6,6 +6,7 @@
import (
"bytes"
+ "unicode/utf8"
"github.com/yuin/goldmark"
"github.com/yuin/goldmark/parser"
@@ -18,6 +19,7 @@
// The Markdown source may contain raw HTML,
// but Go templates have already been processed.
func renderMarkdown(src []byte) ([]byte, error) {
+ src = replaceTabs(src)
// parser.WithHeadingAttribute allows custom ids on headings.
// html.WithUnsafe allows use of raw HTML, which we need for tables.
md := goldmark.New(
@@ -29,3 +31,42 @@
}
return buf.Bytes(), nil
}
+
+// replaceTabs replaces all tabs in text with spaces up to a 4-space tab stop.
+//
+// In Markdown, tabs used for indentation are required to be interpreted as
+// 4-space tab stops. See https://spec.commonmark.org/0.30/#tabs.
+// Go also renders nicely and more compactly on the screen with 4-space
+// tab stops, while browsers often use 8-space.
+// And Goldmark crashes in some inputs that mix spaces and tabs.
+// Fix the crashes and make the Go code consistently compact across browsers,
+// all while staying Markdown-compatible, by expanding to 4-space tab stops.
+//
+// This function does not handle multi-codepoint Unicode sequences correctly.
+func replaceTabs(text []byte) []byte {
+ var buf bytes.Buffer
+ col := 0
+ for len(text) > 0 {
+ r, size := utf8.DecodeRune(text)
+ text = text[size:]
+
+ switch r {
+ case '\n':
+ buf.WriteByte('\n')
+ col = 0
+
+ case '\t':
+ buf.WriteByte(' ')
+ col++
+ for col%4 != 0 {
+ buf.WriteByte(' ')
+ col++
+ }
+
+ default:
+ buf.WriteRune(r)
+ col++
+ }
+ }
+ return buf.Bytes()
+}
