From e8e00d1884af9d36a2335cb44f4248d309a819b2 Mon Sep 17 00:00:00 2001
From: viettranx <viettranx@gmail.com>
Date: Sat, 11 Apr 2026 15:40:39 +0700
Subject: [PATCH] fix(telegram): preserve italic content in htmlTagToMarkdown

Go regexp's $1_ is parsed as a reference to a named group "1_" (underscores are
valid identifier chars), not $1 followed by literal underscore. This dropped the
captured italic content, turning <i>foo</i> into "__" instead of "_foo_".

Use ${1}_ with explicit braces to delimit the group reference, and update the
format_extended_test suite to assert the full round-trip for <i> and <em>.
---
 internal/channels/telegram/format.go          |  6 +++--
 .../channels/telegram/format_extended_test.go | 22 ++++++++++---------
 2 files changed, 16 insertions(+), 12 deletions(-)

diff --git a/internal/channels/telegram/format.go b/internal/channels/telegram/format.go
index e81bf411..5ab481b2 100644
--- a/internal/channels/telegram/format.go
+++ b/internal/channels/telegram/format.go
@@ -21,8 +21,10 @@ var htmlToMdReplacers = []struct {
 	{regexp.MustCompile(`(?i)</?p\s*>`), "\n"},
 	{regexp.MustCompile(`(?i)<b>([\s\S]*?)</b>`), "**$1**"},
 	{regexp.MustCompile(`(?i)<strong>([\s\S]*?)</strong>`), "**$1**"},
-	{regexp.MustCompile(`(?i)<i>([\s\S]*?)</i>`), "_$1_"},
-	{regexp.MustCompile(`(?i)<em>([\s\S]*?)</em>`), "_$1_"},
+	// ${1}_ (not $1_) — Go regexp treats $1_ as a named group reference (identifier
+	// characters include `_`), which drops the capture. Curly braces delimit the group.
+	{regexp.MustCompile(`(?i)<i>([\s\S]*?)</i>`), "_${1}_"},
+	{regexp.MustCompile(`(?i)<em>([\s\S]*?)</em>`), "_${1}_"},
 	{regexp.MustCompile(`(?i)<s>([\s\S]*?)</s>`), "~~$1~~"},
 	{regexp.MustCompile(`(?i)<strike>([\s\S]*?)</strike>`), "~~$1~~"},
 	{regexp.MustCompile(`(?i)<del>([\s\S]*?)</del>`), "~~$1~~"},
diff --git a/internal/channels/telegram/format_extended_test.go b/internal/channels/telegram/format_extended_test.go
index e5eb09e6..8b7cd4fa 100644
--- a/internal/channels/telegram/format_extended_test.go
+++ b/internal/channels/telegram/format_extended_test.go
@@ -161,12 +161,17 @@ func TestMarkdownToTelegramHTML_HTMLTagsConvertedFirst(t *testing.T) {
 			t.Errorf("<b>bold</b> should remain bold in output, got: %q", got)
 		}
 	})
-	t.Run("html italic input produces output", func(t *testing.T) {
-		// <i>italic</i> goes through htmlTagToMarkdown (→ _italic_ with known $1_ caveat)
-		// then through markdownToTelegramHTML — the result is non-empty, no panics.
+	t.Run("html italic tag round-trips", func(t *testing.T) {
+		// <i>italic</i> → _italic_ → <i>italic</i> through the full pipeline.
 		got := markdownToTelegramHTML("<i>italic</i>")
-		if got == "" {
-			t.Error("markdownToTelegramHTML should return non-empty for <i>italic</i>")
+		if !strings.Contains(got, "<i>italic</i>") {
+			t.Errorf("<i>italic</i> should remain italic in output, got: %q", got)
+		}
+	})
+	t.Run("html em tag round-trips", func(t *testing.T) {
+		got := markdownToTelegramHTML("<em>emphasis</em>")
+		if !strings.Contains(got, "<i>emphasis</i>") {
+			t.Errorf("<em>emphasis</em> should become <i>emphasis</i>, got: %q", got)
 		}
 	})
 	t.Run("html strike tag round-trips", func(t *testing.T) {
@@ -419,11 +424,8 @@ func TestHTMLTagToMarkdown(t *testing.T) {
 		{"<br/>", []string{"\n"}},
 		{"<b>bold</b>", []string{"**bold**"}},
 		{"<strong>bold</strong>", []string{"**bold**"}},
-		// NOTE: <i>/<em> use "_$1_" replacement where Go regexp treats $1_ as
-		// a named group reference (not group 1 + literal _), so content is lost.
-		// This is documented behavior of the current implementation.
-		// The pipeline still works end-to-end because <i> only appears in LLM
-		// output which gets processed through the full markdownToTelegramHTML chain.
+		{"<i>italic</i>", []string{"_italic_"}},
+		{"<em>italic</em>", []string{"_italic_"}},
 		{"<s>struck</s>", []string{"~~struck~~"}},
 		{"<del>deleted</del>", []string{"~~deleted~~"}},
 		{"<code>var</code>", []string{"`var`"}},