telegram: improve markdown formatting and table rendering

This commit is contained in:
dj-oyu 2026-02-20 00:54:55 +09:00
parent 82aae32fe7
commit 9c4b49d818
2 changed files with 352 additions and 5 deletions

View file

@ -10,6 +10,7 @@ import (
"strings"
"sync"
"time"
"unicode"
th "github.com/mymmrac/telego/telegohandler"
@ -46,6 +47,8 @@ func (c *thinkingCancel) Cancel() {
}
const telegramMaxMessageChars = 3900
const markdownTableMaxWidth = 90
const markdownTableMinColWidth = 6
var thinkBlockPattern = regexp.MustCompile(`(?is)<think>.*?</think>`)
@ -452,12 +455,17 @@ func markdownToTelegramHTML(text string) string {
codeBlocks := extractCodeBlocks(text)
text = codeBlocks.text
tables := extractMarkdownTables(text)
text = tables.text
inlineCodes := extractInlineCodes(text)
text = inlineCodes.text
text = regexp.MustCompile(`^#{1,6}\s+(.+)$`).ReplaceAllString(text, "$1")
// Convert headings to bold markdown first, then transform to HTML later.
text = regexp.MustCompile(`(?m)^#{1,6}\s+(.+)$`).ReplaceAllString(text, "**$1**")
text = regexp.MustCompile(`^>\s*(.*)$`).ReplaceAllString(text, "$1")
// Strip quote markers per line.
text = regexp.MustCompile(`(?m)^>\s*(.*)$`).ReplaceAllString(text, "$1")
text = escapeHTML(text)
@ -478,13 +486,16 @@ func markdownToTelegramHTML(text string) string {
text = regexp.MustCompile(`~~(.+?)~~`).ReplaceAllString(text, "<s>$1</s>")
text = regexp.MustCompile(`^[-*]\s+`).ReplaceAllString(text, "• ")
for i, code := range inlineCodes.codes {
escaped := escapeHTML(code)
text = strings.ReplaceAll(text, fmt.Sprintf("\x00IC%d\x00", i), fmt.Sprintf("<code>%s</code>", escaped))
}
for i, table := range tables.tables {
escaped := escapeHTML(table)
text = strings.ReplaceAll(text, fmt.Sprintf("\x00TB%d\x00", i), fmt.Sprintf("<pre><code>%s</code></pre>", escaped))
}
for i, code := range codeBlocks.codes {
escaped := escapeHTML(code)
text = strings.ReplaceAll(text, fmt.Sprintf("\x00CB%d\x00", i), fmt.Sprintf("<pre><code>%s</code></pre>", escaped))
@ -498,6 +509,11 @@ type codeBlockMatch struct {
codes []string
}
type tableBlockMatch struct {
text string
tables []string
}
func extractCodeBlocks(text string) codeBlockMatch {
re := regexp.MustCompile("```[\\w]*\\n?([\\s\\S]*?)```")
matches := re.FindAllStringSubmatch(text, -1)
@ -541,6 +557,278 @@ func extractInlineCodes(text string) inlineCodeMatch {
return inlineCodeMatch{text: text, codes: codes}
}
func extractMarkdownTables(text string) tableBlockMatch {
lines := strings.Split(text, "\n")
out := make([]string, 0, len(lines))
tables := make([]string, 0)
placeholderIdx := 0
for i := 0; i < len(lines); {
if i+1 < len(lines) && isTableRowLine(lines[i]) && isTableSeparatorLine(lines[i+1]) {
start := i
i += 2
for i < len(lines) && isTableRowLine(lines[i]) {
i++
}
block := lines[start:i]
formatted := formatMarkdownTable(block)
placeholder := fmt.Sprintf("\x00TB%d\x00", placeholderIdx)
placeholderIdx++
tables = append(tables, formatted)
out = append(out, placeholder)
continue
}
out = append(out, lines[i])
i++
}
return tableBlockMatch{
text: strings.Join(out, "\n"),
tables: tables,
}
}
func isTableRowLine(line string) bool {
trimmed := strings.TrimSpace(line)
return strings.Count(trimmed, "|") >= 2
}
func isTableSeparatorLine(line string) bool {
trimmed := strings.TrimSpace(line)
if trimmed == "" {
return false
}
trimmed = strings.Trim(trimmed, "|")
parts := strings.Split(trimmed, "|")
if len(parts) == 0 {
return false
}
for _, p := range parts {
c := strings.TrimSpace(p)
if c == "" {
return false
}
for _, r := range c {
if r != '-' && r != ':' && r != ' ' {
return false
}
}
if !strings.Contains(c, "-") {
return false
}
}
return true
}
func parseTableCells(line string) []string {
trimmed := strings.TrimSpace(line)
trimmed = strings.Trim(trimmed, "|")
raw := strings.Split(trimmed, "|")
cells := make([]string, 0, len(raw))
for _, c := range raw {
cells = append(cells, strings.TrimSpace(c))
}
return cells
}
func formatMarkdownTable(lines []string) string {
if len(lines) < 2 {
return strings.Join(lines, "\n")
}
rows := make([][]string, 0, len(lines)-1)
rows = append(rows, parseTableCells(lines[0])) // header
for _, l := range lines[2:] { // skip separator
rows = append(rows, parseTableCells(l))
}
cols := 0
for _, r := range rows {
if len(r) > cols {
cols = len(r)
}
}
if cols == 0 {
return strings.Join(lines, "\n")
}
widths := make([]int, cols)
for _, r := range rows {
for i := 0; i < cols; i++ {
cell := ""
if i < len(r) {
cell = r[i]
}
w := displayWidth(cell)
if w > widths[i] {
widths[i] = w
}
}
}
// Enforce horizontal width limit by shrinking widest columns first.
totalWidth := func(ws []int) int {
sum := 0
for _, w := range ws {
sum += w
}
if len(ws) == 0 {
return 0
}
// "| " + " | ".join(cols) + " |"
return sum + 4 + (len(ws)-1)*3
}
for totalWidth(widths) > markdownTableMaxWidth {
maxIdx := -1
maxW := 0
for i, w := range widths {
if w > maxW && w > markdownTableMinColWidth {
maxW = w
maxIdx = i
}
}
if maxIdx == -1 {
break
}
widths[maxIdx]--
}
padToWidth := func(s string, w int) string {
dw := displayWidth(s)
if dw >= w {
return s
}
return s + strings.Repeat(" ", w-dw)
}
renderLogicalRow := func(r []string) []string {
wrapped := make([][]string, cols)
maxLines := 1
for c := 0; c < cols; c++ {
cell := ""
if c < len(r) {
cell = r[c]
}
lines := wrapByDisplayWidth(cell, widths[c])
if len(lines) == 0 {
lines = []string{""}
}
wrapped[c] = lines
if len(lines) > maxLines {
maxLines = len(lines)
}
}
out := make([]string, 0, maxLines)
for lineIdx := 0; lineIdx < maxLines; lineIdx++ {
var rowBuilder strings.Builder
rowBuilder.WriteString("| ")
for c := 0; c < cols; c++ {
cellLine := ""
if lineIdx < len(wrapped[c]) {
cellLine = wrapped[c][lineIdx]
}
rowBuilder.WriteString(padToWidth(cellLine, widths[c]))
if c == cols-1 {
rowBuilder.WriteString(" |")
} else {
rowBuilder.WriteString(" | ")
}
}
out = append(out, rowBuilder.String())
}
return out
}
var b strings.Builder
for rIdx, r := range rows {
rowLines := renderLogicalRow(r)
for _, line := range rowLines {
b.WriteString(line)
b.WriteString("\n")
}
if rIdx == 0 {
b.WriteString("| ")
for c := 0; c < cols; c++ {
b.WriteString(strings.Repeat("-", widths[c]))
if c == cols-1 {
b.WriteString(" |\n")
} else {
b.WriteString(" | ")
}
}
}
}
return strings.TrimRight(b.String(), "\n")
}
func displayWidth(s string) int {
w := 0
for _, r := range s {
switch {
case r == '\u200d' || r == '\u200c' || r == '\ufe0f':
continue
case unicode.Is(unicode.Mn, r):
continue
case unicode.In(r,
unicode.Han,
unicode.Hiragana,
unicode.Katakana,
unicode.Hangul):
w += 2
case (r >= 0x3000 && r <= 0x303F) || (r >= 0xFF00 && r <= 0xFFEF):
// CJK symbols/punctuation and half/fullwidth forms.
w += 2
default:
w++
}
}
return w
}
func wrapByDisplayWidth(s string, maxWidth int) []string {
if maxWidth <= 0 {
return []string{s}
}
if s == "" {
return []string{""}
}
lines := make([]string, 0, 1)
var cur strings.Builder
curWidth := 0
flush := func() {
lines = append(lines, strings.TrimRight(cur.String(), " "))
cur.Reset()
curWidth = 0
}
for _, r := range s {
if r == '\n' {
flush()
continue
}
rw := displayWidth(string(r))
if rw == 0 {
cur.WriteRune(r)
continue
}
if curWidth+rw > maxWidth && cur.Len() > 0 {
flush()
}
cur.WriteRune(r)
curWidth += rw
}
if cur.Len() > 0 || len(lines) == 0 {
flush()
}
return lines
}
func escapeHTML(text string) string {
text = strings.ReplaceAll(text, "&", "&amp;")
text = strings.ReplaceAll(text, "<", "&lt;")

View file

@ -1,6 +1,9 @@
package channels
import "testing"
import (
"strings"
"testing"
)
func TestSanitizeTelegramOutgoingContent_RemovesThinkBlock(t *testing.T) {
in := "<think>\nsecret reasoning\n</think>\n\nユーザー向け本文"
@ -20,3 +23,59 @@ func TestSanitizeTelegramOutgoingContent_EmptyAfterThink(t *testing.T) {
}
}
func TestMarkdownToTelegramHTML_Heading(t *testing.T) {
in := "### メッセージ\n- A\n- B"
got := markdownToTelegramHTML(in)
if !strings.Contains(got, "<b>メッセージ</b>") {
t.Fatalf("expected heading to be bold, got: %q", got)
}
}
func TestMarkdownToTelegramHTML_Table(t *testing.T) {
in := "| col1 | col2 |\n| --- | --- |\n| a | b |"
got := markdownToTelegramHTML(in)
if !strings.Contains(got, "<pre><code>") {
t.Fatalf("expected table to render as code block, got: %q", got)
}
if !strings.Contains(got, "col1") || !strings.Contains(got, "a") {
t.Fatalf("expected table content to remain, got: %q", got)
}
}
func TestMarkdownToTelegramHTML_TableCJKAlignment(t *testing.T) {
in := "| 項目 | value |\n| --- | --- |\n| 温度 | 23.5 |\n| 状態 | 正常 |"
got := markdownToTelegramHTML(in)
mustContain := []string{
"<pre><code>",
"| 項目",
"| value",
"| ----",
"| 温度",
"| 状態",
}
for _, s := range mustContain {
if !strings.Contains(got, s) {
t.Fatalf("expected output to contain %q, got: %q", s, got)
}
}
}
func TestMarkdownToTelegramHTML_TableWrapsLongCell(t *testing.T) {
in := "| col1 | col2 |\n| --- | --- |\n| short | this is a very very very very long cell that should wrap |"
got := markdownToTelegramHTML(in)
if !strings.Contains(got, "<pre><code>") {
t.Fatalf("expected table to render as code block, got: %q", got)
}
if !strings.Contains(got, "| col1") || !strings.Contains(got, "| col2") {
t.Fatalf("expected header row, got: %q", got)
}
if !strings.Contains(got, "| short") {
t.Fatalf("expected data row with short cell, got: %q", got)
}
// Wrapped content should include a prefix from the long sentence.
if !strings.Contains(got, "this is a very") {
t.Fatalf("expected wrapped long cell content, got: %q", got)
}
}