Split long replies between Markdown blocks

splitRich cuts the Markdown of a reply into parts of at most 32768
bytes, the limit of the Markdown of one rich message. It cuts after a
blank line outside a code block, so a table, a code block, and a
paragraph stay whole, and each part renders as the reply would. When no
such cut fits, it cuts after a line break outside a code block, and then
inside a code block, which it closes at the end of the part and opens
again with the same fence line in the next part. A single line over the
limit is cut at the limit, never inside a character.

The Bot API allows 32768 characters of rendered text and does not say if
a character is a byte. The bytes of the Markdown are at least the
characters of the text, so a part is never over the limit. The bot does
not use splitRich yet.

Issue: fec23f9

diff --git a/markdown.go b/markdown.go
index 48e9ace0e75970dd8be5f18331edd4b44babb586..1a2f4feb1343723a704a874c2c586e78a0fee8d6 100644 (file)
--- a/markdown.go
+++ b/markdown.go
@@ -169,8 +169,7 @@ func closingFence(lines []string, index int, fence string) int {
        }
 
        for end := index + 1; end < len(lines); end++ {
        }
 
        for end := index + 1; end < len(lines); end++ {

-               trimmed := strings.TrimSpace(lines[end])
-               if strings.HasPrefix(trimmed, fence) && strings.Trim(trimmed, "`") == "" {
+               if closesFence(lines[end], fence) {

                        return end
                }
        }
                        return end
                }
        }


@@ -178,6 +177,12 @@ func closingFence(lines []string, index int, fence string) int {
        return -1
 }
 
        return -1
 }
 

+// closesFence tells if line closes the code block of fence.
+func closesFence(line, fence string) bool {
+       trimmed := strings.TrimSpace(line)
+       return strings.HasPrefix(trimmed, fence) && strings.Trim(trimmed, "`") == ""
+}
+

 // headingText returns the text of an ATX heading line without its marks, or ""
 // when line is not a heading or has no text.
 func headingText(line string) string {
 // headingText returns the text of an ATX heading line without its marks, or ""
 // when line is not a heading or has no text.
 func headingText(line string) string {

diff --git a/telegram.go b/telegram.go
index 9a24a4e861d7436bcd0a109480b50fbbb538377e..9dbe86e1cf24e6b629245bd71414aa9d2334bf58 100644 (file)
--- a/telegram.go
+++ b/telegram.go
@@ -14,6 +14,7 @@ import (
        "strings"
        "sync"
        "time"
        "strings"
        "sync"
        "time"

+       "unicode/utf8"

 )
 
 const (
 )
 
 const (


@@ -33,6 +34,12 @@ const (
        // messageUnits is the most UTF-16 code units of one Telegram message.
        messageUnits = 4096
 
        // messageUnits is the most UTF-16 code units of one Telegram message.
        messageUnits = 4096
 

+       // richMessageBytes is the most bytes of the Markdown of one rich
+       // message. The Bot API allows 32768 characters of rendered text, and
+       // the Markdown of a text has at least as many bytes as the text has
+       // characters.
+       richMessageBytes = 32768
+

        // typingInterval is how often the bot shows the typing action while a
        // turn runs. Telegram shows the action for about five seconds.
        typingInterval = 4 * time.Second
        // typingInterval is how often the bot shows the typing action while a
        // turn runs. Telegram shows the action for about five seconds.
        typingInterval = 4 * time.Second


@@ -548,3 +555,83 @@ func prefixEnd(text string, limit int) int {
 
        return len(text)
 }
 
        return len(text)
 }

+
+// splitRich cuts the Markdown of a reply into parts of at most limit bytes. It
+// cuts after a blank line outside a code block, so a table, a code block, or a
+// paragraph stays whole. When no such cut fits in a part, it cuts after the
+// last line break outside a code block, then after the last line break in a
+// code block, which it closes at the end of the part and opens again in the
+// next part with the same fence line. A line longer than limit is cut at the
+// limit. It never cuts a character.
+func splitRich(markdown string, limit int) []string {
+       var parts []string
+       for len(markdown) > limit {
+               cut, fence, opening := richCut(markdown, limit)
+               if fence == "" {
+                       parts = append(parts, markdown[:cut])
+                       markdown = markdown[cut:]
+
+                       continue
+               }
+
+               parts = append(parts, markdown[:cut]+fence+"\n")
+               markdown = opening + "\n" + markdown[cut:]
+       }
+
+       return append(parts, markdown)
+}
+
+// richCut returns where splitRich cuts markdown, which is longer than limit.
+// For a cut in a code block, it also returns the fence that closes the block
+// and the line that opens it. Each part before a cut holds text other than
+// space, and a cut in a code block leaves room for the fence, so each cut makes
+// the rest shorter.
+func richCut(markdown string, limit int) (int, string, string) {
+       blank, line, code := 0, 0, 0
+       var fence, opening, inFence, inOpening string
+       for start := 0; start < limit; {
+               end := strings.IndexByte(markdown[start:], '\n')
+               if end < 0 || start+end+1 > limit {
+                       break
+               }
+
+               end += start + 1
+               text := strings.TrimRight(markdown[start:end], "\r\n")
+               _, opens := codeFence(text)
+               switch {
+               case fence != "":
+                       if closesFence(text, fence) {
+                               fence = ""
+                       } else if end > len(opening)+1 && end+len(fence)+1 <= limit {
+                               code, inFence, inOpening = end, fence, opening
+                       }
+               case opens != "":
+                       fence, opening = opens, text
+               case strings.TrimSpace(markdown[:end]) == "":
+               case strings.TrimSpace(text) == "":
+                       blank = end
+               }
+
+               if fence == "" && strings.TrimSpace(markdown[:end]) != "" {
+                       line = end
+               }
+
+               start = end
+       }
+
+       switch {
+       case blank > 0:
+               return blank, "", ""
+       case line > 0:
+               return line, "", ""
+       case code > 0:
+               return code, inFence, inOpening
+       }
+
+       cut := limit
+       for !utf8.RuneStart(markdown[cut]) {
+               cut--
+       }
+
+       return cut, "", ""
+}

diff --git a/telegram_test.go b/telegram_test.go
index 2578b77b1c97b98561f3718e56a6c041ec2cc6a7..30b2fa1bb686f730963c2426bdb88e3761949d5b 100644 (file)
--- a/telegram_test.go
+++ b/telegram_test.go
@@ -409,6 +409,64 @@ func TestSplitEntitiesCutsAnEntityAcrossASplit(t *testing.T) {
        }, parts)
 }
 
        }, parts)
 }
 

+func TestSplitRich(t *testing.T) {
+       testCases := []struct {
+               name     string
+               markdown string
+               limit    int
+               parts    []string
+       }{
+               {
+                       name:     "short reply stays whole",
+                       markdown: "| a | b |",
+                       limit:    100,
+                       parts:    []string{"| a | b |"},
+               },
+               {
+                       name:     "cut after a blank line keeps a table whole",
+                       markdown: "Intro.\n\n| a | b |\n|---|---|\n| 1 | 2 |\n",
+                       limit:    30,
+                       parts:    []string{"Intro.\n\n", "| a | b |\n|---|---|\n| 1 | 2 |\n"},
+               },
+               {
+                       name:     "blank line in a code block is not a cut",
+                       markdown: "Text.\n\n```go\nx\n\ny\n```\n\nEnd.\n",
+                       limit:    20,
+                       parts:    []string{"Text.\n\n", "```go\nx\n\ny\n```\n\n", "End.\n"},
+               },
+               {
+                       name:     "paragraph with no blank line is cut after a line",
+                       markdown: "aaa\nbbb\nccc\n",
+                       limit:    9,
+                       parts:    []string{"aaa\nbbb\n", "ccc\n"},
+               },
+               {
+                       name:     "long code block is closed and opened again",
+                       markdown: "```sh\none\ntwo\nthree\n```\n",
+                       limit:    16,
+                       parts:    []string{"```sh\none\n```\n", "```sh\ntwo\n```\n", "```sh\nthree\n```\n"},
+               },
+               {
+                       name:     "long line is cut at the limit and not inside a character",
+                       markdown: "ab\u00e9cd",
+                       limit:    3,
+                       parts:    []string{"ab", "\u00e9c", "d"},
+               },
+       }
+
+       for _, tc := range testCases {
+               t.Run(tc.name, func(t *testing.T) {
+                       require.Equal(t, tc.parts, splitRich(tc.markdown, tc.limit))
+               })
+       }
+
+       long := strings.Repeat("A paragraph of the reply.\n\n", 2000)
+       parts := splitRich(long, richMessageBytes)
+       require.Len(t, parts, 2)
+       require.Len(t, parts[0], richMessageBytes-richMessageBytes%len("A paragraph of the reply.\n\n"))
+       require.Equal(t, long, strings.Join(parts, ""))
+}
+

 func TestLoadBotConfig(t *testing.T) {
        testCases := []struct {
                name      string
 func TestLoadBotConfig(t *testing.T) {
        testCases := []struct {
                name      string