perf: pre-compile regex patterns in markdownToTelegramHTML

[chillum-codeX]

perf: pre-compile regex patterns in markdownToTelegramHTML

Problem

The markdownToTelegramHTML(), extractCodeBlocks(), and extractInlineCodes() functions in pkg/channels/telegram.go compile 9 regex patterns from scratch on every call using regexp.MustCompile() inside the function body.

Each regexp.MustCompile() allocates ~2-4 KB on the heap for the compiled automaton. For a single outbound message, this creates ~20 KB of unnecessary heap allocation that immediately becomes garbage, increasing GC pressure.

Solution

Move all 9 regexp.MustCompile() calls to package-level var declarations. The compiled patterns are created once at program start and reused across all calls. This is safe because Go's regexp.Regexp is goroutine-safe.

Before

func markdownToTelegramHTML(text string) string {
    text = regexp.MustCompile(`^#{1,6}\s+(.+)$`).ReplaceAllString(text, "$1")
    text = regexp.MustCompile(`^>\s*(.*)$`).ReplaceAllString(text, "$1")
    // ... 5 more inline compilations
}

func extractCodeBlocks(text string) codeBlockMatch {
    re := regexp.MustCompile("```[\\w]*\\n?([\\s\\S]*?)```")
    // ...
}

After

var (
    reHeader     = regexp.MustCompile(`^#{1,6}\s+(.+)$`)
    reBlockquote = regexp.MustCompile(`^>\s*(.*)$`)
    reLink       = regexp.MustCompile(`\[([^\]]+)\]\(([^)]+)\)`)
    reBold       = regexp.MustCompile(`\*\*(.+?)\*\*`)
    reBoldAlt    = regexp.MustCompile(`__(.+?)__`)
    reItalic     = regexp.MustCompile(`_([^_]+)_`)
    reStrike     = regexp.MustCompile(`~~(.+?)~~`)
    reListItem   = regexp.MustCompile(`^[-*]\s+`)
    reCodeBlock  = regexp.MustCompile("```[\\w]*\\n?([\\s\\S]*?)```")
    reInlineCode = regexp.MustCompile("`([^`]+)`")
)

func markdownToTelegramHTML(text string) string {
    text = reHeader.ReplaceAllString(text, "$1")
    text = reBlockquote.ReplaceAllString(text, "$1")
    // ...
}

Impact

• Eliminates ~20 KB heap allocation per outbound message
• Reduces GC pressure under high message throughput
• Zero behavioral change — all regex patterns are identical
• Measured context: Independent benchmarking (100-1000 msg bursts) showed the telegram-slim build achieves 1-3 MB RSS on Linux. This optimization reduces the per-message GC cost, further tightening the memory footprint.

Testing

• Build succeeds: go build -tags "telegram pprof smallbuf" ./cmd/picoclaw
• No functional change — identical regex patterns, same replacement logic
• regexp.Regexp is documented as goroutine-safe

References

• Go regexp goroutine safety: https://pkg.go.dev/regexp#Regexp
• Memory analysis: https://github.com/chillum-codeX/picoclaw-under-the-hood

[alexhoshina]

Hey! We are currently refactoring the channel system, and we have opened a refactor branch. It might be better if you could target your pull request to the refactor branch.