package main

import (
	"strings"
	"unicode"
	"unicode/utf8"
)

// markdownPunctuation holds the ASCII characters that a backslash makes plain.
const markdownPunctuation = "!\"#$%&'()*+,-./:;<=>?@[\\]^_`{|}~"

// MarkdownSpan is an inline mark of a line: the mark kind, the content inside
// the marks, the index after the closing mark, and the URL of a link.
type MarkdownSpan struct {
	kind    string
	content string
	end     int
	url     string
}

// markdownCodeFence returns the language of a line that opens a fenced code
// block, and the fence. The fence is "" when line does not open a block.
func markdownCodeFence(line string) (string, string) {
	trimmed := strings.TrimLeft(line, " ")
	if len(line)-len(trimmed) > 3 || !strings.HasPrefix(trimmed, "```") {
		return "", ""
	}

	run := len(trimmed) - len(strings.TrimLeft(trimmed, "`"))
	info := strings.TrimSpace(trimmed[run:])
	if strings.Contains(info, "`") {
		return "", ""
	}

	language, _, _ := strings.Cut(info, " ")

	return language, trimmed[:run]
}

// markdownClosingFence returns the index of the line after index that closes
// the code block of fence, or -1 when fence is "" or no line closes it.
func markdownClosingFence(lines []string, index int, fence string) int {
	if fence == "" {
		return -1
	}

	for end := index + 1; end < len(lines); end++ {
		if markdownClosesFence(lines[end], fence) {
			return end
		}
	}

	return -1
}

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

// markdownHeadingText returns the text of an ATX heading line without its
// marks, or "" when line is not a heading or has no text.
func markdownHeadingText(line string) string {
	trimmed := strings.TrimLeft(line, " ")
	if len(line)-len(trimmed) > 3 {
		return ""
	}

	level := len(trimmed) - len(strings.TrimLeft(trimmed, "#"))
	if level == 0 || level > 6 {
		return ""
	}

	rest := trimmed[level:]
	if rest != "" && rest[0] != ' ' && rest[0] != '\t' {
		return ""
	}

	rest = strings.TrimSpace(rest)
	closed := strings.TrimRight(rest, "#")
	if closed == "" || strings.HasSuffix(closed, " ") {
		rest = strings.TrimSpace(closed)
	}

	return rest
}

// markdownMarkSpan returns the inline mark that starts at index of line and
// closes in line. It returns false when no mark starts there or the mark does
// not close.
func markdownMarkSpan(line string, index int) (MarkdownSpan, bool) {
	switch line[index] {
	case '`':
		return markdownCodeSpan(line, index)
	case '[':
		return markdownLinkSpan(line, index)
	case '~':
		return markdownDelimited(line, index, "~~", "strikethrough")
	case '*', '_':
		double := line[index : index+1]
		double += double
		span, found := markdownDelimited(line, index, double, "bold")
		if found {
			return span, true
		}

		return markdownDelimited(line, index, line[index:index+1], "italic")
	default:
		return MarkdownSpan{}, false
	}
}

// markdownCodeSpan returns the inline code that starts at index: a run of
// backticks, the code, and a run of the same length. One space on each side of
// the code is left out, as in Markdown.
func markdownCodeSpan(line string, index int) (MarkdownSpan, bool) {
	run := markdownBackticks(line[index:])
	for search := index + run; search < len(line); {
		next := strings.IndexByte(line[search:], '`')
		if next < 0 {
			break
		}

		next += search
		length := markdownBackticks(line[next:])
		if length != run {
			search = next + length
			continue
		}

		content := line[index+run : next]
		if strings.TrimSpace(content) == "" {
			break
		}

		if len(content) > 1 && content[0] == ' ' && content[len(content)-1] == ' ' {
			content = content[1 : len(content)-1]
		}

		return MarkdownSpan{
			kind:    "code",
			content: content,
			end:     next + run,
		}, true
	}

	return MarkdownSpan{}, false
}

// markdownLinkSpan returns the link [text](url) that starts at index. The text
// can hold brackets in pairs, and the URL holds no space.
func markdownLinkSpan(line string, index int) (MarkdownSpan, bool) {
	depth := 0
	for closing := index; closing < len(line); closing++ {
		switch line[closing] {
		case '\\':
			closing++
		case '[':
			depth++
		case ']':
			depth--
		}

		if depth > 0 {
			continue
		}

		rest := line[closing+1:]
		if closing == index+1 || !strings.HasPrefix(rest, "(") {
			return MarkdownSpan{}, false
		}

		url, _, found := strings.Cut(rest[1:], ")")
		if !found || url == "" || strings.ContainsAny(url, " \t") {
			return MarkdownSpan{}, false
		}

		return MarkdownSpan{
			kind:    "text_link",
			content: line[index+1 : closing],
			end:     closing + 2 + len(url) + 1,
			url:     url,
		}, true
	}

	return MarkdownSpan{}, false
}

// markdownDelimited returns the span of kind between the mark at index and the
// same mark later in line. An opening mark is followed by a character that is
// not a space, and a closing mark follows one. A single mark is not a part of a
// double one. A mark of underscores does not open or close inside a word, so
// snake_case stays as text. A code span inside the content is skipped whole.
func markdownDelimited(line string, index int, mark, kind string) (MarkdownSpan, bool) {
	char := mark[0]
	after := index + len(mark)
	if !strings.HasPrefix(line[index:], mark) || after >= len(line) || markdownIsSpace(line[after]) {
		return MarkdownSpan{}, false
	}

	if len(mark) == 1 && (line[after] == char || index > 0 && line[index-1] == char) {
		return MarkdownSpan{}, false
	}

	if char == '_' && index > 0 && markdownWordCharBefore(line[:index]) {
		return MarkdownSpan{}, false
	}

	for closing := after + 1; closing < len(line); closing++ {
		switch {
		case line[closing] == '\\':
			closing++
			continue
		case line[closing] == '`':
			span, found := markdownCodeSpan(line, closing)
			if found {
				closing = span.end - 1
			}

			continue
		case !strings.HasPrefix(line[closing:], mark) || markdownIsSpace(line[closing-1]):
			continue
		}

		end := closing + len(mark)
		if len(mark) == 1 && (line[closing-1] == char || end < len(line) && line[end] == char) {
			continue
		}

		if char == '_' && markdownWordChar(line[end:]) {
			continue
		}

		return MarkdownSpan{
			kind:    kind,
			content: line[after:closing],
			end:     end,
		}, true
	}

	return MarkdownSpan{}, false
}

// markdownBackticks returns the length of the run of backticks at the start of
// text.
func markdownBackticks(text string) int {
	return len(text) - len(strings.TrimLeft(text, "`"))
}

// markdownIsSpace tells if char is a space, a tab, or a line break.
func markdownIsSpace(char byte) bool {
	return char == ' ' || char == '\t' || char == '\n' || char == '\r'
}

// markdownWordCharBefore tells if the character that ends text is a letter or a
// digit.
func markdownWordCharBefore(text string) bool {
	char, _ := utf8.DecodeLastRuneInString(text)
	return text != "" && (unicode.IsLetter(char) || unicode.IsDigit(char))
}

// markdownWordChar tells if the character that starts text is a letter or a
// digit. An underscore next to one is a part of a word, such as in snake_case.
func markdownWordChar(text string) bool {
	char, _ := utf8.DecodeRuneInString(text)
	return text != "" && (unicode.IsLetter(char) || unicode.IsDigit(char))
}