weatherfeeder/internal/providers/nws/forecast_discussion.go

package nws

import (
	"fmt"
	"html"
	"regexp"
	"strconv"
	"strings"
	"time"
)

type ForecastDiscussion struct {
	OfficeID   string
	OfficeName string
	Product    string
	IssuedAt   time.Time
	UpdatedAt  *time.Time

	KeyMessages []string
	ShortTerm   *ForecastDiscussionSection
	LongTerm    *ForecastDiscussionSection
}

type ForecastDiscussionSection struct {
	Qualifier string
	IssuedAt  *time.Time
	Text      string
}

var (
	forecastDiscussionHeaderRE = regexp.MustCompile(`^\.(KEY MESSAGES|SHORT TERM|LONG TERM|AVIATION)\.\.\.(.*)$`)
	forecastDiscussionAFDRE    = regexp.MustCompile(`^AFD([A-Z]{3})$`)
	forecastDiscussionWMORE    = regexp.MustCompile(`\bK([A-Z]{3})\b`)
	forecastDiscussionSigRE    = regexp.MustCompile(`^[A-Z]{2,6}$`)
)

func ParseForecastDiscussionHTML(raw string) (ForecastDiscussion, error) {
	text, err := ExtractForecastDiscussionText(raw)
	if err != nil {
		return ForecastDiscussion{}, err
	}

	parsed, err := ParseForecastDiscussionText(text)
	if err != nil {
		return ForecastDiscussion{}, err
	}

	parsed.UpdatedAt = parseForecastDiscussionUpdatedAt(raw)
	return parsed, nil
}

func ExtractForecastDiscussionText(raw string) (string, error) {
	lower := strings.ToLower(raw)
	searchFrom := 0
	for {
		openStart := strings.Index(lower[searchFrom:], "<pre")
		if openStart < 0 {
			return "", fmt.Errorf("missing <pre class=\"glossaryProduct\"> block")
		}
		openStart += searchFrom

		openEnd := strings.Index(lower[openStart:], ">")
		if openEnd < 0 {
			return "", fmt.Errorf("unterminated <pre> tag")
		}
		openEnd += openStart

		tag := lower[openStart : openEnd+1]
		if isGlossaryProductTag(tag) {
			closeStart := strings.Index(lower[openEnd+1:], "</pre>")
			if closeStart < 0 {
				return "", fmt.Errorf("missing closing </pre> for glossaryProduct block")
			}
			closeStart += openEnd + 1

			text := html.UnescapeString(raw[openEnd+1 : closeStart])
			text = strings.ReplaceAll(text, "\r\n", "\n")
			text = strings.ReplaceAll(text, "\r", "\n")
			return text, nil
		}

		searchFrom = openEnd + 1
	}
}

func ParseForecastDiscussionText(text string) (ForecastDiscussion, error) {
	lines := splitLines(text)

	officeID := parseForecastDiscussionOfficeID(lines)
	officeName, issuedAt, err := parseForecastDiscussionHeader(lines)
	if err != nil {
		return ForecastDiscussion{}, err
	}

	out := ForecastDiscussion{
		OfficeID:   officeID,
		OfficeName: officeName,
		Product:    "afd",
		IssuedAt:   issuedAt.UTC(),
	}

	if block, ok := extractForecastDiscussionSection(lines, "KEY MESSAGES"); ok {
		out.KeyMessages = parseForecastDiscussionKeyMessages(block)
	}
	if block, ok := extractForecastDiscussionSection(lines, "SHORT TERM"); ok {
		section, err := parseForecastDiscussionTextSection(block)
		if err != nil {
			return ForecastDiscussion{}, fmt.Errorf("parse SHORT TERM: %w", err)
		}
		out.ShortTerm = &section
	}
	if block, ok := extractForecastDiscussionSection(lines, "LONG TERM"); ok {
		section, err := parseForecastDiscussionTextSection(block)
		if err != nil {
			return ForecastDiscussion{}, fmt.Errorf("parse LONG TERM: %w", err)
		}
		out.LongTerm = &section
	}

	return out, nil
}

func isGlossaryProductTag(tag string) bool {
	tag = strings.ToLower(tag)
	return strings.Contains(tag, `class="glossaryproduct"`) ||
		strings.Contains(tag, `class='glossaryproduct'`) ||
		strings.Contains(tag, `class="glossaryproduct `) ||
		strings.Contains(tag, `class='glossaryproduct `)
}

func parseForecastDiscussionUpdatedAt(raw string) *time.Time {
	lower := strings.ToLower(raw)
	searchFrom := 0
	for {
		metaStart := strings.Index(lower[searchFrom:], "<meta")
		if metaStart < 0 {
			return nil
		}
		metaStart += searchFrom

		metaEnd := strings.Index(lower[metaStart:], ">")
		if metaEnd < 0 {
			return nil
		}
		metaEnd += metaStart

		tag := raw[metaStart : metaEnd+1]
		if !strings.EqualFold(strings.TrimSpace(extractHTMLAttr(tag, "name")), "DC.date.created") {
			searchFrom = metaEnd + 1
			continue
		}

		content := strings.TrimSpace(extractHTMLAttr(tag, "content"))
		if content == "" {
			return nil
		}
		t, err := ParseTime(content)
		if err != nil {
			return nil
		}
		tt := t.UTC()
		return &tt
	}
}

func extractHTMLAttr(tag, attr string) string {
	lower := strings.ToLower(tag)
	attrLower := strings.ToLower(attr)
	for i := 0; i < len(lower); i++ {
		idx := strings.Index(lower[i:], attrLower)
		if idx < 0 {
			return ""
		}
		idx += i
		if idx > 0 {
			prev := lower[idx-1]
			if isAttrNameChar(prev) {
				i = idx + len(attrLower)
				continue
			}
		}
		j := idx + len(attrLower)
		for j < len(lower) && isHTMLSpace(lower[j]) {
			j++
		}
		if j >= len(lower) || lower[j] != '=' {
			i = idx + len(attrLower)
			continue
		}
		j++
		for j < len(lower) && isHTMLSpace(lower[j]) {
			j++
		}
		if j >= len(tag) {
			return ""
		}
		quote := tag[j]
		if quote != '"' && quote != '\'' {
			return ""
		}
		j++
		k := j
		for k < len(tag) && tag[k] != quote {
			k++
		}
		if k >= len(tag) {
			return ""
		}
		return html.UnescapeString(tag[j:k])
	}
	return ""
}

func isHTMLSpace(b byte) bool {
	switch b {
	case ' ', '\n', '\r', '\t', '\f':
		return true
	default:
		return false
	}
}

func isAttrNameChar(b byte) bool {
	switch {
	case b >= 'a' && b <= 'z':
		return true
	case b >= 'A' && b <= 'Z':
		return true
	case b >= '0' && b <= '9':
		return true
	case b == '-' || b == '_' || b == ':':
		return true
	default:
		return false
	}
}

func splitLines(text string) []string {
	text = strings.ReplaceAll(text, "\r\n", "\n")
	text = strings.ReplaceAll(text, "\r", "\n")
	return strings.Split(text, "\n")
}

func parseForecastDiscussionOfficeID(lines []string) string {
	for _, raw := range lines {
		line := strings.TrimSpace(raw)
		if m := forecastDiscussionAFDRE.FindStringSubmatch(line); len(m) == 2 {
			return m[1]
		}
	}
	for _, raw := range lines {
		line := strings.TrimSpace(raw)
		if m := forecastDiscussionWMORE.FindStringSubmatch(line); len(m) == 2 {
			return m[1]
		}
	}
	return ""
}

func parseForecastDiscussionHeader(lines []string) (string, time.Time, error) {
	for i, raw := range lines {
		line := strings.TrimSpace(raw)
		if !strings.HasPrefix(line, "National Weather Service ") {
			continue
		}

		officeName := line
		for j := i + 1; j < len(lines); j++ {
			tsLine := strings.TrimSpace(lines[j])
			if tsLine == "" {
				continue
			}
			issuedAt, err := parseForecastDiscussionIssueTime(tsLine)
			if err != nil {
				return "", time.Time{}, fmt.Errorf("parse bulletin issuedAt %q: %w", tsLine, err)
			}
			return officeName, issuedAt.UTC(), nil
		}

		return "", time.Time{}, fmt.Errorf("missing bulletin issue time after office line")
	}

	return "", time.Time{}, fmt.Errorf("missing office header")
}

func parseForecastDiscussionIssueTime(line string) (time.Time, error) {
	line = strings.TrimSpace(line)
	line = strings.TrimPrefix(line, "Issued at ")
	line = strings.TrimSpace(line)

	parts := strings.Fields(line)
	if len(parts) != 7 {
		return time.Time{}, fmt.Errorf("unexpected issue time format")
	}

	loc, err := forecastDiscussionLocation(parts[2])
	if err != nil {
		return time.Time{}, err
	}

	datePart, err := time.Parse("Mon Jan 2 2006", strings.Join(parts[3:], " "))
	if err != nil {
		return time.Time{}, err
	}

	hour, minute, err := parseForecastDiscussionClock(parts[0], parts[1])
	if err != nil {
		return time.Time{}, err
	}

	return time.Date(
		datePart.Year(),
		datePart.Month(),
		datePart.Day(),
		hour,
		minute,
		0,
		0,
		loc,
	), nil
}

func parseForecastDiscussionClock(rawClock, rawAMPM string) (int, int, error) {
	clock := strings.TrimSpace(rawClock)
	ampm := strings.ToUpper(strings.TrimSpace(rawAMPM))
	if ampm != "AM" && ampm != "PM" {
		return 0, 0, fmt.Errorf("unexpected meridiem %q", rawAMPM)
	}

	n, err := strconv.Atoi(clock)
	if err != nil {
		return 0, 0, fmt.Errorf("invalid clock %q", rawClock)
	}

	hour := n
	minute := 0
	if len(clock) >= 3 {
		hour = n / 100
		minute = n % 100
	}

	if hour < 1 || hour > 12 {
		return 0, 0, fmt.Errorf("invalid hour %q", rawClock)
	}
	if minute < 0 || minute > 59 {
		return 0, 0, fmt.Errorf("invalid minute %q", rawClock)
	}

	if ampm == "AM" {
		if hour == 12 {
			hour = 0
		}
		return hour, minute, nil
	}

	if hour != 12 {
		hour += 12
	}
	return hour, minute, nil
}

func forecastDiscussionLocation(abbrev string) (*time.Location, error) {
	offsets := map[string]int{
		"AST":  -4 * 3600,
		"ADT":  -3 * 3600,
		"EST":  -5 * 3600,
		"EDT":  -4 * 3600,
		"CST":  -6 * 3600,
		"CDT":  -5 * 3600,
		"MST":  -7 * 3600,
		"MDT":  -6 * 3600,
		"PST":  -8 * 3600,
		"PDT":  -7 * 3600,
		"AKST": -9 * 3600,
		"AKDT": -8 * 3600,
		"HST":  -10 * 3600,
		"UTC":  0,
		"GMT":  0,
	}

	abbr := strings.ToUpper(strings.TrimSpace(abbrev))
	offset, ok := offsets[abbr]
	if !ok {
		return nil, fmt.Errorf("unsupported time zone %q", abbrev)
	}
	return time.FixedZone(abbr, offset), nil
}

func extractForecastDiscussionSection(lines []string, section string) ([]string, bool) {
	target := "." + section + "..."
	for i, raw := range lines {
		line := strings.TrimSpace(raw)
		if !strings.HasPrefix(line, target) {
			continue
		}

		out := []string{line}
		for j := i + 1; j < len(lines); j++ {
			next := strings.TrimSpace(lines[j])
			if next == "&&" || next == "$$" || strings.Contains(next, "WATCHES/WARNINGS/ADVISORIES") {
				break
			}
			if j > i+1 && isForecastDiscussionSectionHeader(next) {
				break
			}
			out = append(out, lines[j])
		}
		return out, true
	}
	return nil, false
}

func isForecastDiscussionSectionHeader(line string) bool {
	return forecastDiscussionHeaderRE.MatchString(strings.TrimSpace(line))
}

func parseForecastDiscussionKeyMessages(block []string) []string {
	if len(block) <= 1 {
		return nil
	}

	body := trimBlankLines(block[1:])
	var messages []string
	var current strings.Builder

	flush := func() {
		msg := strings.TrimSpace(current.String())
		if msg != "" {
			messages = append(messages, msg)
		}
		current.Reset()
	}

	for _, raw := range body {
		line := strings.TrimSpace(raw)
		if line == "" {
			continue
		}
		if strings.HasPrefix(line, "-") {
			flush()
			line = strings.TrimSpace(strings.TrimPrefix(line, "-"))
			current.WriteString(line)
			continue
		}
		if current.Len() > 0 {
			current.WriteByte(' ')
		}
		current.WriteString(line)
	}
	flush()

	return messages
}

func parseForecastDiscussionTextSection(block []string) (ForecastDiscussionSection, error) {
	if len(block) == 0 {
		return ForecastDiscussionSection{}, fmt.Errorf("empty section")
	}

	section := ForecastDiscussionSection{
		Qualifier: parseForecastDiscussionQualifier(strings.TrimSpace(block[0])),
	}

	body := trimBlankLines(block[1:])
	if len(body) == 0 {
		return section, nil
	}

	first := strings.TrimSpace(body[0])
	if strings.HasPrefix(first, "Issued at ") {
		issuedAt, err := parseForecastDiscussionIssueTime(first)
		if err != nil {
			return ForecastDiscussionSection{}, fmt.Errorf("parse section issuedAt %q: %w", first, err)
		}
		tt := issuedAt.UTC()
		section.IssuedAt = &tt
		body = trimBlankLines(body[1:])
	}

	body = trimForecastDiscussionSignatureLines(body)
	section.Text = joinForecastDiscussionParagraphs(body)
	return section, nil
}

func parseForecastDiscussionQualifier(header string) string {
	m := forecastDiscussionHeaderRE.FindStringSubmatch(header)
	if len(m) != 3 {
		return ""
	}
	return strings.TrimSpace(m[2])
}

func trimBlankLines(lines []string) []string {
	start := 0
	for start < len(lines) && strings.TrimSpace(lines[start]) == "" {
		start++
	}

	end := len(lines)
	for end > start && strings.TrimSpace(lines[end-1]) == "" {
		end--
	}

	return lines[start:end]
}

func trimForecastDiscussionSignatureLines(lines []string) []string {
	lines = trimBlankLines(lines)
	for len(lines) > 0 {
		last := strings.TrimSpace(lines[len(lines)-1])
		if last == "" {
			lines = lines[:len(lines)-1]
			continue
		}
		if forecastDiscussionSigRE.MatchString(last) {
			lines = trimBlankLines(lines[:len(lines)-1])
			continue
		}
		break
	}
	return lines
}

func joinForecastDiscussionParagraphs(lines []string) string {
	lines = trimBlankLines(lines)
	if len(lines) == 0 {
		return ""
	}

	var paragraphs []string
	current := make([]string, 0, len(lines))

	flush := func() {
		if len(current) == 0 {
			return
		}
		paragraphs = append(paragraphs, strings.Join(current, " "))
		current = current[:0]
	}

	for _, raw := range lines {
		line := strings.TrimSpace(raw)
		if line == "" {
			flush()
			continue
		}
		current = append(current, line)
	}
	flush()

	return strings.Join(paragraphs, "\n\n")
}