Handle additional NWS forecast discussion formats

docs/roadmap/future.md

@@ -4,6 +4,45 @@
 
 This document is the catch-all roadmap for planned, deferred, aspirational, experimental, or unimplemented weatherfeeder work. Current behavior belongs in the canonical docs outside `docs/roadmap/`.
 
+## NWS AFD Parsing Resilience
+
+The current parser handles the concrete RAH, LWX, and MFR variants that
+motivated these ideas. Future work should keep those extension points
+maintainable as additional evidence appears.
+
+### Generalize Wrapper-Scoped Embedded Sections
+
+The scanner currently permits undotted nested headings only inside `PREV
+DISCUSSION`. If other wrapper identities are observed, replace the single
+wrapper check with a small explicit provider-local registry and add a fixture
+for each wrapper family. Do not make the leading dot globally optional: wrapper
+scope is the safeguard against classifying uppercase prose as a section.
+
+### Extend Conservative Preamble Classification
+
+Leading key-message metadata currently supports validated `Issued at`, `Updated
+at`, and `As of <clock> <weekday>...` forms. Add future wording variants as
+small, ordered classifiers with strict label boundaries and value grammars.
+Every addition should include collision tests proving that similar message prose
+and malformed metadata remain canonical content.
+
+### Keep List-Marker Recognition Extensible
+
+The marker parser currently supports hyphens, asterisks, `N)`, `N.`, `(N)`, and
+composite forms such as `- (N)`. If new decorators appear, evolve the helper
+toward an explicit marker grammar or typed classification result rather than a
+broad punctuation heuristic. Preserve positive-number and whitespace-boundary
+checks so ordinary prose is not stripped.
+
+### Maintain a Cross-Office Fixture Corpus
+
+The compact RAH, LWX, and current MFR fixtures seed regression coverage for the
+observed layouts. Future parser changes should add concise, deterministic HTML
+fixtures for materially distinct office formats and exercise them through both
+the provider parser and normalizer. Fixture comments should identify the format
+family and state that edited prose is not an archived product; tests must remain
+offline and assert both intended extraction and adjacent-section isolation.
+
 ## SPC Convective Outlook Follow-Ups
 
 ### Weatherapi Outlook Endpoints

internal/normalizers/nws/forecast_discussion_test.go

@@ -4,6 +4,7 @@ import (
 	"encoding/json"
 	"os"
 	"path/filepath"
+	"reflect"
 	"strings"
 	"testing"
 	"time"
@@ -71,6 +72,277 @@ func TestForecastDiscussionNormalizerProducesCanonicalSchema(t *testing.T) {
 	}
 }
 
+func TestForecastDiscussionNormalizerSupportsMixedHeadingFormats(t *testing.T) {
+	out, err := (ForecastDiscussionNormalizer{}).Normalize(nil, event.Event{
+		ID:        "evt-discussion-mixed-format",
+		Kind:      event.Kind(standards.KindForecastDiscussion),
+		Source:    "nws-discussion-test",
+		EmittedAt: time.Date(2026, 3, 28, 19, 25, 0, 0, time.UTC),
+		Schema:    standards.SchemaRawNWSForecastDiscussionV1,
+		Payload:   loadMixedFormatForecastDiscussionSampleHTML(t),
+	})
+	if err != nil {
+		t.Fatalf("Normalize() error = %v", err)
+	}
+	if out == nil {
+		t.Fatalf("Normalize() returned nil output")
+	}
+	if out.Kind != event.Kind(standards.KindForecastDiscussion) {
+		t.Fatalf("Kind = %q, want forecast_discussion", out.Kind)
+	}
+	if out.Schema != standards.SchemaWeatherForecastDiscussionV1 {
+		t.Fatalf("Schema = %q, want %q", out.Schema, standards.SchemaWeatherForecastDiscussionV1)
+	}
+	wantEffectiveAt := time.Date(2026, 3, 28, 19, 24, 0, 0, time.UTC)
+	if out.EffectiveAt == nil || !out.EffectiveAt.Equal(wantEffectiveAt) {
+		t.Fatalf("EffectiveAt = %v, want %s", out.EffectiveAt, wantEffectiveAt.Format(time.RFC3339))
+	}
+
+	payload, ok := out.Payload.(model.WeatherForecastDiscussion)
+	if !ok {
+		t.Fatalf("Payload type = %T, want model.WeatherForecastDiscussion", out.Payload)
+	}
+	if payload.ShortTerm == nil || payload.LongTerm == nil {
+		t.Fatalf("ShortTerm=%v LongTerm=%v, want both populated", payload.ShortTerm, payload.LongTerm)
+	}
+	if payload.ShortTerm.Qualifier != "Through Late Sunday Night" {
+		t.Fatalf("ShortTerm.Qualifier = %q", payload.ShortTerm.Qualifier)
+	}
+	if !strings.Contains(payload.ShortTerm.Text, "After a chilly morning") {
+		t.Fatalf("ShortTerm.Text missing expected prose: %q", payload.ShortTerm.Text)
+	}
+	if payload.LongTerm.Qualifier != "Monday through Next Saturday" {
+		t.Fatalf("LongTerm.Qualifier = %q", payload.LongTerm.Qualifier)
+	}
+	if !strings.Contains(payload.LongTerm.Text, "The peak of the warmth arrives Monday and Tuesday") {
+		t.Fatalf("LongTerm.Text missing expected prose: %q", payload.LongTerm.Text)
+	}
+	if strings.Contains(payload.LongTerm.Text, "AVIATION") || strings.Contains(payload.LongTerm.Text, "VFR conditions are expected") {
+		t.Fatalf("LongTerm.Text includes aviation content: %q", payload.LongTerm.Text)
+	}
+
+	b, err := json.Marshal(out.Payload)
+	if err != nil {
+		t.Fatalf("json.Marshal(payload) error = %v", err)
+	}
+	var fields map[string]any
+	if err := json.Unmarshal(b, &fields); err != nil {
+		t.Fatalf("json.Unmarshal(payload) error = %v", err)
+	}
+	for _, key := range []string{"aviation", "sections"} {
+		if _, ok := fields[key]; ok {
+			t.Fatalf("unexpected key %q in canonical payload", key)
+		}
+	}
+}
+
+func TestForecastDiscussionNormalizerSupportsCrossOfficeLayout(t *testing.T) {
+	in := event.Event{
+		ID:        "evt-discussion-bou",
+		Kind:      event.Kind(standards.KindForecastDiscussion),
+		Source:    "nws-discussion-bou-test",
+		EmittedAt: time.Date(2026, 4, 7, 19, 1, 0, 0, time.UTC),
+		Schema:    standards.SchemaRawNWSForecastDiscussionV1,
+		Payload:   loadForecastDiscussionBOUSampleHTML(t),
+	}
+
+	out, err := (ForecastDiscussionNormalizer{}).Normalize(nil, in)
+	if err != nil {
+		t.Fatalf("Normalize() error = %v", err)
+	}
+	if out == nil {
+		t.Fatalf("Normalize() returned nil output")
+	}
+	if out.ID != in.ID || out.Source != in.Source || !out.EmittedAt.Equal(in.EmittedAt) {
+		t.Fatalf("envelope = %#v, want ID/source/emittedAt from input", out)
+	}
+	if out.Kind != event.Kind(standards.KindForecastDiscussion) {
+		t.Fatalf("Kind = %q, want forecast_discussion", out.Kind)
+	}
+	if out.Schema != standards.SchemaWeatherForecastDiscussionV1 {
+		t.Fatalf("Schema = %q, want %q", out.Schema, standards.SchemaWeatherForecastDiscussionV1)
+	}
+	wantEffectiveAt := time.Date(2026, 4, 7, 19, 0, 0, 0, time.UTC)
+	if out.EffectiveAt == nil || !out.EffectiveAt.Equal(wantEffectiveAt) {
+		t.Fatalf("EffectiveAt = %v, want %s", out.EffectiveAt, wantEffectiveAt.Format(time.RFC3339))
+	}
+
+	payload, ok := out.Payload.(model.WeatherForecastDiscussion)
+	if !ok {
+		t.Fatalf("Payload type = %T, want model.WeatherForecastDiscussion", out.Payload)
+	}
+	if payload.OfficeID != "BOU" || payload.OfficeName != "National Weather Service Denver CO" {
+		t.Fatalf("OfficeID=%q OfficeName=%q", payload.OfficeID, payload.OfficeName)
+	}
+	wantMessages := []string{
+		"Strong winds are expected along the Front Range this evening.",
+		"Cooler temperatures arrive on Wednesday.",
+	}
+	if len(payload.KeyMessages) != len(wantMessages) {
+		t.Fatalf("KeyMessages = %#v, want %#v", payload.KeyMessages, wantMessages)
+	}
+	for i := range wantMessages {
+		if payload.KeyMessages[i] != wantMessages[i] {
+			t.Fatalf("KeyMessages[%d] = %q, want %q", i, payload.KeyMessages[i], wantMessages[i])
+		}
+	}
+	if payload.ShortTerm == nil || payload.LongTerm == nil {
+		t.Fatalf("ShortTerm=%v LongTerm=%v, want both populated", payload.ShortTerm, payload.LongTerm)
+	}
+	if payload.ShortTerm.Qualifier != "(Tonight through Wednesday)" || payload.ShortTerm.Text != "Gusty west winds will continue through the evening before decreasing overnight." {
+		t.Fatalf("ShortTerm = %#v", payload.ShortTerm)
+	}
+	if payload.LongTerm.Qualifier != "(Thursday through Saturday)" || payload.LongTerm.Text != "Warmer and drier conditions return Thursday, followed by a chance of showers Friday." {
+		t.Fatalf("LongTerm = %#v", payload.LongTerm)
+	}
+	if payload.ShortTerm.IssuedAt == nil || payload.LongTerm.IssuedAt == nil ||
+		!payload.ShortTerm.IssuedAt.Equal(wantEffectiveAt) || !payload.LongTerm.IssuedAt.Equal(wantEffectiveAt) {
+		t.Fatalf("section issue times = short %v long %v, want %s", payload.ShortTerm.IssuedAt, payload.LongTerm.IssuedAt, wantEffectiveAt.Format(time.RFC3339))
+	}
+
+	b, err := json.Marshal(out.Payload)
+	if err != nil {
+		t.Fatalf("json.Marshal(payload) error = %v", err)
+	}
+	var fields map[string]any
+	if err := json.Unmarshal(b, &fields); err != nil {
+		t.Fatalf("json.Unmarshal(payload) error = %v", err)
+	}
+	for _, key := range []string{"aviation", "discussion", "sections"} {
+		if _, ok := fields[key]; ok {
+			t.Fatalf("unexpected key %q in canonical payload", key)
+		}
+	}
+}
+
+func TestForecastDiscussionNormalizerSupportsCrossOfficeKeyMessageFixtures(t *testing.T) {
+	tests := []struct {
+		name        string
+		filename    string
+		id          string
+		source      string
+		emittedAt   time.Time
+		effectiveAt time.Time
+		messages    []string
+	}{
+		{
+			name:        "numbered key messages",
+			filename:    "forecast_discussion_bgm_numbered_sample.html",
+			id:          "evt-discussion-bgm",
+			source:      "nws-discussion-bgm-test",
+			emittedAt:   time.Date(2026, 4, 10, 17, 31, 0, 0, time.UTC),
+			effectiveAt: time.Date(2026, 4, 10, 17, 30, 0, 0, time.UTC),
+			messages: []string{
+				"Periods of rain are expected through Saturday, with locally heavier amounts possible.",
+				"Cooler temperatures return late this weekend.",
+			},
+		},
+		{
+			name:        "key points alias",
+			filename:    "forecast_discussion_mfr_key_points_sample.html",
+			id:          "evt-discussion-mfr",
+			source:      "nws-discussion-mfr-test",
+			emittedAt:   time.Date(2026, 4, 10, 19, 46, 0, 0, time.UTC),
+			effectiveAt: time.Date(2026, 4, 10, 19, 45, 0, 0, time.UTC),
+			messages: []string{
+				"Gusty winds will develop over exposed ridges, especially during the afternoon.",
+				"Inland valleys remain dry through Saturday.",
+			},
+		},
+		{
+			name:        "as of preamble",
+			filename:    "forecast_discussion_rah_as_of_sample.html",
+			id:          "evt-discussion-rah",
+			source:      "nws-discussion-rah-test",
+			emittedAt:   time.Date(2026, 8, 2, 16, 36, 0, 0, time.UTC),
+			effectiveAt: time.Date(2026, 8, 2, 16, 35, 0, 0, time.UTC),
+			messages: []string{
+				"Scattered storms may produce locally heavy rain this afternoon.",
+				"Drier weather arrives Monday.",
+			},
+		},
+		{
+			name:        "parenthesized numeric markers",
+			filename:    "forecast_discussion_lwx_parenthesized_number_sample.html",
+			id:          "evt-discussion-lwx",
+			source:      "nws-discussion-lwx-test",
+			emittedAt:   time.Date(2026, 8, 2, 18, 1, 0, 0, time.UTC),
+			effectiveAt: time.Date(2026, 8, 2, 18, 0, 0, 0, time.UTC),
+			messages: []string{
+				"Thunderstorms remain possible near the Blue Ridge this evening.",
+				"Seasonably warm conditions continue Monday.",
+			},
+		},
+		{
+			name:        "embedded key messages in previous discussion",
+			filename:    "forecast_discussion_mfr_prev_discussion_sample.html",
+			id:          "evt-discussion-mfr-previous",
+			source:      "nws-discussion-mfr-previous-test",
+			emittedAt:   time.Date(2026, 8, 2, 22, 20, 0, 0, time.UTC),
+			effectiveAt: time.Date(2026, 8, 2, 22, 19, 0, 0, time.UTC),
+			messages: []string{
+				"Heat returns to inland valleys Monday.",
+				"Gusty afternoon winds develop east of the Cascades.",
+			},
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			in := event.Event{
+				ID:        tt.id,
+				Kind:      event.Kind(standards.KindForecastDiscussion),
+				Source:    tt.source,
+				EmittedAt: tt.emittedAt,
+				Schema:    standards.SchemaRawNWSForecastDiscussionV1,
+				Payload:   loadForecastDiscussionFixtureHTML(t, tt.filename),
+			}
+
+			out, err := (ForecastDiscussionNormalizer{}).Normalize(nil, in)
+			if err != nil {
+				t.Fatalf("Normalize() error = %v", err)
+			}
+			if out == nil {
+				t.Fatalf("Normalize() returned nil output")
+			}
+			if out.ID != in.ID || out.Source != in.Source || !out.EmittedAt.Equal(in.EmittedAt) {
+				t.Fatalf("envelope = %#v, want ID/source/emittedAt from input", out)
+			}
+			if out.Kind != event.Kind(standards.KindForecastDiscussion) {
+				t.Fatalf("Kind = %q, want forecast_discussion", out.Kind)
+			}
+			if out.Schema != standards.SchemaWeatherForecastDiscussionV1 {
+				t.Fatalf("Schema = %q, want %q", out.Schema, standards.SchemaWeatherForecastDiscussionV1)
+			}
+			if out.EffectiveAt == nil || !out.EffectiveAt.Equal(tt.effectiveAt) {
+				t.Fatalf("EffectiveAt = %v, want %s", out.EffectiveAt, tt.effectiveAt.Format(time.RFC3339))
+			}
+
+			payload, ok := out.Payload.(model.WeatherForecastDiscussion)
+			if !ok {
+				t.Fatalf("Payload type = %T, want model.WeatherForecastDiscussion", out.Payload)
+			}
+			if !reflect.DeepEqual(payload.KeyMessages, tt.messages) {
+				t.Fatalf("KeyMessages = %#v, want %#v", payload.KeyMessages, tt.messages)
+			}
+
+			b, err := json.Marshal(payload)
+			if err != nil {
+				t.Fatalf("json.Marshal(payload) error = %v", err)
+			}
+			var fields map[string]any
+			if err := json.Unmarshal(b, &fields); err != nil {
+				t.Fatalf("json.Unmarshal(payload) error = %v", err)
+			}
+			for _, key := range []string{"nearTerm", "discussion", "aviation", "sections"} {
+				if _, ok := fields[key]; ok {
+					t.Fatalf("unexpected key %q in canonical payload", key)
+				}
+			}
+		})
+	}
+}
+
 func TestForecastDiscussionNormalizerRejectsMissingIssueTime(t *testing.T) {
 	_, err := (ForecastDiscussionNormalizer{}).Normalize(nil, event.Event{
 		ID:        "evt-discussion-bad",
@@ -128,3 +400,56 @@ func loadForecastDiscussionSampleHTML(t *testing.T) string {
 	}
 	return string(b)
 }
+
+func loadForecastDiscussionBOUSampleHTML(t *testing.T) string {
+	t.Helper()
+
+	path := filepath.Join("..", "..", "providers", "nws", "testdata", "forecast_discussion_bou_sample.html")
+	b, err := os.ReadFile(path)
+	if err != nil {
+		t.Fatalf("os.ReadFile(%q) error = %v", path, err)
+	}
+	return string(b)
+}
+
+func loadForecastDiscussionFixtureHTML(t *testing.T, filename string) string {
+	t.Helper()
+
+	path := filepath.Join("..", "..", "providers", "nws", "testdata", filename)
+	b, err := os.ReadFile(path)
+	if err != nil {
+		t.Fatalf("os.ReadFile(%q) error = %v", path, err)
+	}
+	return string(b)
+}
+
+func loadMixedFormatForecastDiscussionSampleHTML(t *testing.T) string {
+	t.Helper()
+
+	raw := loadForecastDiscussionSampleHTML(t)
+	replacements := []struct {
+		original    string
+		replacement string
+	}{
+		{
+			original:    ".SHORT TERM...  (Through Late Sunday Night)",
+			replacement: ".SHORT TERM /Through Late Sunday Night/...",
+		},
+		{
+			original:    ".LONG TERM...  (Monday through Next Saturday)",
+			replacement: ".LONG TERM /Monday through Next Saturday/...",
+		},
+		{
+			original:    ".AVIATION...  (For the 18z TAFs through 18z Sunday Afternoon)",
+			replacement: ".AVIATION /For the 18z TAFs through 18z Sunday Afternoon/...",
+		},
+	}
+	for _, replacement := range replacements {
+		if !strings.Contains(raw, replacement.original) {
+			t.Fatalf("fixture missing heading %q", replacement.original)
+		}
+		raw = strings.Replace(raw, replacement.original, replacement.replacement, 1)
+	}
+
+	return raw
+}

internal/providers/nws/forecast_discussion.go

@@ -27,8 +27,31 @@ type ForecastDiscussionSection struct {
 	Text      string
 }
 
+type forecastDiscussionSectionRole uint8
+
+const (
+	forecastDiscussionSectionRoleKeyMessages forecastDiscussionSectionRole = iota
+	forecastDiscussionSectionRoleShortTerm
+	forecastDiscussionSectionRoleLongTerm
+)
+
+type forecastDiscussionSectionHeading struct {
+	section   string
+	qualifier string
+}
+
+type forecastDiscussionSectionBlock struct {
+	heading forecastDiscussionSectionHeading
+	body    []string
+}
+
 var (
-	forecastDiscussionHeaderRE = regexp.MustCompile(`^\.(KEY MESSAGES|SHORT TERM|LONG TERM|AVIATION)\.\.\.(.*)$`)
+	forecastDiscussionSectionRoles = map[string]forecastDiscussionSectionRole{
+		"KEY MESSAGES": forecastDiscussionSectionRoleKeyMessages,
+		"KEY POINTS":   forecastDiscussionSectionRoleKeyMessages,
+		"SHORT TERM":   forecastDiscussionSectionRoleShortTerm,
+		"LONG TERM":    forecastDiscussionSectionRoleLongTerm,
+	}
 	forecastDiscussionAFDRE = regexp.MustCompile(`^AFD([A-Z]{3})$`)
 	forecastDiscussionWMORE = regexp.MustCompile(`\bK([A-Z]{3})\b`)
 	forecastDiscussionSigRE = regexp.MustCompile(`^[A-Z]{2,6}$`)
@@ -99,23 +122,31 @@ func ParseForecastDiscussionText(text string) (ForecastDiscussion, error) {
 		IssuedAt:   issuedAt.UTC(),
 	}
 
-	if block, ok := extractForecastDiscussionSection(lines, "KEY MESSAGES"); ok {
+	seenRoles := make(map[forecastDiscussionSectionRole]bool, len(forecastDiscussionSectionRoles))
-		out.KeyMessages = parseForecastDiscussionKeyMessages(block)
+	for _, block := range parseForecastDiscussionSectionBlocks(lines) {
+		role, ok := forecastDiscussionSectionRoles[block.heading.section]
+		if !ok || seenRoles[role] {
+			continue
 		}
-	if block, ok := extractForecastDiscussionSection(lines, "SHORT TERM"); ok {
+		seenRoles[role] = true
+
+		switch role {
+		case forecastDiscussionSectionRoleKeyMessages:
+			out.KeyMessages = parseForecastDiscussionKeyMessages(block.body)
+		case forecastDiscussionSectionRoleShortTerm:
 			section, err := parseForecastDiscussionTextSection(block)
 			if err != nil {
-			return ForecastDiscussion{}, fmt.Errorf("parse SHORT TERM: %w", err)
+				return ForecastDiscussion{}, fmt.Errorf("parse %s: %w", block.heading.section, err)
 			}
 			out.ShortTerm = &section
-	}
+		case forecastDiscussionSectionRoleLongTerm:
-	if block, ok := extractForecastDiscussionSection(lines, "LONG TERM"); ok {
 			section, err := parseForecastDiscussionTextSection(block)
 			if err != nil {
-			return ForecastDiscussion{}, fmt.Errorf("parse LONG TERM: %w", err)
+				return ForecastDiscussion{}, fmt.Errorf("parse %s: %w", block.heading.section, err)
 			}
 			out.LongTerm = &section
 		}
+	}
 
 	return out, nil
 }
@@ -285,8 +316,9 @@ func parseForecastDiscussionHeader(lines []string) (string, time.Time, error) {
 
 func parseForecastDiscussionIssueTime(line string) (time.Time, error) {
 	line = strings.TrimSpace(line)
-	line = strings.TrimPrefix(line, "Issued at ")
+	if isForecastDiscussionIssuedAtLine(line) {
-	line = strings.TrimSpace(line)
+		line = strings.TrimSpace(line[len("Issued at"):])
+	}
 
 	parts := strings.Fields(line)
 	if len(parts) != 7 {
@@ -386,40 +418,217 @@ func forecastDiscussionLocation(abbrev string) (*time.Location, error) {
 	return time.FixedZone(abbr, offset), nil
 }
 
-func extractForecastDiscussionSection(lines []string, section string) ([]string, bool) {
+func parseForecastDiscussionSectionHeading(line string) (forecastDiscussionSectionHeading, bool) {
-	target := "." + section + "..."
+	line = strings.TrimSpace(line)
-	for i, raw := range lines {
+	if len(line) < 2 || line[0] != '.' {
+		return forecastDiscussionSectionHeading{}, false
+	}
+
+	if strings.HasSuffix(line, "/...") {
+		return parseForecastDiscussionSlashQualifiedHeading(line)
+	}
+	if strings.HasSuffix(line, "...") {
+		if heading, ok := parseForecastDiscussionParenthesizedTerminalHeading(line); ok {
+			return heading, true
+		}
+	}
+	return parseForecastDiscussionEllipsisHeading(line)
+}
+
+func parseForecastDiscussionSlashQualifiedHeading(line string) (forecastDiscussionSectionHeading, bool) {
+	content := strings.TrimSuffix(line[1:], "/...")
+	separator := -1
+	for i := 1; i < len(content); i++ {
+		if content[i] == '/' && isForecastDiscussionHorizontalWhitespace(content[i-1]) {
+			separator = i
+			break
+		}
+	}
+	if separator < 0 {
+		return forecastDiscussionSectionHeading{}, false
+	}
+
+	section, ok := normalizeForecastDiscussionSectionIdentity(content[:separator])
+	if !ok {
+		return forecastDiscussionSectionHeading{}, false
+	}
+	qualifier := strings.TrimSpace(content[separator+1:])
+	if qualifier == "" {
+		return forecastDiscussionSectionHeading{}, false
+	}
+	return forecastDiscussionSectionHeading{section: section, qualifier: qualifier}, true
+}
+
+func parseForecastDiscussionParenthesizedTerminalHeading(line string) (forecastDiscussionSectionHeading, bool) {
+	if len(line) < 4 || line[0] != '.' || !strings.HasSuffix(line, "...") {
+		return forecastDiscussionSectionHeading{}, false
+	}
+
+	content := strings.TrimRight(line[1:len(line)-3], " \t")
+	if !strings.HasSuffix(content, ")") {
+		return forecastDiscussionSectionHeading{}, false
+	}
+
+	separator := -1
+	for i := 1; i < len(content); i++ {
+		if content[i] == '(' && isForecastDiscussionHorizontalWhitespace(content[i-1]) {
+			separator = i
+			break
+		}
+	}
+	if separator < 0 {
+		return forecastDiscussionSectionHeading{}, false
+	}
+
+	section, ok := normalizeForecastDiscussionSectionIdentity(content[:separator])
+	if !ok {
+		return forecastDiscussionSectionHeading{}, false
+	}
+	qualifier := content[separator:]
+	if len(qualifier) <= 2 || strings.TrimSpace(qualifier[1:len(qualifier)-1]) == "" {
+		return forecastDiscussionSectionHeading{}, false
+	}
+	return forecastDiscussionSectionHeading{section: section, qualifier: qualifier}, true
+}
+
+func parseForecastDiscussionEllipsisHeading(line string) (forecastDiscussionSectionHeading, bool) {
+	content := line[1:]
+	delimiter := strings.Index(content, "...")
+	if delimiter < 0 {
+		return forecastDiscussionSectionHeading{}, false
+	}
+
+	section, ok := normalizeForecastDiscussionSectionIdentity(content[:delimiter])
+	if !ok {
+		return forecastDiscussionSectionHeading{}, false
+	}
+	return forecastDiscussionSectionHeading{
+		section:   section,
+		qualifier: strings.TrimSpace(content[delimiter+3:]),
+	}, true
+}
+
+func normalizeForecastDiscussionSectionIdentity(raw string) (string, bool) {
+	var normalized strings.Builder
+	pendingSpace := false
+	hasLetterOrDigit := false
+
+	for i := 0; i < len(raw); i++ {
+		b := raw[i]
+		switch {
+		case isForecastDiscussionIdentityLetterOrDigit(b):
+			hasLetterOrDigit = true
+		case b == ' ' || b == '\t':
+			pendingSpace = normalized.Len() > 0
+			continue
+		case b == '/' && i > 0 && isForecastDiscussionHorizontalWhitespace(raw[i-1]):
+			return "", false
+		case b != '/' && b != '&' && b != '\'' && b != '-':
+			return "", false
+		}
+
+		if pendingSpace {
+			normalized.WriteByte(' ')
+			pendingSpace = false
+		}
+		normalized.WriteByte(b)
+	}
+	if !hasLetterOrDigit {
+		return "", false
+	}
+	return normalized.String(), true
+}
+
+func isForecastDiscussionIdentityLetterOrDigit(b byte) bool {
+	return b >= 'A' && b <= 'Z' || b >= '0' && b <= '9'
+}
+
+func isForecastDiscussionHorizontalWhitespace(b byte) bool {
+	return b == ' ' || b == '\t'
+}
+
+func parseForecastDiscussionSectionBlocks(lines []string) []forecastDiscussionSectionBlock {
+	var blocks []forecastDiscussionSectionBlock
+	var active *forecastDiscussionSectionBlock
+	embeddedHeadings := false
+
+	finish := func() {
+		if active == nil {
+			return
+		}
+		blocks = append(blocks, *active)
+		active = nil
+	}
+
+	for _, raw := range lines {
 		line := strings.TrimSpace(raw)
-		if !strings.HasPrefix(line, target) {
+		if line == "$$" {
+			finish()
+			break
+		}
+		if line == "&&" || strings.Contains(line, "WATCHES/WARNINGS/ADVISORIES") {
+			finish()
+			embeddedHeadings = false
 			continue
 		}
 
-		out := []string{line}
+		heading, ok := parseForecastDiscussionSectionHeading(raw)
-		for j := i + 1; j < len(lines); j++ {
+		if ok {
-			next := strings.TrimSpace(lines[j])
+			finish()
-			if next == "&&" || next == "$$" || strings.Contains(next, "WATCHES/WARNINGS/ADVISORIES") {
+			active = &forecastDiscussionSectionBlock{heading: heading}
-				break
+			embeddedHeadings = isForecastDiscussionEmbeddedSectionWrapper(heading.section)
+			continue
 		}
-			if j > i+1 && isForecastDiscussionSectionHeader(next) {
+		if embeddedHeadings {
-				break
+			heading, ok = parseForecastDiscussionEmbeddedSectionHeading(raw)
+			if ok {
+				finish()
+				active = &forecastDiscussionSectionBlock{heading: heading}
+				continue
 			}
-			out = append(out, lines[j])
 		}
-		return out, true
+		if active != nil {
+			active.body = append(active.body, raw)
 		}
-	return nil, false
+	}
+	finish()
+
+	return blocks
 }
 
-func isForecastDiscussionSectionHeader(line string) bool {
+func isForecastDiscussionEmbeddedSectionWrapper(section string) bool {
-	return forecastDiscussionHeaderRE.MatchString(strings.TrimSpace(line))
+	return section == "PREV DISCUSSION"
 }
 
-func parseForecastDiscussionKeyMessages(block []string) []string {
+func parseForecastDiscussionEmbeddedSectionHeading(line string) (forecastDiscussionSectionHeading, bool) {
-	if len(block) <= 1 {
+	line = strings.TrimSpace(line)
+	if line == "" || line[0] == '.' {
+		return forecastDiscussionSectionHeading{}, false
+	}
+	return parseForecastDiscussionSectionHeading("." + line)
+}
+
+func parseForecastDiscussionKeyMessages(body []string) []string {
+	body = removeForecastDiscussionPresentationMarkers(body)
+	body = trimBlankLines(body)
+	if len(body) > 0 && isForecastDiscussionKeyMessageMetadataLine(body[0]) {
+		body = trimBlankLines(body[1:])
+	}
+	if len(body) == 0 {
 		return nil
 	}
+	hasMarkers := false
+	for _, raw := range body {
+		line := strings.TrimSpace(raw)
+		if line == "" {
+			continue
+		}
+		if _, ok := stripForecastDiscussionKeyMessageMarker(line); ok {
+			hasMarkers = true
+			break
+		}
+	}
 
-	body := trimBlankLines(block[1:])
 	var messages []string
 	var current strings.Builder
 
@@ -431,15 +640,21 @@ func parseForecastDiscussionKeyMessages(block []string) []string {
 		current.Reset()
 	}
 
+	seenMarker := false
 	for _, raw := range body {
 		line := strings.TrimSpace(raw)
 		if line == "" {
+			if !hasMarkers || !seenMarker {
+				flush()
+			}
 			continue
 		}
-		if strings.HasPrefix(line, "-") {
+		if stripped, ok := stripForecastDiscussionKeyMessageMarker(line); ok {
 			flush()
-			line = strings.TrimSpace(strings.TrimPrefix(line, "-"))
+			seenMarker = true
-			current.WriteString(line)
+			line = stripped
+		}
+		if line == "" {
 			continue
 		}
 		if current.Len() > 0 {
@@ -452,25 +667,24 @@ func parseForecastDiscussionKeyMessages(block []string) []string {
 	return messages
 }
 
-func parseForecastDiscussionTextSection(block []string) (ForecastDiscussionSection, error) {
+func parseForecastDiscussionTextSection(block forecastDiscussionSectionBlock) (ForecastDiscussionSection, error) {
-	if len(block) == 0 {
-		return ForecastDiscussionSection{}, fmt.Errorf("empty section")
-	}
-
 	section := ForecastDiscussionSection{
-		Qualifier: parseForecastDiscussionQualifier(strings.TrimSpace(block[0])),
+		Qualifier: block.heading.qualifier,
 	}
 
-	body := trimBlankLines(block[1:])
+	body := trimBlankLines(removeForecastDiscussionPresentationMarkers(block.body))
+	if section.Qualifier == "" && len(body) > 0 && isForecastDiscussionStandaloneParenthetical(body[0]) {
+		section.Qualifier = strings.TrimSpace(body[0])
+		body = trimBlankLines(body[1:])
+	}
 	if len(body) == 0 {
 		return section, nil
 	}
 
-	first := strings.TrimSpace(body[0])
+	if isForecastDiscussionIssuedAtLine(body[0]) {
-	if strings.HasPrefix(first, "Issued at ") {
+		issuedAt, err := parseForecastDiscussionIssueTime(body[0])
-		issuedAt, err := parseForecastDiscussionIssueTime(first)
 		if err != nil {
-			return ForecastDiscussionSection{}, fmt.Errorf("parse section issuedAt %q: %w", first, err)
+			return ForecastDiscussionSection{}, fmt.Errorf("parse section issuedAt %q: %w", strings.TrimSpace(body[0]), err)
 		}
 		tt := issuedAt.UTC()
 		section.IssuedAt = &tt
@@ -482,12 +696,147 @@ func parseForecastDiscussionTextSection(block []string) (ForecastDiscussionSecti
 	return section, nil
 }
 
-func parseForecastDiscussionQualifier(header string) string {
+func isForecastDiscussionPresentationMarker(line string) bool {
-	m := forecastDiscussionHeaderRE.FindStringSubmatch(header)
+	switch {
-	if len(m) != 3 {
+	case strings.EqualFold(strings.TrimSpace(line), "-- Changed Discussion --"):
-		return ""
+		return true
+	case strings.EqualFold(strings.TrimSpace(line), "-- End Changed Discussion --"):
+		return true
+	default:
+		return false
 	}
-	return strings.TrimSpace(m[2])
+}
+
+func removeForecastDiscussionPresentationMarkers(lines []string) []string {
+	body := make([]string, 0, len(lines))
+	for _, line := range lines {
+		if !isForecastDiscussionPresentationMarker(line) {
+			body = append(body, line)
+		}
+	}
+	return body
+}
+
+func isForecastDiscussionStandaloneParenthetical(line string) bool {
+	line = strings.TrimSpace(line)
+	return len(line) > 2 && line[0] == '(' && line[len(line)-1] == ')' && strings.TrimSpace(line[1:len(line)-1]) != ""
+}
+
+func isForecastDiscussionIssuedAtLine(line string) bool {
+	line = strings.TrimSpace(line)
+	return len(line) > len("Issued at") &&
+		strings.EqualFold(line[:len("Issued at")], "Issued at") &&
+		isForecastDiscussionHorizontalWhitespace(line[len("Issued at")])
+}
+
+func isForecastDiscussionKeyMessageMetadataLine(line string) bool {
+	line = strings.TrimSpace(line)
+	for _, label := range []string{"Issued at", "Updated at"} {
+		if !hasForecastDiscussionASCIIPrefix(line, label) || len(line) == len(label) || !isForecastDiscussionHorizontalWhitespace(line[len(label)]) {
+			continue
+		}
+		if _, err := parseForecastDiscussionIssueTime(strings.TrimSpace(line[len(label):])); err == nil {
+			return true
+		}
+	}
+	return isForecastDiscussionKeyMessageAsOfLine(line)
+}
+
+func isForecastDiscussionKeyMessageAsOfLine(line string) bool {
+	const label = "As of"
+
+	if !hasForecastDiscussionASCIIPrefix(line, label) || len(line) == len(label) || !isForecastDiscussionHorizontalWhitespace(line[len(label)]) {
+		return false
+	}
+	remainder := strings.TrimSpace(line[len(label):])
+	if !strings.HasSuffix(remainder, "...") {
+		return false
+	}
+	fields := strings.Fields(strings.TrimSpace(strings.TrimSuffix(remainder, "...")))
+	if len(fields) != 3 {
+		return false
+	}
+	if _, _, err := parseForecastDiscussionClock(fields[0], fields[1]); err != nil {
+		return false
+	}
+	switch strings.ToLower(fields[2]) {
+	case "monday", "tuesday", "wednesday", "thursday", "friday", "saturday", "sunday":
+		return true
+	default:
+		return false
+	}
+}
+
+func stripForecastDiscussionKeyMessageMarker(line string) (string, bool) {
+	if line == "" {
+		return "", false
+	}
+
+	if line[0] == '-' || line[0] == '*' {
+		content := line[1:]
+		hadWhitespace := len(content) > 0 && isForecastDiscussionHorizontalWhitespace(content[0])
+		content = strings.TrimLeft(content, " \t")
+		if hadWhitespace {
+			if stripped, ok := stripForecastDiscussionKeyMessageNumericMarker(content); ok {
+				content = stripped
+			}
+		}
+		return content, true
+	}
+
+	return stripForecastDiscussionKeyMessageNumericMarker(line)
+}
+
+func stripForecastDiscussionKeyMessageNumericMarker(line string) (string, bool) {
+	digitStart := 0
+	digitEnd := 0
+	parenthesized := len(line) > 0 && line[0] == '('
+	if parenthesized {
+		digitStart = 1
+		digitEnd = 1
+	}
+	for digitEnd < len(line) && line[digitEnd] >= '0' && line[digitEnd] <= '9' {
+		digitEnd++
+	}
+	if digitEnd == digitStart || digitEnd == len(line) {
+		return "", false
+	}
+	if parenthesized && line[digitEnd] != ')' {
+		return "", false
+	}
+	if !parenthesized && line[digitEnd] != ')' && line[digitEnd] != '.' {
+		return "", false
+	}
+
+	markerEnd := digitEnd + 1
+	if markerEnd < len(line) && !isForecastDiscussionHorizontalWhitespace(line[markerEnd]) {
+		return "", false
+	}
+	value, err := strconv.ParseUint(line[digitStart:digitEnd], 10, 0)
+	if err != nil || value == 0 {
+		return "", false
+	}
+	return strings.TrimLeft(line[markerEnd:], " \t"), true
+}
+
+func hasForecastDiscussionASCIIPrefix(line, prefix string) bool {
+	if len(line) < len(prefix) {
+		return false
+	}
+	for i := range prefix {
+		actual := line[i]
+		if actual >= 'A' && actual <= 'Z' {
+			actual += 'a' - 'A'
+		}
+		expected := prefix[i]
+		if expected >= 'A' && expected <= 'Z' {
+			expected += 'a' - 'A'
+		}
+		if actual != expected {
+			return false
+		}
+	}
+	return true
 }
 
 func trimBlankLines(lines []string) []string {

internal/providers/nws/testdata/forecast_discussion_bgm_numbered_sample.html

@@ -0,0 +1,26 @@
+<!DOCTYPE html>
+<!-- Representative BGM/CTP-style layout; prose is concise edited test data, not an archived product. -->
+<html>
+  <body>
+    <pre class="glossaryProduct">
+FXUS61 KBGM 101730
+AFDBGM
+
+Area Forecast Discussion
+National Weather Service Binghamton NY
+130 PM EDT Fri Apr 10 2026
+
+.KEY MESSAGES...
+1) Periods of rain are expected through Saturday,
+with locally heavier amounts possible.
+2. Cooler temperatures return late this weekend.
+
+.DISCUSSION...
+Discussion details remain boundary-only content.
+
+$$
+
+WFO BGM
+    </pre>
+  </body>
+</html>

internal/providers/nws/testdata/forecast_discussion_bou_sample.html

@@ -0,0 +1,48 @@
+<!DOCTYPE html>
+<html>
+  <body>
+    <pre class="glossaryProduct">
+FXUS65 KBOU 071900
+AFDBOU
+
+Area Forecast Discussion
+National Weather Service Denver CO
+100 PM MDT Tue Apr 7 2026
+
+.KEY MESSAGES...
+-- Changed Discussion --
+Updated at 100 PM MDT Tue Apr 7 2026
+- Strong winds are expected along the Front Range this evening.
+- Cooler temperatures arrive on Wednesday.
+-- End Changed Discussion --
+
+&&
+
+.SHORT TERM...
+(Tonight through Wednesday)
+Issued at 100 PM MDT Tue Apr 7 2026
+
+Gusty west winds will continue through the evening before decreasing overnight.
+
+&&
+
+.LONG TERM...
+(Thursday through Saturday)
+ISSUED AT 100 PM MDT Tue Apr 7 2026
+
+Warmer and drier conditions return Thursday, followed by a chance of showers Friday.
+
+&&
+
+.AVIATION...
+
+VFR conditions are expected at Denver-area terminals through Wednesday morning.
+
+&&
+
+$$
+
+WFO BOU
+    </pre>
+  </body>
+</html>

internal/providers/nws/testdata/forecast_discussion_lwx_parenthesized_number_sample.html

@@ -0,0 +1,27 @@
+<!DOCTYPE html>
+<!-- Representative LWX-style layout; prose is concise edited test data, not an archived product. -->
+<html>
+  <body>
+    <pre class="glossaryProduct">
+FXUS61 KLWX 021800
+AFDLWX
+
+Area Forecast Discussion
+National Weather Service Baltimore MD/Washington DC
+200 PM EDT Sun Aug 2 2026
+
+.KEY MESSAGES...
+- (1) Thunderstorms remain possible near the Blue Ridge this evening.
+- (2) Seasonably warm conditions continue Monday.
+
+&&
+
+.AVIATION...
+Aviation details remain boundary-only content.
+
+$$
+
+WFO LWX
+    </pre>
+  </body>
+</html>

internal/providers/nws/testdata/forecast_discussion_mfr_key_points_sample.html

@@ -0,0 +1,25 @@
+<!DOCTYPE html>
+<!-- Representative MFR-style layout; prose is concise edited test data, not an archived product. -->
+<html>
+  <body>
+    <pre class="glossaryProduct">
+FXUS66 KMFR 101945
+AFDMFR
+
+Area Forecast Discussion
+National Weather Service Medford OR
+1245 PM PDT Fri Apr 10 2026
+
+.KEY POINTS...
+* Gusty winds will develop over exposed ridges,
+especially during the afternoon.
+* Inland valleys remain dry through Saturday.
+.DISCUSSION (Today through Thursday)...
+Discussion details must not be included with key points.
+
+$$
+
+WFO MFR
+    </pre>
+  </body>
+</html>

internal/providers/nws/testdata/forecast_discussion_mfr_prev_discussion_sample.html

@@ -0,0 +1,33 @@
+<!DOCTYPE html>
+<!-- Representative current MFR previous-discussion wrapper layout; prose is concise edited test data, not an archived product. -->
+<html>
+  <body>
+    <pre class="glossaryProduct">
+FXUS66 KMFR 022219
+AFDMFR
+
+Area Forecast Discussion
+National Weather Service Medford OR
+319 PM PDT Sun Aug 2 2026
+
+.PREV DISCUSSION... /Issued 319 PM PDT Sun Aug 2 2026/
+
+KEY MESSAGES...
+
+* Heat returns to inland valleys Monday.
+* Gusty afternoon winds develop east of the Cascades.
+
+DISCUSSION...
+Discussion details must not be included with key messages.
+
+&&
+
+.MFR WATCHES/WARNINGS/ADVISORIES...
+None.
+
+$$
+
+WFO MFR
+    </pre>
+  </body>
+</html>

internal/providers/nws/testdata/forecast_discussion_rah_as_of_sample.html

@@ -0,0 +1,29 @@
+<!DOCTYPE html>
+<!-- Representative RAH-style layout; prose is concise edited test data, not an archived product. -->
+<html>
+  <body>
+    <pre class="glossaryProduct">
+FXUS62 KRAH 021635
+AFDRAH
+
+Area Forecast Discussion
+National Weather Service Raleigh NC
+1235 PM EDT Sun Aug 2 2026
+
+.KEY MESSAGES...
+As of 1235 PM Sunday...
+
+1) Scattered storms may produce locally heavy rain this afternoon.
+2) Drier weather arrives Monday.
+
+&&
+
+.DISCUSSION...
+Discussion details remain boundary-only content.
+
+$$
+
+WFO RAH
+    </pre>
+  </body>
+</html>