1 files changed, 156 insertions, 0 deletions
diff --git a/internal/liturgy/parse.go b/internal/liturgy/parse.go
new file mode 100644
index 0000000..64e6928
--- /dev/null
+++ b/internal/liturgy/parse.go
@@ -0,0 +1,156 @@
+package liturgy
+
+import (
+	"fmt"
+	"html"
+	"regexp"
+	"strings"
+)
+
+// lectionaryTabs lists the two reading sets carried on the page: the
+// lectionary in force since Advent 2015 ("nowy") and the one it replaced
+// ("stary"). They are not interchangeable -- the acclamation can cite a
+// different book entirely, and the older set contains malformed citations.
+// Pick by id, never by position in the markup, or a reordering silently
+// switches edition.
+var lectionaryTabs = []string{"tabnowy0all", "tabstary0all"}
+
+var (
+	brRe       = regexp.MustCompile(`(?i)<br\s*/?>`)
+	tagRe      = regexp.MustCompile(`<[^>]+>`)
+	wsRe       = regexp.MustCompile(`[ \t]+`)
+	h2Re       = regexp.MustCompile(`(?s)<h2>(.*?)</h2>`)
+	h4Re       = regexp.MustCompile(`(?s)<h4>(.*?)</h4>`)
+	pRe        = regexp.MustCompile(`(?s)<p>(.*?)</p>`)
+	citationRe = regexp.MustCompile(`\((.+)\)\s*$`)
+)
+
+// panePattern matches the opening tag of the tab-pane div carrying the given
+// tab id, e.g. `<div class="tab-pane fade " id="tabnowy0all">`.
+func panePattern(tab string) *regexp.Regexp {
+	return regexp.MustCompile(`<div class="tab-pane[^"]*"\s+id="` + regexp.QuoteMeta(tab) + `">`)
+}
+
+// htmlToLines turns an HTML fragment into a list of non-empty text lines.
+// <br> marks a verse line break (used in psalms/acclamations); other tags
+// are dropped, entities decoded, and intra-line whitespace collapsed.
+func htmlToLines(fragment string) []string {
+	fragment = brRe.ReplaceAllString(fragment, "\n")
+	fragment = tagRe.ReplaceAllString(fragment, "")
+	text := html.UnescapeString(fragment)
+	var lines []string
+	for _, ln := range strings.Split(text, "\n") {
+		ln = strings.TrimSpace(wsRe.ReplaceAllString(ln, " "))
+		if ln != "" {
+			lines = append(lines, ln)
+		}
+	}
+	return lines
+}
+
+// Parse extracts every reading section from the page's preferred lectionary
+// tab (falling back to the superseded one), erroring loudly if neither tab is
+// present on the page, or the tab is found but carries no sections -- a
+// layout change should never be mistaken for a quiet day with no readings.
+func Parse(pageHTML string) ([]Section, error) {
+	var loc []int
+	for _, tab := range lectionaryTabs {
+		if m := panePattern(tab).FindStringIndex(pageHTML); m != nil {
+			loc = m
+			break
+		}
+	}
+	if loc == nil {
+		if strings.Contains(pageHTML, "Przykro nam") {
+			return nil, fmt.Errorf("no reading published for this date yet")
+		}
+		return nil, fmt.Errorf(
+			"no reading tab (%s) found on page -- the site layout may have changed",
+			strings.Join(lectionaryTabs, "/"),
+		)
+	}
+
+	rest := pageHTML[loc[1]:]
+	block := rest
+	if nxt := strings.Index(rest, `<div class="tab-pane`); nxt != -1 {
+		block = rest[:nxt]
+	}
+
+	heads := h2Re.FindAllStringSubmatchIndex(block, -1)
+	var sections []Section
+	czytanieCount := 0
+	for i, h := range heads {
+		bodyEnd := len(block)
+		if i+1 < len(heads) {
+			bodyEnd = heads[i+1][0]
+		}
+		body := block[h[1]:bodyEnd]
+
+		heading := strings.Join(htmlToLines(block[h[2]:h[3]]), " ")
+
+		subtitle := ""
+		if sub := h4Re.FindStringSubmatch(body); sub != nil {
+			subtitle = strings.Join(htmlToLines(sub[1]), " ")
+		}
+
+		var paragraphs [][]string
+		for _, p := range pRe.FindAllStringSubmatch(body, -1) {
+			if lines := htmlToLines(p[1]); len(lines) > 0 {
+				paragraphs = append(paragraphs, lines)
+			}
+		}
+
+		// A heading is expected to always carry a parenthetical citation;
+		// if one is somehow missing, leave Citation empty rather than
+		// failing the whole parse over one section.
+		citation, _ := ExtractCitation(heading)
+
+		sections = append(sections, Section{
+			Heading:    heading,
+			Subtitle:   subtitle,
+			Citation:   citation,
+			PartID:     partID(heading, &czytanieCount),
+			Paragraphs: paragraphs,
+		})
+	}
+
+	// A layout change can leave the tab findable but empty; say so rather
+	// than returning nothing and looking like a quiet day.
+	if len(sections) == 0 {
+		return nil, fmt.Errorf("reading tab found but no sections in it -- the site layout may have changed")
+	}
+	return sections, nil
+}
+
+// partID assigns the stable liturgical-part identifier for a section heading.
+// A second "1. czytanie" heading on the same day (a split feast offering two
+// alternative first readings) becomes "drugie_czytanie" instead of colliding
+// with the first ("pierwsze_czytanie"). Anything unrecognised is "".
+func partID(heading string, czytanieCount *int) string {
+	switch {
+	case strings.HasPrefix(heading, "1. czytanie"):
+		*czytanieCount++
+		if *czytanieCount == 1 {
+			return "pierwsze_czytanie"
+		}
+		return "drugie_czytanie"
+	case strings.HasPrefix(heading, "Psalm"):
+		return "psalm"
+	case strings.HasPrefix(heading, "Aklamacja"):
+		return "aklamacja"
+	case strings.HasPrefix(heading, "Ewangelia"):
+		return "ewangelia"
+	default:
+		return ""
+	}
+}
+
+// ExtractCitation pulls the citation from a section heading:
+// "Ewangelia (Mt 7, 1-5)" -> "Mt 7, 1-5".
+func ExtractCitation(heading string) (string, error) {
+	m := citationRe.FindStringSubmatch(heading)
+	if m == nil {
+		return "", fmt.Errorf("no reference found in heading: %q", heading)
+	}
+	return strings.TrimSpace(m[1]), nil
+}