Initial commit

scraper/bill_id.py

@@ -0,0 +1,75 @@
+"""bill_id codec + session/chamber/number helpers, per spec §4."""
+
+from __future__ import annotations
+
+import re
+from typing import Optional, Tuple
+
+# Universal bill_id pattern across all daily-file links (spec §4).
+# 20252026 0 AB 28 -> 202520260AB28
+BILL_ID_RE = re.compile(
+    r"^(?P<session>\d{8})0?(?P<chamber>AB|SB|ACR|SCR|HR|SJR|ABR)"
+    r"(?P<number>\d+)$"
+)
+
+# Chamber prefix -> origin house (spec §8 #7: origin, not publishing house).
+ASSEMBLY_PREFIXES = {"AB", "ACR", "HR", "ABR"}
+SENATE_PREFIXES = {"SB", "SCR", "SJR"}
+
+_SESSION_RE = re.compile(r"^(?P<a>\d{4})(?P<b>\d{4})$")
+_BILL_DISPLAY_RE = re.compile(r"^(?P<prefix>AB|SB|ACR|SCR|HR|SJR|ABR)\s*(?P<num>\d+)$")
+_SUPPLEMENT_RE = re.compile(r"Supplemental\s+(\d+)", re.IGNORECASE)
+
+
+def origin_house(chamber: str) -> str:
+    if chamber in ASSEMBLY_PREFIXES:
+        return "assembly"
+    if chamber in SENATE_PREFIXES:
+        return "senate"
+    return "unknown"
+
+
+def normalise_session(raw: str) -> str:
+    """'20252026' -> '2025-2026'."""
+    m = _SESSION_RE.match(raw or "")
+    if m:
+        return f"{m['a']}-{m['b']}"
+    return raw
+
+
+def parse_bill_id(bill_id: str) -> Optional[dict]:
+    """Decode a canonical bill_id. Returns dict or None if it does not match.
+
+    Keys: session (YYYY-YYYY), session_raw, chamber (prefix upper), number
+    (int, unpadded), display (e.g. 'AB 28'), house (origin), raw.
+    """
+    if not bill_id:
+        return None
+    m = BILL_ID_RE.match(bill_id.strip())
+    if not m:
+        return None
+    session_raw = m["session"]
+    chamber = m["chamber"].upper()
+    number = int(m["number"])
+    return {
+        "session": normalise_session(session_raw),
+        "session_raw": session_raw,
+        "chamber": chamber,
+        "number": number,
+        "display": f"{chamber} {number}",
+        "house": origin_house(chamber),
+        "raw": bill_id,
+    }
+
+
+def parse_display(display: str) -> Optional[Tuple[str, int]]:
+    """'AB 00028' or 'AB 28' -> ('AB', 28)."""
+    m = _BILL_DISPLAY_RE.match((display or "").strip())
+    if not m:
+        return None
+    return m["prefix"].upper(), int(m["num"])
+
+
+def parse_supplement(text: str) -> Optional[int]:
+    m = _SUPPLEMENT_RE.search(text or "")
+    return int(m.group(1)) if m else None