1 files changed, 111 insertions, 0 deletions
diff --git a/scripts/parse-sanpi-hyo.py b/scripts/parse-sanpi-hyo.py
new file mode 100644
index 0000000..2030549
--- /dev/null
+++ b/scripts/parse-sanpi-hyo.py
@@ -0,0 +1,111 @@
+import sys
+import json
+import re
+import pdfplumber
+
+sys.stdout.reconfigure(encoding="utf-8")
+
+VOTE_CHARS = set("〇○×✕✖◯")  # 賛成・反対の記号候補
+
+
+def norm(c):
+    """セル文字列を正規化（空白・改行を除去）"""
+    if c is None:
+        return ""
+    return "".join(str(c).split())
+
+
+HEADER_WORDS = {"区分", "番号", "件名", "議決結果", "賛成", "反対", "会派", "議員", "賛否"}
+
+
+def is_vote(c):
+    s = norm(c)
+    return len(s) == 1 and s in VOTE_CHARS
+
+
+def is_name(c):
+    s = norm(c)
+    if not (2 <= len(s) <= 8):
+        return False
+    if is_vote(c) or s in HEADER_WORDS:
+        return False
+    return not re.search(r"[0-9０-９]", s)
+
+
+def is_label(c):
+    s = norm(c)
+    if not (1 <= len(s) <= 5):
+        return False
+    if is_vote(c) or s in HEADER_WORDS:
+        return False
+    return not re.search(r"[0-9０-９]", s)
+
+
+def main(path):
+    members = []      # [(col, 会派, 議員名)]
+    rows = []         # [{区分, 番号, 件名, 議決結果, votes:{col: '〇'/'×'}}]
+    carries = {}      # 区分の繰越
+
+    with pdfplumber.open(path) as pdf:
+        for page in pdf.pages:
+            for table in page.extract_tables():
+                if not table:
+                    continue
+                ncols = max(len(r) for r in table)
+                grid = [list(r) + [None] * (ncols - len(r)) for r in table]
+
+                # 議員名の行＝2〜8文字の名前らしいセルが最も多く並ぶ行
+                best = None
+                for i, row in enumerate(grid[:6]):
+                    n = sum(1 for c in row if is_name(c))
+                    if n >= 5 and (best is None or n > best[1]):
+                        best = (i, n)
+                if best is None:
+                    continue
+                member_row_idx = best[0]
+
+                # 会派名は、その上の行（無ければ同じ行の短いラベル）から拾う
+                label_row = grid[member_row_idx - 1] if member_row_idx > 0 else grid[member_row_idx]
+                local = {}
+                group = ""
+                for c in range(ncols):
+                    lab = norm(label_row[c])
+                    if lab and is_label(label_row[c]):
+                        group = lab
+                    if is_name(grid[member_row_idx][c]):
+                        local[c] = (group, norm(grid[member_row_idx][c]))
+                if local and not members:
+                    members = [(c, g, n) for c, (g, n) in sorted(local.items())]
+
+                # データ行（〇×がある行）
+                for row in grid[member_row_idx + 1:]:
+                    votes = {c: norm(row[c]) for c in local if is_vote(row[c])}
+                    if len(votes) < 2:
+                        continue
+                    vals = [norm(x) for x in row]
+                    disp = vals[0] if vals[0] else carries.get("区分", "")
+                    if vals[0]:
+                        carries["区分"] = vals[0]
+                    rows.append({
+                        "区分": disp,
+                        "番号": next((v for v in vals[1:4] if v and "第" in v), ""),
+                        "件名": next((v for v in vals[1:6] if len(v) >= 8), ""),
+                        "議決結果": next((v for v in vals if v in ("認 定", "認定", "原案可決", "可決", "不採択", "採択", "否決", "原案否決")), ""),
+                        "votes": votes,
+                        "raw": vals[:4],
+                    })
+
+    print("議員（会派 / 名前）:")
+    for c, g, n in members:
+        print(f"  col{c:>3}: {g:<6} {n}")
+    print("\n議案行数:", len(rows))
+    for r in rows[:6]:
+        vs = "".join(r["votes"].get(c, "・") for c, _, _ in members)
+        print(f"  {r['区分'][:6]:<6} {r['番号']:<8} {r['議決結果']:<8} {r['件名'][:28]:<28} {vs}")
+
+    with open(".sanpi-parsed.json", "w", encoding="utf-8") as f:
+        json.dump({"members": members, "rows": rows}, f, ensure_ascii=False, indent=1)
+    print("\n.sanpi-parsed.json に保存しました")
+
+
+main(".tmp-hyo.pdf")