Read now works, and terraform directory contains example code

extractors/xe_page.py

@@ -28,19 +28,27 @@ def extract_cell_value(td) -> str:
 
 
 def parse_table(table) -> dict:
+    table_id = table.get("id")
     caption_tag = table.find("caption")
     caption = caption_tag.get_text(strip=True) if caption_tag else None
     headers = [th.get_text(strip=True) for th in table.find_all("th", recursive=False)]
     trs = table.find_all("tr", recursive=False)
 
     if headers:
+        # Some pages repeat a label across a hidden raw-value column and a
+        # visible display column (e.g. PortSummary's/MAC Table's two
+        # "Interface"/"MAC Address" headers). Building the row dict via zip()
+        # means the later (visible) column wins on a duplicate label -- that
+        # has matched the raw column's value in every case seen so far, but
+        # if a future page's hidden/visible pair ever genuinely differs,
+        # only the visible one survives here.
         rows = []
         for tr in trs:
             tds = tr.find_all("td", recursive=False)
             if len(tds) != len(headers):
                 continue
             rows.append({h: extract_cell_value(td) for h, td in zip(headers, tds)})
-        return {"caption": caption, "kind": "tabular", "rows": rows}
+        return {"caption": caption, "kind": "tabular", "rows": rows, "id": table_id}
 
     record = {}
     field_names = {}
@@ -58,7 +66,13 @@ def parse_table(table) -> dict:
             field_names[label] = inp["name"]
     # field_names lets a write action look up the real <INPUT NAME=...> for a
     # label instead of hardcoding a guessed "v_R_C_1"-style field name.
-    return {"caption": caption, "kind": "scalar", "record": record, "field_names": field_names}
+    return {
+        "caption": caption,
+        "kind": "scalar",
+        "record": record,
+        "field_names": field_names,
+        "id": table_id,
+    }
 
 
 def parse_xe_tables(html: str) -> list[dict]:
@@ -69,8 +83,18 @@ def parse_xe_tables(html: str) -> list[dict]:
 def fetch_tables(session, path: str) -> dict[str, dict]:
     """Fetch path and return {caption: table} for every captioned table.
 
-    Tables with no/blank caption (usually just a submit button row) are
-    dropped since there's nothing useful to key them by."""
+    Tables with no/blank caption (usually just a submit button row, but
+    occasionally a real data table -- e.g. FDBSearch.html's MAC entries
+    table has no caption) are dropped since there's nothing useful to key
+    them by. Use fetch_all_tables() + table["id"] for those instead."""
     html = switch_client.fetch(session, path)
     tables = parse_xe_tables(html)
     return {t["caption"]: t for t in tables if t["caption"]}
+
+
+def fetch_all_tables(session, path: str) -> list[dict]:
+    """Fetch path and return every table, including uncaptioned ones, in
+    document order. Look up by table["id"] (the <TABLE ID=...> attribute)
+    when a table has no caption to key it by."""
+    html = switch_client.fetch(session, path)
+    return parse_xe_tables(html)