From bcbb31e609c1f03f8b66b9e0bc0c27ce9cd7e562 Mon Sep 17 00:00:00 2001
From: Adam Janovsky
Date: Mon, 27 Jan 2025 16:11:32 +0100
Subject: forbid empty PP links in ProtectionProfile objects

---
 src/sec_certs/dataset/protection_profile.py | 18 ++++++++----------
 src/sec_certs/sample/protection_profile.py  | 20 +++++++++++++++-----
 2 files changed, 23 insertions(+), 15 deletions(-)

diff --git a/src/sec_certs/dataset/protection_profile.py b/src/sec_certs/dataset/protection_profile.py
index 1bb2e3f1..4a938d1c 100644
--- a/src/sec_certs/dataset/protection_profile.py
+++ b/src/sec_certs/dataset/protection_profile.py
@@ -226,16 +226,14 @@ class ProtectionProfileDataset(Dataset[ProtectionProfile], ComplexSerializableTy
                 return {}
 
             body = list(tables[0].find_all("tr"))[1:]
-            try:
-                table_certs = {
-                    x.dgst: x
-                    for x in [
-                        ProtectionProfile.from_html_row(row, cert_status, category_string, is_collaborative)
-                        for row in body
-                    ]
-                }
-            except ValueError as e:
-                raise ValueError(f"Bad html file: {file.name} ({str(e)})") from e
+            table_certs = {}
+            for row in body:
+                try:
+                    pp = ProtectionProfile.from_html_row(row, cert_status, category_string, is_collaborative)
+                    table_certs[pp.dgst] = pp
+                except ValueError as e:
+                    logger.error(f"Error when creating ProtectionProfile object: {e}")
+
             return table_certs
 
         cert_status: Literal["active", "archived"] = "active" if "active" in file.name else "archived"
diff --git a/src/sec_certs/sample/protection_profile.py b/src/sec_certs/sample/protection_profile.py
index 36f13130..079c9319 100644
--- a/src/sec_certs/sample/protection_profile.py
+++ b/src/sec_certs/sample/protection_profile.py
@@ -19,7 +19,7 @@ from sec_certs.sample.certificate import Heuristics as BaseHeuristics
 from sec_certs.sample.certificate import PdfData as BasePdfData
 from sec_certs.sample.document_state import DocumentState
 from sec_certs.serialization.json import ComplexSerializableType
-from sec_certs.utils import helpers
+from sec_certs.utils import helpers, sanitization
 
 
 class ProtectionProfile(
@@ -83,18 +83,23 @@ class ProtectionProfile(
                     f"Unexpected number of <td> elements in PP html row. Expected: 6, actual: {len(cells)}"
                 )
 
+            pp_link = cls._html_row_get_link(cells[0])
+            pp_name = cls._html_row_get_name(cells[0])
+            if not sanitization.sanitize_cc_link(pp_link):
+                raise ValueError(f"pp_link for PP {pp_name} is empty, cannot create PP record")
+
             # TODO: Parse maintenance div here. See CC parsing.
             return cls(
                 category,
                 status,
                 False,
-                cls._html_row_get_name(cells[0]),
+                pp_name,
                 cls._html_row_get_version(cells[1]),
                 cls._html_row_get_security_level(cells[2]),
                 cls._html_row_get_date(cells[3]),
                 None if status == "active" else cls._html_row_get_date(cells[4]),
                 cls._html_row_get_link(cells[-1]),
-                cls._html_row_get_link(cells[0]),
+                pp_link,
                 cls._html_row_get_scheme(cells[-2]),
                 [],
             )
@@ -107,17 +112,22 @@ class ProtectionProfile(
                     f"Unexpected number of <td> elements in collaborative PP html row. Expected: 5, actual: {len(cells)}"
                 )
 
+            pp_link = cls._html_row_get_collaborative_pp_link(cells[0])
+            pp_name = cls._html_row_get_collaborative_name(cells[0])
+            if not sanitization.sanitize_cc_link(pp_link):
+                raise ValueError(f"pp_link for PP {pp_name} is empty, cannot create PP record")
+
             return cls(
                 category,
                 "active",
                 True,
-                cls._html_row_get_collaborative_name(cells[0]),
+                pp_name,
                 cls._html_row_get_version(cells[1]),
                 cls._html_row_get_security_level(cells[2]),
                 cls._html_row_get_date(cells[3]),
                 None,
                 cls._html_row_get_link(cells[-1]),
-                cls._html_row_get_collaborative_pp_link(cells[0]),
+                pp_link,
                 None,
                 [],
             )
-- 
cgit v1.3.1