Verify each finding against the current code and only fix it if needed.

Inline comments:
In `@scripts/generate_site_assets.py`:
- Around line 127-128: The dict construction unnecessarily wraps iterables in
list() before calling sorted(); change the entries that read "languages":
sorted(list(languages)) and "categories": sorted(list(categories)) to simply use
sorted(languages) and sorted(categories) respectively (update the code where the
dict is built so the sorted() calls accept the iterables directly).
- Around line 98-99: The docstring for function process_entity contains a typo
("Worker for Generates assets"); update it to a correct, concise phrasing such
as "Worker that generates assets for a single Country/Market." to fix grammar
and clarity in the process_entity docstring.
- Around line 177-183: The DEV branch currently hardcodes output_dir =
Path("docs/downloads/ics") and ignores the --output flag; modify the block under
if args.dev to use args.output (e.g., set output_dir = Path(args.output) or
respect args.output when present) before calling clean_output_dir(output_dir),
so args.output is honored in DEV mode; keep the rest of the DEV overrides
(target_years, target_countries, target_financial) unchanged and only replace
the hardcoded Path with the args.output-derived Path.
- Around line 35-39: The clean_output_dir(path) function currently calls
shutil.rmtree on a user-supplied path with no validation; update it to validate
the path before deleting by ensuring the resolved path is a subpath of the
project root (or check for a specific sentinel file inside the directory) and
refuse to operate if the path resolves to a filesystem root or home directory
(e.g., path.resolve() == Path("/") or path.resolve() == Path.home()) or is
outside the project directory; if validation fails, raise a clear exception
instead of deleting, otherwise proceed to rmtree and recreate the directory.
- Around line 147-148: Replace the silent "except Exception: pass" blocks with
proper exception logging: catch the exception as e (e.g., "except Exception as
e") and call the module logger (or logger.exception) to record the error plus
contextual identifiers such as entity, year, language in the first block and
subdivision in the second block so you can trace which item failed; after
logging, continue to the next iteration to preserve existing behavior. Ensure
you reference the existing loop variables (entity, year, language, subdivision)
in the log message and use logger.exception or logger.error with traceback for
full diagnostics.
- Around line 217-221: The as_completed loop currently calls future.result()
directly which will re-raise any exceptions from process_entity and crash before
the manifest is written; wrap the future.result() call in a try/except inside
the loop (catch Exception), log the error and the future identity (or the input
entity), and continue so other futures still contribute to manifest; ensure you
do not lose partial results by leaving manifest population as-is and allowing
the final manifest write to run even if some futures failed (or explicitly write
a partial manifest in the except branch) — update the loop around futures and
the exception handling to reference the futures/as_completed loop and
process_entity results accordingly.
- Around line 113-118: Set default_lang from instance.default_language before
computing languages, then change the languages fallback to use that default
rather than hardcoding ["en"]; specifically, move the default_lang assignment
above where languages is defined and replace the languages fallback expression
(instance.supported_languages if instance.supported_languages else ["en"]) with
(instance.supported_languages if instance.supported_languages else
[default_lang]); also remove any hardcoded "Thailand" or "en_US" example logic
related to supported languages so the fallback uniformly uses default_lang;
leave subdiv_codes and categories logic unchanged.

Verify each finding against the current code and only fix it if needed.

Inline comments:
In `@scripts/generate_site_assets.py`:
- Around line 117-118: The loop that calls write_assets is silently skipping
ChildEntity objects when h_obj.subdiv != subdiv; add a diagnostic log inside
that branch to record when a h_obj is skipped including h_obj identifier (e.g.,
h_obj.subdiv or h_obj.code/name), the expected subdiv value, and context (lang,
cat, year_dir) so developers can trace why files weren’t generated; locate the
check around the call to write_assets (the conditional using h_obj.subdiv ==
subdiv) and insert a processLogger or logging.debug/info line that includes
those symbols.

---

Duplicate comments:
In `@scripts/generate_site_assets.py`:
- Around line 67-96: Wrap the entire body of the worker function
(process_entity) in a top-level try/except to prevent uncaught exceptions from
escaping executor.map: catch Exception around all metadata extraction and file
operations (including calls to get_subdivision_aliases(), reading
supported_categories/supported_languages/default_language, and
year_dir.mkdir()), log the error with context (code, name, year) and return a
safe value (e.g., None or a dict indicating failure) so the main executor.map
iteration can continue and the manifest-writing logic still runs; alternatively,
add similar exception handling around the executor.map consumption in main() to
catch worker exceptions, log them, and continue processing.
- Around line 71-73: The fallback language is computed after languages, so
languages uses the hardcoded ["en_US"] instead of the instance.default_language;
compute default_lang first (use getattr(instance, "default_language", "en_US"))
and then set languages = instance.supported_languages if
instance.supported_languages else [default_lang], leaving categories =
instance.supported_categories unchanged and updating any references to
languages/default_lang accordingly.

Verify each finding against the current code and only fix it if needed.

Inline comments:
In `@scripts/generate_site_assets.py`:
- Around line 126-153: The current main() uses
ProcessPoolExecutor().map(process_entity, work_items) but if a worker raises an
uncaught exception the map will propagate and stop iteration losing partial
manifest; wrap the for result in executor.map(...) loop in a try/except to catch
any Exception (or BaseException for extreme cases like MemoryError), log the
error, and continue/flush the manifest (i.e., ensure manifest writing still
runs). Specifically update the loop in main() that iterates over executor.map to
handle and recover from exceptions raised during iteration so
manifest[etype][code] entries already collected are preserved and written out.
- Around line 32-48: The function write_assets currently returns early when
h_obj is falsy, skipping creation of .ics/.json files; instead, handle empty
HolidayBase by always generating files: remove the early return and ensure you
still instantiate ICalExporter(h_obj) and call exporter.save_ics(...) for an
empty calendar, and write an empty-but-valid JSON (e.g. "[]") to (year_dir /
f"{filename_base}.json"). Keep the try/except around exporter.save_ics and the
JSON write so errors are caught, and use filename_base and year_dir as in the
existing code paths.

Verify each finding against the current code and only fix it if needed.

Duplicate comments:
In `@scripts/generate_site_assets.py`:
- Around line 140-144: executor.map will re-raise worker exceptions and abort
the loop, losing a partially-built manifest; change the pattern around
ProcessPoolExecutor so that you submit tasks with
executor.submit(process_entity, item) and iterate over
concurrent.futures.as_completed(futures) to collect results into manifest,
catching and logging exceptions per-future (so MemoryError/pickle errors don't
stop processing) and still write out index.json from the partially-populated
manifest; update references to executor.map -> executor.submit/as_completed,
keep using process_entity and manifest, and ensure any raised exceptions are
logged but do not stop other tasks from being processed.
- Around line 130-132: Before calling shutil.rmtree(OUTPUT_DIR) add a safety
check that verifies OUTPUT_DIR.resolve() is a subpath of the project root and is
not a dangerous absolute path (e.g., "/" or home). Use the project root Path
(e.g., PROJECT_ROOT or derive with Path(__file__).resolve().parents[n]) and
assert OUTPUT_DIR.resolve().is_relative_to(PROJECT_ROOT.resolve()) (or compare
path parts/prefix if Python <3.9), and raise an error or skip removal if the
check fails; only then call shutil.rmtree(OUTPUT_DIR) and proceed to
OUTPUT_DIR.mkdir(...).
- Around line 34-35: The current early return on "if not h_obj: return" silently
skips generating files for combinations with zero holidays; change this so that
when h_obj is empty you instead build and write an empty-but-valid .ics (a
VCALENDAR with no VEVENTs) using the same writer path used for non-empty h_obj
outputs (i.e., the code that writes holiday .ics files), so download links
always have a file; keep the same filename/metadata and only omit VEVENT entries
when h_obj is empty.