Fix 500 on whitespace-only hadith/chapter text (#158)

[wakqasahmed]

Summary

Fixes #158 (duplicate: #403). Bukhari chapters 89 & 96 (and likely others) returned:

{"error": {"details": "The server encountered an internal error and was unable to complete your request. Either the server is overloaded or there is an error in the application.", "code": 500}}

Root cause

fix_html() in text_transform.py strips the text and, if the result is empty, passes it straight to lxml.html.document_fromstring(). document_fromstring("") raises lxml.etree.ParserError: Document is empty. This is not a Flask HTTPException, so it is never caught by the app's @app.errorhandler(HTTPException) handler and bubbles up as an unhandled 500.

cleanup_text()/cleanup_chapter_title() only short-circuit when the raw field is falsy (None or ""), but a hadith/chapter text field can be a non-empty, whitespace-only string (e.g. a single space or newline). Such a value passes the if not text guard, survives the earlier re.sub calls unchanged, and only becomes empty after fix_html()'s internal .strip() — at which point lxml chokes.

This affects any hadith or chapter row where englishText/arabicText/englishBabName/arabicBabName/intro/ending is whitespace-only, which is the case for at least one row under Bukhari books 89 and 96 in production. This is a code robustness bug, not a data gap — the data is legitimately near-empty, the code should simply produce an empty string rather than crash.

Fix

Return "" early from fix_html() once the stripped text is empty, before calling into lxml.

Reproduction & verification

• Reproduced locally via docker compose up (repo's sample MySQL dataset) by inserting a Bukhari book-89 hadith row with a whitespace-only englishText field, then hitting GET /v1/collections/bukhari/books/89/hadiths — got the exact same 500 / lxml.etree.ParserError: Document is empty traceback reported in the issue.
• Applied the fix, re-hit the same endpoint — now returns 200 with "body": "" for the affected hadith, and unaffected rows are unchanged.
• Removed the test row afterward; no schema/data changes shipped.

Tests

Added tests/test_text_transform.py covering whitespace-only input for all four cleanup functions, plus a sanity check that None/""/normal HTML are unaffected.

docker compose exec web python3 -m unittest tests.test_text_transform -v
# Ran 6 tests in 0.003s - OK

Also ran black --check and flake8 on the changed files — clean.