[AdrianAtZyte]

Fixes #1020, fixes #1032

[codspeed]

Merging this PR will not alter performance

✅ 10 untouched benchmarks

---

_{Comparing AdrianAtZyte:reject-dateless-strings (7612ad0) with master (699855a)}

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 99.18%. Comparing base (699855a) to head (7612ad0).

Additional details and impacted files

@@           Coverage Diff           @@
##           master    #1440   +/-   ##
=======================================
  Coverage   99.18%   99.18%           
=======================================
  Files         239      239           
  Lines        4054     4058    +4     
=======================================
+ Hits         4021     4025    +4     
  Misses         33       33           

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[serhii73]

Code review (claude code-review, at e198d57). The full suite passes on this head (24612 passed) and I found no case that was right on master and is wrong here; the strings it changes ('year', 'hour minute', 't') returned today's date before.

Non-blocking notes:

• Because the skipped-word-only parse is now rejected in the parser, the language loop falls through to the next locale, so a single letter can produce a different spurious date: parse('h', settings={'RELATIVE_BASE': datetime(2020,5,5,10)}) gave 2020-05-05 (en) on master and now gives 2020-05-04 via Hungarian h (Monday). Both are wrong, but a check in the applicability or skip-word layer would not change which locale wins.
• all() over an empty filtered_tokens is True, so an empty token list now raises "Nothing to parse" too; please note it or add a test.
• The check compares token strings and ignores token_type, and duplicates the skip logic of the loop below it; a single pass counting non-skipped tokens would be simpler.
• Tests cover year, hour minute, t and one search sentence. The docs text about PARSERS=[custom-formats] with date_formats and the single-letter behaviour across locales have no test.

[serhii73]

I guess we can feel free to merge this PR, @AdrianAtZyte