mirror of
https://github.com/kevinveenbirkenbach/computer-playbook.git
synced 2026-11-09 04:06:42 +00:00
kevinveenbirkenbachandClaude Opus 5.5
de29b99a89
PROTECTED masked identifiers with an underscore and bare numbers, so S3, OAuth2, POP3 and HTML5 went to the engines as a word plus a separately protected digit. Engines tore them ('Amazon S ilə uyğun ...' for 'S3-compatible') or glued the digit onto another word ('Distribuid3-compatible'); the masking audit over core found 159 such entries.
A letter-digit token is now one protected span, ahead of the bare-number alternative. The client masks it whole, and damage.reason() condemns a translation that lost or moved it, so prune and retranslation pick those entries up.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
246 lines
10 KiB
Python
246 lines
10 KiB
Python
"""Unit tests for the criteria that condemn an answer carrying its source or
|
||
coming back in the wrong script.
|
||
|
||
The shape is one the gateway produced across twenty languages: the English
|
||
source, a markdown rule, then a rendering. Every other criterion passed it,
|
||
because the rendering keeps the spans and the names and the separator is made
|
||
of characters the markup comparison does not count.
|
||
"""
|
||
|
||
from __future__ import annotations
|
||
|
||
import unittest
|
||
from typing import ClassVar
|
||
|
||
from utils.i18n.damage import reason
|
||
|
||
PHP_SOURCE = (
|
||
"Shared PHP hardening ini and PHP-FPM pool drop-in that PHP services of "
|
||
"compose and swarm stacks mount as configs"
|
||
)
|
||
STORAGE_SOURCE = "Shared storage for backups and media files"
|
||
STORAGE_RENDERINGS = {
|
||
"de": "Gemeinsamer Speicher für Sicherungen und Mediendateien",
|
||
"hi": "बैकअप और मीडिया फ़ाइलों के लिए साझा भंडारण",
|
||
"ar": "تخزين مشترك للنسخ الاحتياطية وملفات الوسائط",
|
||
"ur": "بیک اپ اور میڈیا فائلوں کے لیے مشترکہ اسٹوریج",
|
||
"vi": "Bộ nhớ dùng chung cho bản sao lưu và tệp đa phương tiện",
|
||
"ja": "バックアップとメディアファイル用の共有ストレージ",
|
||
}
|
||
|
||
|
||
class TestSoundRenderingsPass(unittest.TestCase):
|
||
"""A sound rendering passes every criterion, whatever script its language writes."""
|
||
|
||
def test_every_sound_rendering_is_kept(self) -> None:
|
||
for language, rendering in STORAGE_RENDERINGS.items():
|
||
with self.subTest(language=language):
|
||
self.assertEqual(reason(STORAGE_SOURCE, rendering, language), "")
|
||
|
||
|
||
class TestAddedLines(unittest.TestCase):
|
||
"""A translation breaking lines its source does not break carries text besides the rendering."""
|
||
|
||
def test_a_preamble_line_naming_the_task_is_condemned(self) -> None:
|
||
answer = (
|
||
"Translation of the text to Vietnamese:\n\n"
|
||
"Đã chia sẻ PHP hardening và pool drop-in PHP dịch vụ của tài nguyên "
|
||
"compose và tài nguyên swarm mounts làm cấu hình"
|
||
)
|
||
self.assertEqual(reason(PHP_SOURCE, answer, "vi"), "added-line")
|
||
|
||
def test_a_trailing_note_is_condemned(self) -> None:
|
||
answer = 'Outils\n---\n\nNote: The term "Tools" is translated as "Outils".'
|
||
self.assertEqual(reason("Tools", answer, "bi"), "added-line")
|
||
|
||
def test_a_label_on_a_line_of_its_own_is_condemned(self) -> None:
|
||
self.assertEqual(reason("Added:", "Added:\n\nPřidáno:", "cs"), "added-line")
|
||
|
||
def test_a_line_break_the_source_carries_is_kept(self) -> None:
|
||
self.assertEqual(
|
||
reason("First line\nSecond line", "Erste Zeile\nZweite Zeile", "de"), ""
|
||
)
|
||
|
||
|
||
class TestLetterDigitTokens(unittest.TestCase):
|
||
"""A name that carries its version digit is one token the translation must keep whole."""
|
||
|
||
SOURCE: ClassVar[str] = "S3-compatible object storage shared by the roles"
|
||
|
||
def test_a_token_torn_from_its_digit_is_condemned(self) -> None:
|
||
answer = "Amazon S ilə uyğun obyekt saxlama, rollar arasında paylaşılan"
|
||
self.assertEqual(reason(self.SOURCE, answer, "az"), "protected-span")
|
||
|
||
def test_a_token_moved_onto_another_word_is_condemned(self) -> None:
|
||
answer = "Distribuid3-compatible obyekt stora agaaniin share"
|
||
self.assertEqual(reason(self.SOURCE, answer, "aa"), "protected-span")
|
||
|
||
def test_a_rendering_keeping_the_token_is_kept(self) -> None:
|
||
answer = "S3-kompatibler Objektspeicher, den die Rollen teilen"
|
||
self.assertEqual(reason(self.SOURCE, answer, "de"), "")
|
||
|
||
|
||
class TestWrongNonLatinScript(unittest.TestCase):
|
||
"""A language written in a non-Latin script came back in another non-Latin one."""
|
||
|
||
def test_urdu_written_in_devanagari_is_condemned(self) -> None:
|
||
answer = (
|
||
"Shared PHP हथियार रोजन और PHP-FPM प्लून जो PHP सेवाओं ऑस्टर्म "
|
||
"यूडियोक्सस चालक क्रिप्टोग्राफी इकाइ मount करते हैं"
|
||
)
|
||
self.assertEqual(reason(PHP_SOURCE, answer, "ur"), "wrong-script")
|
||
|
||
def test_hindi_written_in_devanagari_is_kept(self) -> None:
|
||
answer = (
|
||
"साझा PHP सख्त ini और PHP-FPM पूल ड्रॉप-इन कि PHP सेवाओं की रचना और "
|
||
"swarm स्टैक विन्यास के रूप में माउंट"
|
||
)
|
||
self.assertEqual(reason(PHP_SOURCE, answer, "hi"), "")
|
||
|
||
def test_amharic_answered_in_armenian_is_condemned(self) -> None:
|
||
self.assertEqual(
|
||
reason("AI Services", "AI լացոցություներ", "am"), "wrong-script"
|
||
)
|
||
|
||
def test_a_word_in_a_script_the_source_lacks_outweighing_the_own_one_is_condemned(
|
||
self,
|
||
) -> None:
|
||
self.assertEqual(
|
||
reason(
|
||
"Switch the interface to Russian",
|
||
"インターフェースを Русский に切り替える",
|
||
"ja",
|
||
),
|
||
"wrong-script",
|
||
)
|
||
|
||
def test_a_script_the_source_carries_is_not_counted(self) -> None:
|
||
self.assertEqual(
|
||
reason(
|
||
"Switch the interface to Русский",
|
||
"インターフェースを Русский に切り替える",
|
||
"ja",
|
||
),
|
||
"",
|
||
)
|
||
|
||
def test_a_lone_letter_of_another_script_is_not_counted(self) -> None:
|
||
self.assertEqual(reason("alpha", "α", "zh"), "")
|
||
self.assertEqual(
|
||
reason("Gated to alpha and beta lifecycles", "仅限 α 和 β 生命周期", "zh"),
|
||
"",
|
||
)
|
||
|
||
def test_a_latin_script_target_is_not_judged_by_this_rule(self) -> None:
|
||
self.assertEqual(reason("Applications", "Anwendungen", "de"), "")
|
||
|
||
|
||
class TestEchoAndContinue(unittest.TestCase):
|
||
"""A translation holding its whole source is the server failing, not translating."""
|
||
|
||
SOURCE: ClassVar[str] = (
|
||
"Corporate design injection for NGINX-served apps, unifying light and "
|
||
"dark theming from OKLCH design tokens"
|
||
)
|
||
RENDERING: ClassVar[str] = (
|
||
"Corporate-Design-Injection für NGINX-bediente Anwendungen, vereinheitlicht "
|
||
"helles und dunkles Theming aus OKLCH-Design-Tokens"
|
||
)
|
||
|
||
def test_an_answer_that_prepends_its_source_is_condemned(self) -> None:
|
||
answer = f"{self.SOURCE}\n---\n\n{self.RENDERING}"
|
||
self.assertEqual(reason(self.SOURCE, answer, "de"), "echo")
|
||
|
||
def test_an_answer_that_appends_its_source_is_condemned(self) -> None:
|
||
answer = f"{self.RENDERING}\n\n{self.SOURCE}"
|
||
self.assertEqual(reason(self.SOURCE, answer, "de"), "echo")
|
||
|
||
def test_a_rendering_that_drops_the_source_is_kept(self) -> None:
|
||
self.assertEqual(reason(self.SOURCE, self.RENDERING, "de"), "")
|
||
|
||
def test_a_short_source_may_live_inside_a_longer_rendering(self) -> None:
|
||
source = "Mail"
|
||
self.assertEqual(reason(source, "Mail-Konto des Benutzers", "de"), "")
|
||
|
||
|
||
class TestWrongScript(unittest.TestCase):
|
||
"""A language written in another script came back in Latin letters.
|
||
|
||
The catalogs are mostly one- to four-word labels, and a six-word floor
|
||
once stood in front of this criterion, so it never fired on them: a whole
|
||
run filled Belarusian with Polish and Abkhaz with Spanish while every
|
||
criterion passed.
|
||
"""
|
||
|
||
def test_a_one_word_answer_in_the_wrong_script_is_condemned(self) -> None:
|
||
self.assertEqual(reason("Applications", "Aplikacje", "be"), "untranslated")
|
||
|
||
def test_a_label_answered_in_another_language_is_condemned(self) -> None:
|
||
self.assertEqual(reason("Communication", "Comunicación", "ab"), "untranslated")
|
||
|
||
def test_an_answer_in_the_language_s_own_script_is_kept(self) -> None:
|
||
self.assertEqual(reason("Applications", "Праграмы", "be"), "")
|
||
|
||
def test_a_latin_script_target_is_not_judged_by_script(self) -> None:
|
||
self.assertEqual(reason("Applications", "Anwendungen", "de"), "")
|
||
|
||
def test_a_translation_without_a_word_is_kept(self) -> None:
|
||
self.assertEqual(reason("Status", "100 %", "be"), "")
|
||
|
||
def test_a_separatorless_script_counts_one_word_per_character(self) -> None:
|
||
"""Japanese and Chinese write without spaces, so a run-based pattern
|
||
returns the whole sentence as one token, and two Latin product names
|
||
inside it carry that token past a half and the ratio to 1.0."""
|
||
self.assertEqual(
|
||
reason(
|
||
"Runs Playwright in Docker with stable browser settings",
|
||
"安定したブラウザ設定でDockerでPlaywrightを実行",
|
||
"ja",
|
||
),
|
||
"",
|
||
)
|
||
self.assertEqual(
|
||
reason("Run Playwright in Docker", "在Docker中运行Playwright", "zh"), ""
|
||
)
|
||
|
||
def test_a_separatorless_script_answered_in_english_is_still_condemned(
|
||
self,
|
||
) -> None:
|
||
self.assertEqual(
|
||
reason(
|
||
"Runs Playwright in Docker with stable settings",
|
||
"Runs Playwright in Docker with stable browser settings",
|
||
"ja",
|
||
),
|
||
"untranslated",
|
||
)
|
||
|
||
|
||
class TestForeignScript(unittest.TestCase):
|
||
"""A Latin-script language came back in a script it does not use.
|
||
|
||
The model answers a language it cannot write with one it can. Chamorro
|
||
came back in Han because the ISO code ``ch`` reads as Chinese, Turkish in
|
||
Bengali, Avestan in Arabic, and none of it was caught while only the
|
||
non-Latin direction was checked.
|
||
"""
|
||
|
||
def test_han_in_a_latin_script_catalog_is_condemned(self) -> None:
|
||
self.assertEqual(reason("AI Services", "人工智能服务", "ch"), "foreign-script")
|
||
|
||
def test_arabic_in_a_latin_script_catalog_is_condemned(self) -> None:
|
||
self.assertEqual(reason("Sales", "إدارة", "tr"), "foreign-script")
|
||
|
||
def test_a_latin_script_answer_is_kept(self) -> None:
|
||
self.assertEqual(reason("Applications", "Anwendungen", "de"), "")
|
||
|
||
def test_a_non_latin_target_is_not_judged_by_this_rule(self) -> None:
|
||
self.assertEqual(reason("Applications", "Праграмы", "be"), "")
|
||
|
||
def test_a_translation_without_a_word_is_kept(self) -> None:
|
||
self.assertEqual(reason("Status", "100 %", "ch"), "")
|
||
|
||
|
||
if __name__ == "__main__":
|
||
unittest.main()
|