"""tools/gen_tables.py writes README.nl.md's brand block, so it has to write that page's counts too.
The script gained a translated README: README.nl.md carries the same generated brand block and short TV
table as README.md, and the count line the block sits under stayed where it was - "316 tv-modellen · 89
merken · 28 talen" over a grid built from 345 lists in 93 brands - because counts() only knew the
English words ("N TV models · B brands · L languages · " and "| **N languages** |"). The page
contradicted itself, one line apart, and nothing read either number: the English card's alt text is held
by test_gen_tables.py, the Dutch one by nothing.
TRANSLATIONS is now {page: the three words that page states its counts in}, so a translated README gets
its own numbers moved with its grid, and this file reads the mapping: each page it names must state those
three words in its card's alt text, one triple.
This file judges that mapping on stand-in checkouts, and the words on this one. Whether the numbers on
disk match examples/*/tv.json is tools/gen_tables.py --check's, run by CI's "README tables" job on main
only - the rule test_examples.py states for the TV tables, for the same reason: a list PR adds its folder
and its tv.json and changes nothing outside it (docs/workflows/tv-list.md, step 5), so the counts lag on
purpose until a maintainer runs the script on main. Holding the page on disk to the checkout's numbers
here turned that lag red on every list PR: one folder added the way the recipe says - debloat.txt and a
tv.json with a new brand - failed test_the_translated_card_and_row_carry_the_numbers_the_examples_give
("'356 tv-modellen · 98 merken · 29 talen · ' != '355 tv-modellen · 97 merken · 29 talen · '") and
test_the_page_on_disk_is_what_the_script_writes, two failures a contributor is told not to fix. The check
itself is not lost: test_a_stale_translated_page_is_what_the_check_reports runs --check against a page
whose numbers did not move, so a translated page that lags is still caught, where it is actionable.
The mapping is also the whole list of pages the script writes, so a translation cannot copy the brand block
by hand and be left out of it: README.de.md did exactly that when the German page landed, and its grid said
"Xiaomi · 32" and "+49 more brands" over 97 brands while README.md said 34 and +53 - the block's own header
says it is generated, and no run touched it. One test here walks the checkout for the markers and holds
every page that carries them to being written.
Run from the repository root: python3 -m unittest discover -s tests -v
"""
import io
import re
import sys
import tempfile
import unittest
from contextlib import redirect_stdout
from pathlib import Path
ROOT = Path(__file__).resolve().parents[1]
sys.path.insert(0, str(ROOT / "tests"))
sys.path.insert(0, str(ROOT / "tools"))
import gen_tables # noqa: E402 (needs tools/ on the path first)
from test_gen_tables import START, END, add, section, tree, tv # noqa: E402
# The Dutch page's card, as README.nl.md has it: the three counts, then the launcher's size (not one of
# them, so it does not move), then the press counts test_press_counts.py reads.
NL = f"""# Titel
Een gewone regel Nederlandse tekst die niet van deze generator is.
| **28 talen** | Volgt de taal van de tv. |
| **Thema's** | Achtergrond en indeling uit een `config.json`. |
{START.format("tv-brands")}
oude merkregel
{END.format("tv-brands")}
{START.format("tv-table")}
| TV | OS |
|---|---|
| oude rij | x |
{END.format("tv-table")}
"""
LISTS = [("acme-a", tv("Acme A", "Acme")), ("acme-b", tv("Acme B", "Acme")), ("bolt-c", tv("Bolt C", "Bolt"))]
# The German page's card, as README.de.md has it: the same three counts in German, then the launcher's size,
# which is not one of them and does not move. No short TV table - that page has the brand block only.
DE = f"""# Titel
Eine gewöhnliche deutsche Zeile, die nicht von diesem Generator stammt.
| **28 Sprachen** | Folgt der Sprache des Fernsehers. |
| **Themes** | Hintergrund und Aufteilung aus einer `config.json`. |
{START.format("tv-brands")}
alte Markenzeile
{END.format("tv-brands")}
"""
def checkout(tmp, dutch: str | None = NL, german: str | None = DE):
"""The fake checkout test_gen_tables.py builds, with the translated READMEs beside its README.md."""
root = tree(tmp, LISTS)
for page, text in (("README.nl.md", dutch), ("README.de.md", german)):
if text is not None:
(root / page).write_text(text, encoding="utf-8")
return root
def numbers(text, words):
"""(models, brands, languages) as `text` states them in `words`, or None."""
found = re.search(rf"(\d+) {re.escape(words[0])} · (\d+) {re.escape(words[1])} · "
rf"(\d+) {re.escape(words[2])} · ", text)
return tuple(int(n) for n in found.groups()) if found else None
def alt_triple(text, words):
"""The card's alt text up to and including the three counts, so a stray second triple cannot match."""
found = re.search(rf'alt="(\d+ {re.escape(words[0])} · \d+ {re.escape(words[1])} · '
rf'\d+ {re.escape(words[2])} · )', text)
return found.group(1) if found else None
def outside(text):
"""The lines of a page that are not inside a generated block, so prose can be compared."""
keeping, inside = [], False
for line in text.split("\n"):
if line.startswith(("", ""):
inside = False
return keeping
class TranslatedCountsTest(unittest.TestCase):
"""A translated page in TRANSLATIONS gets its own numbers; a page that is not is not touched."""
def render(self, root):
return gen_tables.render(root)
def test_the_translated_page_states_this_checkout_s_numbers_in_its_own_words(self):
with tempfile.TemporaryDirectory() as tmp:
root = checkout(tmp)
dutch = self.render(root)["README.nl.md"]
entries = gen_tables.load(root)
self.assertEqual((3, 2, 3), (len(entries), len(gen_tables.brands(entries)),
gen_tables.languages(root)))
self.assertIn("3 tv-modellen · 2 merken · 3 talen · ", dutch)
self.assertIn("| **3 talen** |", dutch)
self.assertNotIn("316", dutch)
self.assertNotIn("28 talen", dutch)
def test_the_numbers_the_generator_does_not_own_stay_where_they_are(self):
# The launcher's size and the Dutch prose are not counts of the examples: a run that rewrote
# them would be editing the translation, which is not what this script is for.
with tempfile.TemporaryDirectory() as tmp:
root = checkout(tmp)
dutch = self.render(root)["README.nl.md"]
self.assertIn("launcher\nvan 523 KB", dutch)
self.assertIn("Een gewone regel Nederlandse tekst die niet van deze generator is.", dutch)
self.assertIn("| **Thema's** | Achtergrond en indeling uit een `config.json`. |", dutch)
def test_only_the_two_count_lines_move_outside_the_generated_blocks(self):
with tempfile.TemporaryDirectory() as tmp:
root = checkout(tmp)
dutch = self.render(root)["README.nl.md"]
before, after = outside(NL), outside(dutch)
self.assertEqual(len(before), len(after), "the block padding changed the page's shape")
moved = [(a, b) for a, b in zip(before, after) if a != b]
self.assertEqual(2, len(moved), f"more than the two count lines changed: {moved}")
self.assertIn("2 merken · 3 talen", moved[0][1])
self.assertTrue(moved[1][1].startswith("| **3 talen** |"))
def test_the_english_page_keeps_the_english_words(self):
with tempfile.TemporaryDirectory() as tmp:
root = checkout(tmp)
readme = self.render(root)["README.md"]
self.assertEqual((3, 2, 3), numbers(readme, ("TV models", "brands", "languages")))
self.assertIn("| **3 languages** |", readme)
self.assertNotIn("tv-modellen", readme)
def test_a_README_nl_that_is_not_there_is_not_written(self):
with tempfile.TemporaryDirectory() as tmp:
root = checkout(tmp, dutch=None)
self.assertNotIn("README.nl.md", self.render(root))
def test_each_translated_page_states_the_numbers_in_its_own_words(self):
# Two pages in the mapping, two sets of words: the Dutch one must not take the German words or the
# other way round, which is what one shared triple would do.
with tempfile.TemporaryDirectory() as tmp:
root = checkout(tmp)
rendered = self.render(root)
self.assertIn("3 tv-modellen · 2 merken · 3 talen · ", rendered["README.nl.md"])
self.assertIn("| **3 talen** |", rendered["README.nl.md"])
self.assertIn("3 TV-Modelle · 2 Marken · 3 Sprachen · ", rendered["README.de.md"])
self.assertIn("| **3 Sprachen** |", rendered["README.de.md"])
self.assertNotIn("Marken", rendered["README.nl.md"])
self.assertNotIn("merken", rendered["README.de.md"])
def test_the_german_page_keeps_what_the_generator_does_not_own(self):
with tempfile.TemporaryDirectory() as tmp:
root = checkout(tmp)
german = self.render(root)["README.de.md"]
self.assertIn("Launcher\nmit 523 KB", german)
self.assertIn("Eine gewöhnliche deutsche Zeile, die nicht von diesem Generator stammt.", german)
self.assertIn("| **Themes** | Hintergrund und Aufteilung aus einer `config.json`. |", german)
def test_a_README_de_that_is_not_there_is_not_written(self):
with tempfile.TemporaryDirectory() as tmp:
root = checkout(tmp, german=None)
rendered = self.render(root)
self.assertNotIn("README.de.md", rendered)
self.assertIn("README.nl.md", rendered)
def test_both_translated_pages_get_the_brand_block_beside_their_counts(self):
with tempfile.TemporaryDirectory() as tmp:
root = checkout(tmp)
rendered = self.render(root)
for page in ("README.nl.md", "README.de.md"):
with self.subTest(page=page):
block = section(rendered[page], "tv-brands")
self.assertIn("[Acme](https://cleanyour.tv/tvs?q=Acme) · 2", " ".join(block))
self.assertNotIn("alte Markenzeile", block)
def test_the_brand_block_is_written_beside_the_counts(self):
# The point of moving the numbers: the grid under them is generated from every tv.json.
with tempfile.TemporaryDirectory() as tmp:
root = checkout(tmp)
dutch = self.render(root)["README.nl.md"]
block = section(dutch, "tv-brands")
self.assertIn("[Acme](https://cleanyour.tv/tvs?q=Acme) · 2", " ".join(block))
self.assertIn("| Acme A | Android TV 11 |", dutch)
class ThisRepositoryTest(unittest.TestCase):
"""README.nl.md on this checkout: the mapping names its words, and a lagging page is what --check reports."""
def test_every_translated_page_the_generator_writes_names_its_three_words(self):
self.assertTrue(gen_tables.TRANSLATIONS, "no translated page is written at all")
for page, words in gen_tables.TRANSLATIONS.items():
with self.subTest(page=page):
self.assertEqual(3, len(words), f"{page}'s counts are stated in three words, not {words}")
self.assertTrue(all(word.strip() for word in words), words)
self.assertEqual(3, len(set(words)), f"{page} uses one word twice: {words}")
text = (ROOT / page).read_text(encoding="utf-8")
self.assertIsNotNone(alt_triple(text, words),
f"{page}'s card alt text does not state {words[0]} · {words[1]} · "
f"{words[2]} in the words TRANSLATIONS gives")
def test_a_stale_translated_page_is_what_the_check_reports(self):
"""A translated page whose numbers did not move is --check's to report, on a checkout of its own.
Not this one: the counts over a generated grid lag on purpose between a list PR and the
maintainer's run of the script on main (see test_examples.py's docstring), so reading them here
turned every list PR red over a file the contributor was told to leave alone.
"""
with tempfile.TemporaryDirectory() as tmp:
root = checkout(tmp)
self.assertIn("README.nl.md", gen_tables.stale(root))
out = io.StringIO()
with redirect_stdout(out):
status = gen_tables.main(["--check", "--root", str(root)])
self.assertEqual(1, status, "a lagging translated page is not an error for --check")
self.assertIn("README.nl.md", out.getvalue(), "the report does not name the translated page")
for path, text in gen_tables.render(root).items():
(root / path).write_text(text, encoding="utf-8")
self.assertNotIn("README.nl.md", gen_tables.stale(root))
with redirect_stdout(io.StringIO()):
self.assertEqual(0, gen_tables.main(["--check", "--root", str(root)]))
def test_every_page_that_carries_the_brand_block_is_written(self):
# The block's header says it is generated, so a page that carries it has to be in the list this
# script writes: README.de.md copied it by hand and was left out, and its grid said "Xiaomi · 32"
# and "+49 more brands" over 97 brands. A translation that copies the block next is held here.
pages, marker = [], "