Port tests/decode.js to Python (issue #2)

Add the full contents of tests/decode.js to tests/test_decode.py:
%xx decoding, invalid-sequence passthrough, the reserved-set variants,
and the complete 36-case binary UTF-8 sample matrix (validity judged
by a strict UTF-8 oracle that mirrors decodeURIComponent). Keep the
pre-existing decode tests unchanged.

Author: cjchanh

tests/test_decode.py

@@ -1,3 +1,8 @@
+import re
+import urllib.parse
+
+import pytest
+
 from mdurl import decode
 
 
@@ -14,3 +19,117 @@ def test_decode_invalid_utf8():
         decode("https://host.invalid/%F1%81%d1%45")
         == "https://host.invalid/\ufffd\ufffd\ufffdE"
     )
+
+
+def _percent_encode(binary):
+    # Mirror the ``encodeBinary`` helper in ``tests/decode.js``: strip all
+    # whitespace, then convert each trailing 8-bit chunk to a two-digit
+    # lowercase hex escape, prepending to the result.
+    binary = re.sub(r"\s+", "", binary)
+    result = ""
+    while binary:
+        chunk = binary[-8:]
+        result = "%" + format(int(chunk, 2), "02x") + result
+        binary = binary[:-8]
+    return result
+
+
+def _decode_uri_component(percent):
+    # Mimic JavaScript ``decodeURIComponent``: strict UTF-8 decoding, which
+    # raises ``UnicodeDecodeError`` on malformed sequences just like
+    # ``decodeURIComponent`` throws URIError on them.
+    return urllib.parse.unquote_to_bytes(percent).decode("utf-8", "strict")
+
+
+# Binary UTF-8 sample matrix ported verbatim from tests/decode.js. As in the
+# JS test, the True/False marks are documentation only: each sample's
+# validity is decided by the strict UTF-8 oracle above (the JS test likewise
+# probes decodeURIComponent directly and never reads the marks).
+SAMPLES = {
+    "00000000": True,
+    "01010101": True,
+    "01111111": True,
+    # invalid as 1st byte
+    "10000000": True,
+    "10111111": True,
+    # invalid sequences, 2nd byte should be >= 0x80
+    "11000111 01010101": False,
+    "11100011 01010101": False,
+    "11110001 01010101": False,
+    # invalid sequences, 2nd byte should be < 0xc0
+    "11000111 11000000": False,
+    "11100011 11000000": False,
+    "11110001 11000000": False,
+    # invalid 3rd byte
+    "11100011 10010101 01010101": False,
+    "11110001 10010101 01010101": False,
+    # invalid 4th byte
+    "11110001 10010101 10010101 01010101": False,
+    # valid sequences
+    "11000111 10101010": True,
+    "11100011 10101010 10101010": True,
+    "11110001 10101010 10101010 10101010": True,
+    # minimal chars with given length
+    "11000010 10000000": True,
+    "11100000 10100000 10000000": True,
+    # impossible sequences
+    "11000001 10111111": False,
+    "11100000 10011111 10111111": False,
+    "11000001 10000000": False,
+    "11100000 10010000 10000000": False,
+    # maximum chars with given length
+    "11011111 10111111": True,
+    "11101111 10111111 10111111": True,
+    "11110000 10010000 10000000 10000000": True,
+    "11110000 10010000 10001111 10001111": True,
+    "11110100 10001111 10110000 10000000": True,
+    "11110100 10001111 10111111 10111111": True,
+    # too low
+    "11110000 10001111 10111111 10111111": False,
+    # too high
+    "11110100 10010000 10000000 10000000": False,
+    "11110100 10011111 10111111 10111111": False,
+    # surrogate range
+    "11101101 10011111 10111111": True,
+    "11101101 10100000 10000000": False,
+    "11101101 10111111 10111111": False,
+    "11101110 10000000 10000000": True,
+}
+
+
+def test_decode_percent():
+    assert decode("x%20xx%20%2520") == "x xx %20"
+
+
+def test_decode_invalid_sequences():
+    assert decode("%2g%z1%%") == "%2g%z1%%"
+
+
+def test_decode_reserved_set_percent():
+    assert decode("%20%25%20", "%") == " %25 "
+
+
+def test_decode_reserved_set_space():
+    assert decode("%20%25%20", " ") == "%20%%20"
+
+
+def test_decode_reserved_set_both():
+    assert decode("%20%25%20", " %") == "%20%25%20"
+
+
+@pytest.mark.parametrize("binary", SAMPLES.keys())
+def test_decode_utf8_sample_matrix(binary):
+    percent = _percent_encode(binary)
+    try:
+        expected = _decode_uri_component(percent)
+        error = None
+    except UnicodeDecodeError:
+        error = True
+
+    result = decode(percent)
+
+    if error:
+        assert "\ufffd" in result
+    else:
+        assert result == expected
+        assert "\ufffd" not in result