Commit Graph

74 Commits

Author SHA1 Message Date
Fred Drake 14f6c18b62 Give more useful information about a failing PyUnit-style test. 2001-07-16 18:51:32 +00:00
Mark Hammond ef8b654bbe Add support for Windows using "mbcs" as the default Unicode encoding when dealing with the file system. As discussed on python-dev and in patch 410465. 2001-05-13 08:04:26 +00:00
Tim Peters 2f228e75e4 Get rid of the superstitious "~" in dict hashing's "i = (~hash) & mask".
The comment following used to say:
	/* We use ~hash instead of hash, as degenerate hash functions, such
	   as for ints <sigh>, can have lots of leading zeros. It's not
	   really a performance risk, but better safe than sorry.
	   12-Dec-00 tim:  so ~hash produces lots of leading ones instead --
	   what's the gain? */
That is, there was never a good reason for doing it.  And to the contrary,
as explained on Python-Dev last December, it tended to make the *sum*
(i + incr) & mask (which is the first table index examined in case of
collison) the same "too often" across distinct hashes.

Changing to the simpler "i = hash & mask" reduced the number of string-dict
collisions (== # number of times we go around the lookup for-loop) from about
6 million to 5 million during a full run of the test suite (these are
approximate because the test suite does some random stuff from run to run).
The number of collisions in non-string dicts also decreased, but not as
dramatically.

Note that this may, for a given dict, change the order (wrt previous
releases) of entries exposed by .keys(), .values() and .items().  A number
of std tests suffered bogus failures as a result.  For dicts keyed by
small ints, or (less so) by characters, the order is much more likely to be
in increasing order of key now; e.g.,

>>> d = {}
>>> for i in range(10):
...    d[i] = i
...
>>> d
{0: 0, 1: 1, 2: 2, 3: 3, 4: 4, 5: 5, 6: 6, 7: 7, 8: 8, 9: 9}
>>>

Unfortunately. people may latch on to that in small examples and draw a
bogus conclusion.

test_support.py
    Moved test_extcall's sortdict() into test_support, made it stronger,
    and imported sortdict into other std tests that needed it.
test_unicode.py
    Excluced cp875 from the "roundtrip over range(128)" test, because
    cp875 doesn't have a well-defined inverse for unicode("?", "cp875").
    See Python-Dev for excruciating details.
Cookie.py
    Chaged various output functions to sort dicts before building
    strings from them.
test_extcall
    Fiddled the expected-result file.  This remains sensitive to native
    dict ordering, because, e.g., if there are multiple errors in a
    keyword-arg dict (and test_extcall sets up many cases like that), the
    specific error Python complains about first depends on native dict
    ordering.
2001-05-13 00:19:31 +00:00
Barry Warsaw 559f6680c2 In Jython, `@' is not allowed in module names. Extend the TESTFN test
to use "$test" when in Jython.  Closes SF patch #403668.
2001-03-23 18:04:02 +00:00
Fred Drake 84a5934f8a When the regression test is run in verbose mode, make the PyUNIT-based
tests a little noisier, providing more progress information.
2001-03-23 04:21:17 +00:00
Steve Purcell 5ddd1a8dcb Updated to latest PyUnit version (1.31 in PyUnit CVS); test_support.py
changed accordingly.
2001-03-22 08:45:36 +00:00
Fred Drake cd1b1dd6d2 Just import sys at the top instead of inside lots of functions.
Add some helpers for supporting PyUNIT-based unit testing.
2001-03-21 18:26:33 +00:00
Guido van Rossum a8f7e59761 Oops. A RISCOS patch I forgot to check in. 2001-03-13 09:31:07 +00:00
Tim Peters 0009c4ea59 Whitespace normalization. 2001-02-21 07:29:48 +00:00
Jeremy Hylton 4779399e9f Add test for syntax error on "x = 1 + 1".
Move check_syntax() function into test_support.
2001-02-19 15:35:26 +00:00
Skip Montanaro c955f89225 docstring typo 2001-01-20 19:12:54 +00:00
Guido van Rossum a1374e429b Change verify() function to raise TestFailed, not AssertionError.
(I realize that I didn't really test this, because all the tests
succeed, so verify() never raised an AssertionError -- but the test
suite still succeeds, so I'm not too worried.)
2001-01-19 19:01:56 +00:00
Tim Peters 983874dd32 Use constructor form of "raise"; normalize <wink> docstrings. 2001-01-19 05:59:21 +00:00
Tim Peters d2bf3b7ca6 Whitespace normalization. Leaving tokenize_tests.py alone for now. 2001-01-18 02:22:22 +00:00
Marc-André Lemburg 3661908a6a This patch removes all uses of "assert" in the regression test suite
and replaces them with a new API verify(). As a result the regression
suite will also perform its tests in optimization mode.

Written by Marc-Andre Lemburg. Copyright assigned to Guido van Rossum.
2001-01-17 19:11:13 +00:00
Fred Drake 132dce2246 Update the code to better reflect recommended style:
Use != instead of <> since <> is documented as "obsolescent".
Use "is" and "is not" when comparing with None or type objects.
2000-12-12 23:11:42 +00:00
Fred Drake 004d5e6880 Make reindent.py happy (convert everything to 4-space indents!). 2000-10-23 17:22:08 +00:00
Fred Drake 1790dd4b66 Restore TestSkipped exception; appears to have disappeared in last checkin.
Make both TextFailed and TestSkipped subclasses of Error, which derives
from Exception.  Docstrings have been added for the exceptions and module.
2000-07-24 06:55:00 +00:00
Skip Montanaro 4d06923122 make TestFailed a class exception 2000-07-19 17:14:48 +00:00
Guido van Rossum e614fb12a0 Changed runs of 8 spaces to tab -- to satisfy the tab nanny. 1998-06-09 19:20:12 +00:00
Guido van Rossum e26132cf5e Move unified findfile() into test_support.py 1998-04-23 20:13:30 +00:00
Guido van Rossum 531661c1ae Add verbose flag for regression test to clear. 1996-12-20 02:58:22 +00:00
Guido van Rossum 35fb82a33f * os.py: _exit doesn't exist in all variations of posix
* Added fcmp() to test_support.py and use it in test*.py
1993-01-26 13:04:43 +00:00
Guido van Rossum 3bead0984c Initial revision 1992-01-27 17:00:37 +00:00