1. 20 Eki, 2003 1 kayıt (commit)
  2. 29 Ara, 2002 1 kayıt (commit)
  3. 17 Ara, 2002 2 kayıt (commit)
  4. 26 Haz, 2002 1 kayıt (commit)
  5. 26 Nis, 2002 1 kayıt (commit)
  6. 02 Agu, 2001 1 kayıt (commit)
  7. 20 Tem, 2001 1 kayıt (commit)
  8. 13 May, 2001 1 kayıt (commit)
    • Tim Peters's avatar
      Get rid of the superstitious "~" in dict hashing's "i = (~hash) & mask". · 2f228e75
      Tim Peters yazdı
      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.
      2f228e75
  9. 06 Nis, 2001 1 kayıt (commit)
    • Guido van Rossum's avatar
      Since this module already uses doctest-style examples, I figured I'd · 58b6f5b5
      Guido van Rossum yazdı
      add a self-test using doctest.  Results:
      
      - The docstring needs to be a raw string because it uses \"...\".
      
      - The oreo example was broken: the Set-Cookie output doesn't add
        quotes around "doublestuff".
      
      - I had to change the example that prints the class of a Cookie.Cookie
        instance to avoid incorporating an arbitrary object address in the
        test output.
      
      Pretty good score for both doctest and the doc string, I'd say!
      58b6f5b5
  10. 20 Şub, 2001 1 kayıt (commit)
  11. 20 Ock, 2001 1 kayıt (commit)
    • Skip Montanaro's avatar
      added __all__ lists to a number of Python modules · e99d5ea2
      Skip Montanaro yazdı
      added test script and expected output file as well
      this closes patch 103297.
      __all__ attributes will be added to other modules without first submitting
      a patch, just adding the necessary line to the test script to verify
      more-or-less correct implementation.
      e99d5ea2
  12. 14 Ock, 2001 1 kayıt (commit)
  13. 12 Ara, 2000 1 kayıt (commit)
  14. 24 Agu, 2000 2 kayıt (commit)
  15. 19 Agu, 2000 1 kayıt (commit)