| commit | ed1c89b88454744d0ae08e7048992f80920276d8 | |
| author | Brian Wolff <bawolff+wn@gmail.com> | |
| Thu, 4 Jul 2013 16:19:47 +0000 (4 13:19 -0300) | ||
| committer | Brion Vibber <brion@pobox.com> | |
| Thu, 1 Aug 2013 17:33:04 +0000 (1 10:33 -0700) | ||
| tree | 1dce808d12933ba23088258aa8e4667eb2a75485 | treesnapshot (tar.gz zip) |
| parent | ac4fab6518e090933a25c50267f0e62810ce3475 | commitdiff |
Add Special:RandomInCategory.
The method used is quite biased, but I believe its the best
possible without a schema change and still being efficient.
I think it is good enough for many of the use cases that
need different articles to pop up, but not "real" randomness.
The method used
is to chose a random timestamp and look at cl_timestamp. This method
will give good results if the timestamps are uniformly distributed
(which probably is not usually true). I think it may give acceptable
results in general, especially given most people are not interested
in true randomness, but more in "give me a result I haven't seen before".
(For example, to pick a random entry in a maintenance category to clean
up).
It also fudges the result a little bit using offset to stop really
biased results from happening. This is mostly meant to stop a
category with an extremely clumped distributed from returning
the exact same article every time. It is not meant to generally
increase randomness.
Bug: 25931
Change-Id: I0c48e4a236b50fb627af94f0df47fef8372ea14d
The method used is quite biased, but I believe its the best
possible without a schema change and still being efficient.
I think it is good enough for many of the use cases that
need different articles to pop up, but not "real" randomness.
The method used
is to chose a random timestamp and look at cl_timestamp. This method
will give good results if the timestamps are uniformly distributed
(which probably is not usually true). I think it may give acceptable
results in general, especially given most people are not interested
in true randomness, but more in "give me a result I haven't seen before".
(For example, to pick a random entry in a maintenance category to clean
up).
It also fudges the result a little bit using offset to stop really
biased results from happening. This is mostly meant to stop a
category with an extremely clumped distributed from returning
the exact same article every time. It is not meant to generally
increase randomness.
Bug: 25931
Change-Id: I0c48e4a236b50fb627af94f0df47fef8372ea14d