Author tim.peters
Recipients Dennis Sweeney, Zeturic, ammar2, corona10, josh.r, pmpp, serhiy.storchaka, tim.peters, vstinner
Date 2020-10-14.03:03:55
SpamBayes Score -1.0
Marked as misclassified Yes
Message-id <1602644635.56.0.0442839518228.issue41972@roundup.psfhosted.org>
In-reply-to
Content
> There's no discomfort at all to me if, e.g., it stored
> 32-bit counts and is indexed by the last 6 bits of the
> character.  That's a measly 256 bytes in all.

Or, for the same space, 16-bit counts indexed by the last 7 bits. Then there's no aliasing for 7-bit ASCII, which is still very common in my world ;-)  Needles over 64K characters aren't.

Which is a weird rule of thumb that's served me well, although for no solid reason I can detect:  when faced with a universe of tradeoff possibilities for which it appears impossible to get a handle on "the typical" case, optimize for _your_ cases. Then at least one user will be delighted in the end :-)
History
Date User Action Args
2020-10-14 03:03:55tim.peterssetrecipients: + tim.peters, vstinner, pmpp, serhiy.storchaka, josh.r, ammar2, corona10, Dennis Sweeney, Zeturic
2020-10-14 03:03:55tim.peterssetmessageid: <1602644635.56.0.0442839518228.issue41972@roundup.psfhosted.org>
2020-10-14 03:03:55tim.peterslinkissue41972 messages
2020-10-14 03:03:55tim.peterscreate