--k+w/mQv8wyuph6w0
Content-Type: text/plain; charset=us-ascii



The attached change set (in a ZIP archive) is an update to the
TimeZone goodie which I posted about a week ago.


Change set preamble follows:
---------------------------------------------------------------------
Change Set:		TimeZoneUpdates
Date:			27 December 1999
Author:			David T. Lewis

Incremental changes to TimeZone, to be applied on top of the original
posted change set (TimeZone.21Dec1124am.cs). Includes method name and
comment fixes courtesy of Stan Heckman (stan at stanheckman.com).

Introduced new classes to clarify the distinction between time zone
offsets and leap second offsets. TimeZones now collaborate with
instances of LeapSecondRuleSet to obtain leap second offsets, and
the system TimeZoneDatabase holds an instance of LeapSecondRuleSet
which applies to all time zones. For TimeZones in a system
TimeZoneDatabase, all TimeZones which use leap seconds collaborate
with the single system-wide LeapSecondRuleSet. These changes
also eliminate unnecessary dupication of leap second rule sets.

Moved the tzfile logic out of the TimeZone hierarchy, and put it in a
new TzFileLoader class. TzFileLoader knows how to read tzfile data files
and create instances of TimeZoneRuleSet and LeapSecondRuleSet.

Introduced optimizations in time zone file loading to eliminate duplication
of offset rules. For example, the tzfile data for 'posix/America/Detroit'
is the same as that for 'America/Detroit', so these two equivalent
time zones now can now refer to a single set of time zone offset rules.

Added tree-structured indexing of time zones in the TimeZoneDatabase
to reduce the size of time zone dictionaries, and to make it easier to
browse the system database with an inspector. Time zones can be
directly referenced by name strings such as 'posix/Europe/Paris', but
are stored in the database in a hierarchical tree of Dictionaries
(timeZones->posix->Europe->Paris). The path name separator default
is '/' but can be overridden for purposes of loading data named,
for example, 'posix-Europe-Paris', or for database queries using
names like 'posix%Europe%Paris'. Internally, the time zones have
names like 'posix/Europe/Paris' to encourage consistent time zone
naming.

For a complete time zone database loaded from data files as distributed
on a typical Linux system (for example), the database size is now
reduced by a factor of 3.4 compared to the original posted
version of TimeZone. A full time zone database now occupies 1.8Mb on
disk when copied out on a ReferenceStream (this compresses to under
0.3Mb bytes with gzip). This compares favorably to the size of the
compiled tzfile time zone files, which occupy 1.78Mb on disk.

Changed time zone abbreviations and time zone names to be Symbols
rather than strings. Changed several method names to simplify
expressions. Added more examples and utility methods.


--k+w/mQv8wyuph6w0
Content-Type: text/plain; name="TimeZoneUpdates.zip"
 ; x-mac-type="705A4950"
 ; x-mac-creator="705A4950"
Content-Disposition: attachment; filename="TimeZoneUpdates.zip"
Content-Transfer-Encoding: imap_stub

0,2671,2,18476,0,

--k+w/mQv8wyuph6w0--