./wip/p5-MMapDB, Simple database in shared memory

[ CVSweb ] [ Homepage ] [ RSS ] [ Required by ] [ Add to tracker ]


Branch: CURRENT, Version: 0.15, Package name: p5-MMapDB-0.15, Maintainer: wenheping

MMapDB implements a database similar to a hash of hashes of hashes, ...,
of arrays of data.

It's main design goals were:
* very fast read access
* lock-free read access for massive parallelism
* minimal memory consumption per accessing process
* transaction based write access
* simple backup, compactness, one file


Required to run:
[lang/perl5] [wip/p5-File-Map]

Required to build:
[pkgtools/cwrappers]

Master sites: (Expand)

RMD160: 4fdfeb75f0794075110755df091001e70f69ffd8
Filesize: 88.092 KB

Version history: (Expand)


CVS history: (Expand)


   2013-04-25 04:11:21 by Hiramatsu Yoshifumi | Files touched by this commit (2)
Log message:
Update p5-MMapDB to 0.15.

Changes from previsou:
----------------------
0.15  Mon Sep 10 10:37:44 CEST 2012
      - make "map {local $_} values %{$d->main_index}" work
        map creates an alias ($_) and local tries to assign to it.
        Prior this threw an exception.

0.14  Sun Apr  3 20:57:57 CEST 2011
      - index_lookup and index_lookup_position now accept 0 or undef in
      	place of the index parameter. In this case the mainidx is looked up.
	Should be a bit faster than passing $db->mainidx.
      - index_lookup_records(), index_lookup_values(), index_lookup_sorts()
      - speed up writing: mmap _tmpfh and _stringfh to avoid sysseek+syswrite

        v0.13: 20.7612 (19.09 usr +  1.63 sys = 20.72 CPU) @  1.06/s (n=22)
        v0.14: 20.299 (20.20 usr +  0.05 sys = 20.25 CPU) @  1.68/s (n=34)
                Rate v0.13 v0.14
        v0.13 1.06/s    --  -37%
        v0.14 1.68/s   58%    --

0.13  Thu Mar 31 12:27:39 CEST 2011
      - MMDB-lookup tool
      - data_record, data_value and data_sort now work on a position list
        instead of a single position.

0.12  Sun Oct 31 14:22:53 CET 2010
      - fixed subtle UTF8 problem: 2 strings consisting of the same ascii
      	characters were considered different if one of them had the UTF8
	flag set while the other had not.

	Example:

	  no utf8;      # just to make sure
	  Encode::decode_utf8('hello') eq 'hello'

	In Perl this is true. The MMapDB index_lookup function reported them
	as different up to version 0.11.

0.11  Sat Jun  5 16:37:48 CEST 2010
      - removed _rwdata, since File::Map provides protect() now (since 0.21)
      - avoid lvalue substr() in invalidate since this prevents the
        file map to be destroyed in time.

0.10  Fri May 14 13:58:13 CEST 2010
      - make references to the MMapDB object from MMapDB::_base objects
        (tied interface) weak.

0.09  Mon May  3 19:54:25 CEST 2010
      - declare @attributes as 'our'
      - data_value, data_sort accessors
      - DATAMODE_SIMPLE
      - untie in stop() to break circular refs
   2012-10-04 20:42:20 by Aleksej Saushev | Files touched by this commit (38)
Log message:
Drop superfluous PKG_DESTDIR_SUPPORT, "user-destdir" is default these days.
Mark packages that don't or might probably not have staged installation.
   2010-04-21 17:33:55 by Wen Heping | Files touched by this commit (3) | Imported package
Log message:
Import p5-MMapDB-0.08 as wip/p5-MMapDB.

MMapDB implements a database similar to a hash of hashes of hashes, ...,
of arrays of data.

It's main design goals were:
    * very fast read access
    * lock-free read access for massive parallelism
    * minimal memory consumption per accessing process
    * transaction based write access
    * simple backup, compactness, one file