QUICK DESCRIPTION
Bulk copies data into a Sybase table. Data sources can include a)
another Sybase table, b) the results of any sql, c) the return values
from a perl subroutine called repetitively, or d) a flat file. Comes
with robust error reporting, handling, and intercepting and perl call
back hooks to intercept data before it's sent to the server and
after (if errors.)
Also comes with a command line wrapper, sybxfer.
SYNOPSIS
#!/usr/bin/perl
use Sybase::Xfer;
my %switches = (-from_server=>'CIA', -to_server=>'NSA', -table => 'x-files');
my $h = Sybase::Xfer->new( %switches );
my %status = $h->xfer(-return=>"HASH");
print "xref failed. $status{last_error_msg}\n" unless $status{ok};
#!/bin/ksh
sybxfer -from_server 'CIA' -to_server 'NSA' -table 'x-files'
if [[ $? != 0 ]]; then print "transfer problems"; fi
INSTALLATION (the usual)
perl Makefile.PL [ LIB=<alternate installation> ]
make
make test
make install
DEPENDENCIES
Requires Perl Version 5.005 or beyond
Requires packages:
Sybase::DBlib
CONTACTS
my e-mail: stephen.sprague@morganstanley.com
CHANGE LOG:
-----------------------
Version 0.62 01-MAR-2003 It Lives!!!
-----------------------
* added switch --textsize for use with
bulk inserting text/images.
(rocco ditaranto & jon freeman)
* fixed binary/varbinary/image/timestamp column
xfer bug. (rocco ditaranto)
* added switch --packetsize for potentially
increasing bulk insert performance.
* fixed bug with Sybase::ObjectInfo.pm where
allownulls was always returning true.
(chris anecelle)
* fixed bug with misreporting a float that
underflowed/overflowed was due to non-numeric
data.(Harjit Johal)
* made Xfer.pm warnings (-w) compliant.
(Jun Wu)
* fixed bug with specifying shortcut switches
-T & -D for both source and target.
* removed dependency on DBD::Sybase. Seems it
didn't play nice with Sybase::CTlib under some
conditions.
* fixed bug with xferring identity columns and
requesting the target server to create new ones.
ie. xfer now undef's the corresponding value.
* xfer now explicitly extends the input record
with undef's if it's shorter than the target
table.
pending:
* new switch --method {bcp | insert | sybbcp}
(andrew lowery)
* new switch --filter "{ regex[N] => '[!] /expr/',
firstrow => n,
maxrows => n}"
if regex specified w/o N, then it is applied to each
column of each row. if regex is specifed with N it is
applied to that column only where N is the column number
in target table. (khalid taha)
* db2 support
------------------------
Version 0.53 07-NOV-2001
------------------------
* added "sysUser=>u, sysPwd=>p" to the -truncate
and -drop_and_recreate_indices switches.
This is useful if the bcp needs to be run under a
different user than the truncate/create indices
user. (abhijit subhedar)
* added optimization to -delete, -truncate, and
-auto_delete logic to skip operation if target
table empty. (joe melillo)
* added logic to -from_file s/t if file ends in .gz/.ZIP
to pipe through zcat. (khalid taha)
* added new switch -from_file_pack_format 'format'. Where
format is passed to the unpack command to determine the
fields to transfer to target. This switch is mutex with
-from_file_delimeter. (mark dootson)
------------------------
Version 0.52 10-SEP-2001
------------------------
* corrected more errors in deadlock retry logic (Tom Cornell)
a) {ok} hash status wasn't properly set
b) rows transferred to target incorrectly set
* -from_file can now accept a reference (Graham Ward)
if reference subclassed from IO::File then xfer will
read input from it. Useful for reading from sockets.
------------------------
Version 0.51 15-JUN-2001
------------------------
* added -to_database documentation (Suresh Balakrishnan)
* added use 5.005. (Merijn Broeren)
* added new option -timeout [secs] (Ron Isaacson)
* added new option,
-drop_and_recreate_indices | dari [syts:server.database]
instructs xfer to drop the indices on the target table
before the transfer begins and to recreate them after
the transfer is complete.
* defined scalar return context of method xfer as follows:
0 = success w/o any hitches
>0 = success w/ n number of rows not transferred
-1 = aborted. unrecoverable error
-2 = aborted. interrupt signal
-3 = aborted. timeout signal
* added above scalar return code as 4th element to
list return context. ie. array is as follows:
[0] = num rows read from source
[1] = num rows transferred to target
[2] = last known error message
[3] = scalar return code
* new switch added. -return=> 'HASH' | 'ARRAY' | 'SCALAR'
default is SCALAR if called in scalar context and
ARRAY if called in list context. This is backward
compatible default behavior. However, if HASH is specified
the method will now return:
{rows_read} = number of rows read from target
{rows_transferred} = number of rows transferred to target
{last_error_msg} = last error message encountered
{scalar_return} = scalar return. see above.
{ok} = 1 if all rows were transferred regardless
of retries or warnings along the way.
= 0 if not
* added 'Xfer info' portion to std out. This echoes
-from_table, -to_table and logs pre-xfer activities
like deleting rows, truncating table, dropping indices
etc.
* can now specify -to_table as:
[source.][server.][database.][owner.][table]
where source = 'Sybase' is the only supported source right
now. -to_table settings override -to_server, -to_database
settings.
* fixed double mapping bug when using -from_file_map. Also,
can now use -from_file_map with any source.
* fixed bug with -retry_verbose (Ilya Shulimovich)
* fixed bug when using -from_file and
file had trailing null fields. (Thomas Boccafola)
* rigorous deadlock testing performed by Ilya Shulimovich (i think)
* huge thanks to Ron Isaacson for enlightening me on many
various perl oddities and for volunteering to help me
with more releases of this module. Hell, he's even starting
in own xfer module that'll stretch past this one.
* documentation on error handling expanded.
* added documentation in the bugs section that -to_tables
residing on Sybase Openserver's are not supported for the
target table.
(Peter Somu)